Manuscript under review
Distributional Inverse Reinforcement Learning
Manuscript under review
Feiyang Wu, Ye Zhao, Anqi Wu
2025
@misc{nokey,
title = {Distributional Inverse Reinforcement Learning},
author = {Feiyang Wu and Ye Zhao and Anqi Wu},
year = {2025},
date = {2025-05-23},
urldate = {2025-05-23},
abstract = {We propose a distributional framework for offline Inverse Reinforcement Learning (IRL) that jointly models uncertainty over reward functions and full distribution of returns. Unlike conventional IRL approaches that recover a deterministic reward estimate or match only expected returns, our method captures richer structure in expert behavior by combining distributional reward modeling with Distributional Reinforcement Learning (DistRL) techniques. We formulate reward inference as minimizing first-order stochastic dominance (FSD) violations and integrate distortion risk measures (DRMs) into policy learning, enabling the recovery of both reward distributions and risk-sensitive policies. This formulation is well suited for behavior analysis and risk-aware imitation learning. Empirical results on synthetic benchmarks, real-world dopamine-level neural recordings, and MuJoCo control tasks demonstrate that our method recovers expressive reward representations and achieves state-of-the-art imitation performance.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
We propose a distributional framework for offline Inverse Reinforcement Learning (IRL) that jointly models uncertainty over reward functions and full distribution of returns. Unlike conventional IRL approaches that recover a deterministic reward estimate or match only expected returns, our method captures richer structure in expert behavior by combining distributional reward modeling with Distributional Reinforcement Learning (DistRL) techniques. We formulate reward inference as minimizing first-order stochastic dominance (FSD) violations and integrate distortion risk measures (DRMs) into policy learning, enabling the recovery of both reward distributions and risk-sensitive policies. This formulation is well suited for behavior analysis and risk-aware imitation learning. Empirical results on synthetic benchmarks, real-world dopamine-level neural recordings, and MuJoCo control tasks demonstrate that our method recovers expressive reward representations and achieves state-of-the-art imitation performance.
Manuscript under review
Wei Zhu, Abirath Raju, Abdulaziz Shamsah, Anqi Wu, Seth Hutchinson, Ye Zhao
2025
@misc{zhu_emobipednav_2025,
title = {EmoBipedNav: Emotion-aware Social Navigation for Bipedal Robots with Deep Reinforcement Learning},
author = {Wei Zhu and Abirath Raju and Abdulaziz Shamsah and Anqi Wu and Seth Hutchinson and Ye Zhao},
url = {https://gatech-lidar.github.io/emobipednav.github.io/, project website
http://arxiv.org/abs/2503.12538
https://youtu.be/fNNL56sTSjY?si=gIQ4iqnRF854PYmT},
year = {2025},
date = {2025-03-16},
urldate = {2025-03-16},
publisher = {arXiv},
abstract = {This study presents an emotion-aware navigation framework – EmoBipedNav – using deep reinforcement learning (DRL) for bipedal robots walking in socially interactive environments. The inherent locomotion constraints of bipedal robots challenge their safe maneuvering capabilities in dynamic environments. When combined with the intricacies of social environments, including pedestrian interactions and social cues, such as emotions, these challenges become even more pronounced. To address these coupled problems, we propose a two-stage pipeline that considers both bipedal locomotion constraints and complex social environments. Specifically, social navigation scenarios are represented using sequential LiDAR grid maps (LGMs), from which we extract latent features, including collision regions, emotion-related discomfort zones, social interactions, and the spatio-temporal dynamics of evolving environments. The extracted features are directly mapped to the actions of reduced-order models (ROMs) through a DRL architecture. Furthermore, the proposed framework incorporates full-order dynamics and locomotion constraints during training, effectively accounting for tracking errors and restrictions of the locomotion controller while planning the trajectory with ROMs. Comprehensive experiments demonstrate that our approach exceeds both model-based planners and DRL-based baselines. The hardware videos and open-source code are available at https://gatechlidar.github.io/emobipednav.github.io/.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
This study presents an emotion-aware navigation framework – EmoBipedNav – using deep reinforcement learning (DRL) for bipedal robots walking in socially interactive environments. The inherent locomotion constraints of bipedal robots challenge their safe maneuvering capabilities in dynamic environments. When combined with the intricacies of social environments, including pedestrian interactions and social cues, such as emotions, these challenges become even more pronounced. To address these coupled problems, we propose a two-stage pipeline that considers both bipedal locomotion constraints and complex social environments. Specifically, social navigation scenarios are represented using sequential LiDAR grid maps (LGMs), from which we extract latent features, including collision regions, emotion-related discomfort zones, social interactions, and the spatio-temporal dynamics of evolving environments. The extracted features are directly mapped to the actions of reduced-order models (ROMs) through a DRL architecture. Furthermore, the proposed framework incorporates full-order dynamics and locomotion constraints during training, effectively accounting for tracking errors and restrictions of the locomotion controller while planning the trajectory with ROMs. Comprehensive experiments demonstrate that our approach exceeds both model-based planners and DRL-based baselines. The hardware videos and open-source code are available at https://gatechlidar.github.io/emobipednav.github.io/.
Manuscript under review
Ziyi Zhou, Qian Meng, Hadas Kress-Gazit, Ye Zhao
2025
@misc{zhou_physically-feasible_2025,
title = {Physically-Feasible Reactive Synthesis for Terrain-Adaptive Locomotion via Trajectory Optimization and Symbolic Repair},
author = {Ziyi Zhou and Qian Meng and Hadas Kress-Gazit and Ye Zhao},
url = {http://arxiv.org/abs/2503.03071
https://youtu.be/i2rw1nVTxxY},
year = {2025},
date = {2025-03-01},
urldate = {2025-03-01},
publisher = {arXiv},
abstract = {We propose an integrated planning framework for quadrupedal locomotion over dynamically changing, unforeseen terrains. Existing approaches either rely on heuristics for instantaneous foothold selection–compromising safety and versatility–or solve expensive trajectory optimization problems with complex terrain features and long time horizons. In contrast, our framework leverages reactive synthesis to generate correct-by-construction controllers at the symbolic level, and mixed-integer convex programming (MICP) for dynamic and physically feasible footstep planning for each symbolic transition. We use a high-level manager to reduce the large state space in synthesis by incorporating local environment information, improving synthesis scalability. To handle specifications that cannot be met due to dynamic infeasibility, and to minimize costly MICP solves, we leverage a symbolic repair process to generate only necessary symbolic transitions. During online execution, re-running the MICP with real-world terrain data, along with runtime symbolic repair, bridges the gap between offline synthesis and online execution. We demonstrate, in simulation, our framework’s capabilities to discover missing locomotion skills and react promptly in safety-critical environments, such as scattered stepping stones and rebars.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
We propose an integrated planning framework for quadrupedal locomotion over dynamically changing, unforeseen terrains. Existing approaches either rely on heuristics for instantaneous foothold selection–compromising safety and versatility–or solve expensive trajectory optimization problems with complex terrain features and long time horizons. In contrast, our framework leverages reactive synthesis to generate correct-by-construction controllers at the symbolic level, and mixed-integer convex programming (MICP) for dynamic and physically feasible footstep planning for each symbolic transition. We use a high-level manager to reduce the large state space in synthesis by incorporating local environment information, improving synthesis scalability. To handle specifications that cannot be met due to dynamic infeasibility, and to minimize costly MICP solves, we leverage a symbolic repair process to generate only necessary symbolic transitions. During online execution, re-running the MICP with real-world terrain data, along with runtime symbolic repair, bridges the gap between offline synthesis and online execution. We demonstrate, in simulation, our framework’s capabilities to discover missing locomotion skills and react promptly in safety-critical environments, such as scattered stepping stones and rebars.
Manuscript under review
Feiyang Wu, Xavier Nal, Jaehwi Jang, Wei Zhu, Zhaoyuan Gu, Anqi Wu, Ye Zhao
2025
@misc{wu_learn_2025,
title = {Learn to Teach: Sample-Efficient Privileged Learning for Humanoid Locomotion over Diverse Terrains},
author = {Feiyang Wu and Xavier Nal and Jaehwi Jang and Wei Zhu and Zhaoyuan Gu and Anqi Wu and Ye Zhao},
url = {https://lidar-learn-to-teach.github.io, webpage
http://arxiv.org/abs/2402.06783
https://www.youtube.com/watch?v=pkw4gxOn6Ho
https://github.com/GTLIDAR/IsaacLab/tree/dev},
year = {2025},
date = {2025-03-01},
urldate = {2025-03-01},
publisher = {arXiv},
abstract = {Humanoid robots promise transformative capabilities for industrial and service applications. While recent advances in Reinforcement Learning (RL) yield impressive results in locomotion, manipulation, and navigation, the proposed methods typically require enormous simulation samples to account for real-world variability. This work proposes a novel one-stage training framework—Learn to Teach (L2T)—which unifies teacher and student policy learning. Our approach recycles simulator samples and synchronizes the learning trajectories through shared dynamics, significantly reducing sample complexities and training time while achieving state-of-the-art performance. Furthermore, we validate the RL variant (L2TRL) through extensive simulations and hardware tests on the Digit robot, demonstrating zero-shot sim-to-real transfer and robust performance over 12+ challenging terrains without depth estimation modules. Experimental videos are available online 1.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
Humanoid robots promise transformative capabilities for industrial and service applications. While recent advances in Reinforcement Learning (RL) yield impressive results in locomotion, manipulation, and navigation, the proposed methods typically require enormous simulation samples to account for real-world variability. This work proposes a novel one-stage training framework—Learn to Teach (L2T)—which unifies teacher and student policy learning. Our approach recycles simulator samples and synchronizes the learning trajectories through shared dynamics, significantly reducing sample complexities and training time while achieving state-of-the-art performance. Furthermore, we validate the RL variant (L2TRL) through extensive simulations and hardware tests on the Digit robot, demonstrating zero-shot sim-to-real transfer and robust performance over 12+ challenging terrains without depth estimation modules. Experimental videos are available online 1.
Manuscript under review
Hyunyoung Jung, Zhaoyuan Gu, Ye Zhao, Hae-Won Park, Sehoon Ha
2025
@misc{jung_preci_2025,
title = {PreCi: Pre-training and Continual Improvement of Humanoid Locomotion via Model-Assumption-based Regularization},
author = {Hyunyoung Jung and Zhaoyuan Gu and Ye Zhao and Hae-Won Park and Sehoon Ha},
url = {https://arxiv.org/pdf/2504.09833v1},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
abstract = {Humanoid locomotion is a challenging task due to its inherent complexity and high-dimensional dynamics, as well as the need to adapt to diverse and unpredictable environments. In this work, we introduce a novel learning framework for effectively training a humanoid locomotion policy that imitates the behavior of a model-based controller while extending its capabilities to handle more complex locomotion tasks, such as more challenging terrain and higher velocity commands. Our framework consists of three key components: pre-training through imitation of the model-based controller, fine-tuning via reinforcement learning, and model-assumption-based regularization (MAR) during fine-tuning. In particular, MAR aligns the policy with actions from the model-based controller only in states where the model assumption holds to prevent catastrophic forgetting. We evaluate the proposed framework through comprehensive simulation tests and hardware experiments on a fullsize humanoid robot, Digit, demonstrating a forward speed of 1.5 m/s and robust locomotion across diverse terrains, including slippery, sloped, uneven, and sandy terrains.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
Humanoid locomotion is a challenging task due to its inherent complexity and high-dimensional dynamics, as well as the need to adapt to diverse and unpredictable environments. In this work, we introduce a novel learning framework for effectively training a humanoid locomotion policy that imitates the behavior of a model-based controller while extending its capabilities to handle more complex locomotion tasks, such as more challenging terrain and higher velocity commands. Our framework consists of three key components: pre-training through imitation of the model-based controller, fine-tuning via reinforcement learning, and model-assumption-based regularization (MAR) during fine-tuning. In particular, MAR aligns the policy with actions from the model-based controller only in states where the model assumption holds to prevent catastrophic forgetting. We evaluate the proposed framework through comprehensive simulation tests and hardware experiments on a fullsize humanoid robot, Digit, demonstrating a forward speed of 1.5 m/s and robust locomotion across diverse terrains, including slippery, sloped, uneven, and sandy terrains.
Manuscript under review
Max Asselmeier, Ye Zhao, Patricio A. Vela
2024
@misc{asselmeier_steppability-informed_2024,
title = {Steppability-informed Quadrupedal Contact Planning through Deep Visual Search Heuristics},
author = {Max Asselmeier and Ye Zhao and Patricio A. Vela},
url = {http://arxiv.org/abs/2501.00112},
year = {2024},
date = {2024-12-01},
urldate = {2024-12-01},
publisher = {arXiv},
abstract = {In this work, we introduce a method for predicting environment steppability – the ability of a legged robot platform to place a foothold at a particular location in the local environment – in the image space. This novel environment representation captures this critical geometric property of the local terrain while allowing us to exploit the computational benefits of sensing and planning in the image space. We adapt a primitive shapes-based synthetic data generation scheme to create geometrically rich and diverse simulation scenes and extract ground truth semantic information in order to train a steppability model. We then integrate this steppability model into an existing interleaved graph search and trajectory optimization-based footstep planner to demonstrate how this steppability paradigm can inform footstep planning in complex, unknown environments. We analyze the steppability model performance to demonstrate its validity, and we deploy the perception-informed footstep planner both in offline and online settings to experimentally verify planning performance.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
In this work, we introduce a method for predicting environment steppability – the ability of a legged robot platform to place a foothold at a particular location in the local environment – in the image space. This novel environment representation captures this critical geometric property of the local terrain while allowing us to exploit the computational benefits of sensing and planning in the image space. We adapt a primitive shapes-based synthetic data generation scheme to create geometrically rich and diverse simulation scenes and extract ground truth semantic information in order to train a steppability model. We then integrate this steppability model into an existing interleaved graph search and trajectory optimization-based footstep planner to demonstrate how this steppability paradigm can inform footstep planning in complex, unknown environments. We analyze the steppability model performance to demonstrate its validity, and we deploy the perception-informed footstep planner both in offline and online settings to experimentally verify planning performance.
Manuscript under review
Fukang Liu, Zhaoyuan Gu, Yilin Cai, Ziyi Zhou, Shijie Zhao, Hyunyoung Jung, Sehoon Ha, Yue Chen, Danfei Xu, Ye Zhao
2024
@misc{liu_opt2skill_2024,
title = {Opt2Skill: Imitating Dynamically-feasible Whole-Body Trajectories for Versatile Humanoid Loco-Manipulation},
author = {Fukang Liu and Zhaoyuan Gu and Yilin Cai and Ziyi Zhou and Shijie Zhao and Hyunyoung Jung and Sehoon Ha and Yue Chen and Danfei Xu and Ye Zhao},
url = {https://opt2skill.github.io/, project website
http://arxiv.org/abs/2409.20514
https://youtu.be/DRHfpCYXJfU?si=OYKiitNgiaVy9LMI},
year = {2024},
date = {2024-12-01},
urldate = {2024-12-01},
publisher = {arXiv},
abstract = {Humanoid robots are designed to perform diverse loco-manipulation tasks. However, they face challenges due to their high-dimensional and unstable dynamics, as well as the complex contact-rich nature of the tasks. Model-based optimal control methods offer precise and systematic control but are limited by high computational complexity and accurate contact sensing. On the other hand, reinforcement learning (RL) provides robustness and handles high-dimensional spaces but suffers from inefficient learning, unnatural motion, and sim-to-real gaps. To address these challenges, we introduce Opt2Skill, an end-to-end pipeline that combines model-based trajectory optimization with RL to achieve robust whole-body loco-manipulation. We generate reference motions for the Digit humanoid robot using differential dynamic programming (DDP) and train RL policies to track these trajectories. Our results demonstrate that Opt2Skill outperforms pure RL methods in both training efficiency and task performance, with optimal trajectories that account for torque limits enhancing trajectory tracking. We successfully transfer our approach to real-world applications.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
Humanoid robots are designed to perform diverse loco-manipulation tasks. However, they face challenges due to their high-dimensional and unstable dynamics, as well as the complex contact-rich nature of the tasks. Model-based optimal control methods offer precise and systematic control but are limited by high computational complexity and accurate contact sensing. On the other hand, reinforcement learning (RL) provides robustness and handles high-dimensional spaces but suffers from inefficient learning, unnatural motion, and sim-to-real gaps. To address these challenges, we introduce Opt2Skill, an end-to-end pipeline that combines model-based trajectory optimization with RL to achieve robust whole-body loco-manipulation. We generate reference motions for the Digit humanoid robot using differential dynamic programming (DDP) and train RL policies to track these trajectories. Our results demonstrate that Opt2Skill outperforms pure RL methods in both training efficiency and task performance, with optimal trajectories that account for torque limits enhancing trajectory tracking. We successfully transfer our approach to real-world applications.
Manuscript under review
Zhaoyuan Gu, Yuntian Zhao, Yipu Chen, Rongming Guo, Jennifer K. Leestma, Gregory S. Sawicki, Ye Zhao. Conditionally Accepted by IEEE Transactions on Robotics
2024
@misc{gu_robust-locomotion-by-logic_2024,
title = {Robust-Locomotion-by-Logic: Perturbation-Resilient Bipedal Locomotion via Signal Temporal Logic Guided Model Predictive Control},
author = {Zhaoyuan Gu and Yuntian Zhao and Yipu Chen and Rongming Guo and Jennifer K. Leestma and Gregory S. Sawicki and Ye Zhao. Conditionally Accepted by IEEE Transactions on Robotics},
url = {https://bipedal-stl-mpc.github.io/, project website
http://arxiv.org/abs/2403.15993
https://youtu.be/mhfmr68vW1M?si=OXvFRVqTIQldPlKJ},
year = {2024},
date = {2024-11-01},
urldate = {2024-11-01},
publisher = {arXiv},
abstract = {This study introduces a robust planning framework that utilizes a model predictive control (MPC) approach, enhanced by incorporating signal temporal logic (STL) specifications. This marks the first-ever study to apply STL-guided trajectory optimization for bipedal locomotion, specifically designed to handle both translational and orientational perturbations. Existing recovery strategies often struggle with reasoning complex task logic and evaluating locomotion robustness systematically, making them susceptible to failures caused by inappropriate recovery strategies or lack of robustness. To address these issues, we design an analytical stability metric for bipedal locomotion and quantify this metric using STL specifications, which guide the generation of recovery trajectories to achieve maximum robustness degree. To enable safe and computational-efficient crossed-leg maneuver, we design data-driven self-leg-collision constraints that are 1000 times faster than the traditional inverse-kinematics-based approach. Our framework outperforms a state-of-the-art locomotion controller, a standard MPC without STL, and a linear-temporal-logic-based planner in a high-fidelity dynamic simulation, especially in scenarios involving crossed-leg maneuvers. Additionally, the Cassie bipedal robot achieves robust performance under horizontal and orientational perturbations such as those observed in ship motions. These environments are validated in simulations and deployed on hardware. Furthermore, our proposed method demonstrates versatility on stepping stones and terrain-agnostic features on inclined terrains.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
This study introduces a robust planning framework that utilizes a model predictive control (MPC) approach, enhanced by incorporating signal temporal logic (STL) specifications. This marks the first-ever study to apply STL-guided trajectory optimization for bipedal locomotion, specifically designed to handle both translational and orientational perturbations. Existing recovery strategies often struggle with reasoning complex task logic and evaluating locomotion robustness systematically, making them susceptible to failures caused by inappropriate recovery strategies or lack of robustness. To address these issues, we design an analytical stability metric for bipedal locomotion and quantify this metric using STL specifications, which guide the generation of recovery trajectories to achieve maximum robustness degree. To enable safe and computational-efficient crossed-leg maneuver, we design data-driven self-leg-collision constraints that are 1000 times faster than the traditional inverse-kinematics-based approach. Our framework outperforms a state-of-the-art locomotion controller, a standard MPC without STL, and a linear-temporal-logic-based planner in a high-fidelity dynamic simulation, especially in scenarios involving crossed-leg maneuvers. Additionally, the Cassie bipedal robot achieves robust performance under horizontal and orientational perturbations such as those observed in ship motions. These environments are validated in simulations and deployed on hardware. Furthermore, our proposed method demonstrates versatility on stepping stones and terrain-agnostic features on inclined terrains.
Manuscript under review
Jesse Jiang, Samuel Coogan, Ye Zhao
2024
@misc{jiang_unified_2024,
title = {A Unified Approach to Multi-task Legged Navigation: Temporal Logic Meets Reinforcement Learning},
author = {Jesse Jiang and Samuel Coogan and Ye Zhao},
url = {http://arxiv.org/abs/2407.06931
https://youtu.be/USTBhrJmePM?si=1vM8nJ_V8lQoB2-B},
year = {2024},
date = {2024-07-01},
urldate = {2024-07-01},
publisher = {arXiv},
abstract = {This study examines the problem of hopping robot navigation planning to achieve simultaneous goal-directed and environment exploration tasks. We consider a scenario in which the robot has mandatory goal-directed tasks defined using Linear Temporal Logic (LTL) specifications as well as optional exploration tasks represented using a reward function. Additionally, there exists uncertainty in the robot dynamics which results in motion perturbation. We first propose an abstraction of 3D hopping robot dynamics which enables highlevel planning and a neural-network-based optimization for lowlevel control. We then introduce a Multi-task Product IMDP (MT-PIMDP) model of the system and tasks. We propose a unified control policy synthesis algorithm which enables both task-directed goal-reaching behaviors as well as task-agnostic exploration to learn perturbations and reward. We provide a formal proof of the trade-off induced by prioritizing either LTL or RL actions. We demonstrate our methods with simulation case studies in a 2D world navigation environment.},
keywords = {},
pubstate = {submitted},
tppubtype = {misc}
}
This study examines the problem of hopping robot navigation planning to achieve simultaneous goal-directed and environment exploration tasks. We consider a scenario in which the robot has mandatory goal-directed tasks defined using Linear Temporal Logic (LTL) specifications as well as optional exploration tasks represented using a reward function. Additionally, there exists uncertainty in the robot dynamics which results in motion perturbation. We first propose an abstraction of 3D hopping robot dynamics which enables highlevel planning and a neural-network-based optimization for lowlevel control. We then introduce a Multi-task Product IMDP (MT-PIMDP) model of the system and tasks. We propose a unified control policy synthesis algorithm which enables both task-directed goal-reaching behaviors as well as task-agnostic exploration to learn perturbations and reward. We provide a formal proof of the trade-off induced by prioritizing either LTL or RL actions. We demonstrate our methods with simulation case studies in a 2D world navigation environment.
Journal Articles
Journal Article
Zhaoyuan Gu, Junheng Li, Wenlan Shen, Wenhao Yu, Zhaoming Xie, Stephen McCrory, Xianyi Cheng, Abdulaziz Shamsah, Robert Griffin, C. Karen Liu, Abderrahmane Kheddar, Xue Bin Peng, Yuke Zhu, Guanya Shi, Quan Nguyen, Gordon Cheng, Huijun Gao, Ye Zhao.
IEEE/ASME Transactions on Mechatronics. In Press, 2025
@article{gu_humanoid_2025,
title = {Humanoid Locomotion and Manipulation: Current Progress and Challenges in Control, Planning, and Learning},
author = {Zhaoyuan Gu and Junheng Li and Wenlan Shen and Wenhao Yu and Zhaoming Xie and Stephen McCrory and Xianyi Cheng and Abdulaziz Shamsah and Robert Griffin and C. Karen Liu and Abderrahmane Kheddar and Xue Bin Peng and Yuke Zhu and Guanya Shi and Quan Nguyen and Gordon Cheng and Huijun Gao and Ye Zhao.},
url = {http://arxiv.org/abs/2501.02116},
year = {2025},
date = {2025-05-12},
urldate = {2025-05-12},
journal = {IEEE/ASME Transactions on Mechatronics. In Press},
publisher = {arXiv},
abstract = {Humanoid robots have great potential to perform various human-level skills. These skills involve locomotion, manipulation, and cognitive capabilities. Driven by advances in machine learning and the strength of existing model-based approaches, these capabilities have progressed rapidly, but often separately. Therefore, a timely overview of current progress and future trends in this fast-evolving field is essential. This survey first summarizes the model-based planning and control that have been the backbone of humanoid robotics for the past three decades. We then explore emerging learning-based methods, with a focus on reinforcement learning and imitation learning that enhance the versatility of loco-manipulation skills. We examine the potential of integrating foundation models with humanoid embodiments, assessing the prospects for developing generalist humanoid agents. In addition, this survey covers emerging research for whole-body tactile sensing that unlocks new humanoid skills that involve physical interactions. The survey concludes with a discussion of the challenges and future trends.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Humanoid robots have great potential to perform various human-level skills. These skills involve locomotion, manipulation, and cognitive capabilities. Driven by advances in machine learning and the strength of existing model-based approaches, these capabilities have progressed rapidly, but often separately. Therefore, a timely overview of current progress and future trends in this fast-evolving field is essential. This survey first summarizes the model-based planning and control that have been the backbone of humanoid robotics for the past three decades. We then explore emerging learning-based methods, with a focus on reinforcement learning and imitation learning that enhance the versatility of loco-manipulation skills. We examine the potential of integrating foundation models with humanoid embodiments, assessing the prospects for developing generalist humanoid agents. In addition, this survey covers emerging research for whole-body tactile sensing that unlocks new humanoid skills that involve physical interactions. The survey concludes with a discussion of the challenges and future trends.
Journal Article
Sheeraz Athar, Xinwei Zhang, Jun Ueda, Ye Zhao, Yu She
IEEE Transactions on Haptics. In Press, 2025
@article{10965524,
title = {VibTac: A High-Resolution High-Bandwidth Tactile Sensing Finger for Multi-Modal Perception in Robotic Manipulation},
author = {Sheeraz Athar and Xinwei Zhang and Jun Ueda and Ye Zhao and Yu She},
doi = {10.1109/TOH.2025.3561049},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {IEEE Transactions on Haptics. In Press},
pages = {1-12},
abstract = {Tactile sensing is pivotal for enhancing robot manipulation abilities by providing crucial feedback for localized information. However, existing sensors often lack the necessary resolution and bandwidth required for intricate tasks. To address this gap, we introduce VibTac, a novel multi-modal tactile sensing finger designed to offer high-resolution and high-bandwidth tactile sensing simultaneously. VibTac seamlessly integrates vision-based and vibration-based tactile sensing modes to achieve high-resolution and high-bandwidth tactile sensing respectively, leveraging a streamlined human-inspired design for versatility in tasks. This paper outlines the key design elements of VibTac and its fabrication methods, highlighting the significance of the Elastomer Gel Pad (EGP) in its sensing mechanism. The sensor's multi-modal performance is validated through 3D reconstruction and spectral analysis to discern tactile stimuli effectively. In experimental trials, VibTac demonstrates its efficacy by achieving over 90% accuracy in insertion tasks involving objects emitting distinct sounds, such as ethernet connectors. Leveraging vision-based tactile sensing for object localization and employing a deep learning model for “click” sound classification, VibTac showcases its robustness in real-world scenarios. Video of the sensor working can be accessed at https://youtu.be/kmKIUlXGroo.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Tactile sensing is pivotal for enhancing robot manipulation abilities by providing crucial feedback for localized information. However, existing sensors often lack the necessary resolution and bandwidth required for intricate tasks. To address this gap, we introduce VibTac, a novel multi-modal tactile sensing finger designed to offer high-resolution and high-bandwidth tactile sensing simultaneously. VibTac seamlessly integrates vision-based and vibration-based tactile sensing modes to achieve high-resolution and high-bandwidth tactile sensing respectively, leveraging a streamlined human-inspired design for versatility in tasks. This paper outlines the key design elements of VibTac and its fabrication methods, highlighting the significance of the Elastomer Gel Pad (EGP) in its sensing mechanism. The sensor's multi-modal performance is validated through 3D reconstruction and spectral analysis to discern tactile stimuli effectively. In experimental trials, VibTac demonstrates its efficacy by achieving over 90% accuracy in insertion tasks involving objects emitting distinct sounds, such as ethernet connectors. Leveraging vision-based tactile sensing for object localization and employing a deep learning model for “click” sound classification, VibTac showcases its robustness in real-world scenarios. Video of the sensor working can be accessed at https://youtu.be/kmKIUlXGroo.
Journal Article
Abdulaziz Shamsah, Krishanu Agarwal, Nigam Katta, Abirath Raju, Shreyas Kousik, Ye Zhao
IEEE Transactions on Automation Science and Engineering, 2024
@article{shamsah_socially_2024,
title = {Socially Acceptable Bipedal Robot Navigation via Social Zonotope Network Model Predictive Control},
author = {Abdulaziz Shamsah and Krishanu Agarwal and Nigam Katta and Abirath Raju and Shreyas Kousik and Ye Zhao},
url = {https://szn-mpc.github.io/, project
https://ieeexplore.ieee.org/document/10810741/
https://youtu.be/bUSbsj3cCH4?si=B_thoptPgDu8plNL},
doi = {10.1109/TASE.2024.3519012},
year = {2024},
date = {2024-12-20},
urldate = {2024-12-20},
journal = {IEEE Transactions on Automation Science and Engineering},
volume = {22},
pages = {10130-10148},
abstract = {This study addresses the challenge of social bipedal navigation in a dynamic, human-crowded environment, a research area largely underexplored in legged robot navigation. We present a zonotope-based framework that couples prediction and motion planning for a bipedal ego-agent to account for bidirectional influence with the surrounding pedestrians. This framework incorporates a Social Zonotope Network (SZN), a neural network that predicts future pedestrian reachable sets and plans future socially acceptable reachable set for the ego-agent. SZN generates the reachable sets as zonotopes for efficient reachability-based planning, collision checking, and online uncertainty parameterization. Locomotion-specific losses are added to the SZN training process to adhere to the dynamic limits of the bipedal robot that are not explicitly present in the human crowds data set. These loss functions enable the SZN to generate locomotion paths that are more dynamically feasible for improved tracking. SZN is integrated with a Model Predictive Controller (SZN-MPC) for footstep planning for our bipedal robot Digit. SZN-MPC solves for collision-free trajectory by optimizing through SZN’s gradients. Our results demonstrate the framework’s effectiveness in producing a socially acceptable path, with consistent locomotion velocity, and optimality. The SZN-MPC framework is validated with extensive simulations and hardware experiments.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study addresses the challenge of social bipedal navigation in a dynamic, human-crowded environment, a research area largely underexplored in legged robot navigation. We present a zonotope-based framework that couples prediction and motion planning for a bipedal ego-agent to account for bidirectional influence with the surrounding pedestrians. This framework incorporates a Social Zonotope Network (SZN), a neural network that predicts future pedestrian reachable sets and plans future socially acceptable reachable set for the ego-agent. SZN generates the reachable sets as zonotopes for efficient reachability-based planning, collision checking, and online uncertainty parameterization. Locomotion-specific losses are added to the SZN training process to adhere to the dynamic limits of the bipedal robot that are not explicitly present in the human crowds data set. These loss functions enable the SZN to generate locomotion paths that are more dynamically feasible for improved tracking. SZN is integrated with a Model Predictive Controller (SZN-MPC) for footstep planning for our bipedal robot Digit. SZN-MPC solves for collision-free trajectory by optimizing through SZN’s gradients. Our results demonstrate the framework’s effectiveness in producing a socially acceptable path, with consistent locomotion velocity, and optimality. The SZN-MPC framework is validated with extensive simulations and hardware experiments.
Journal Article
Zhigen Zhao, Shuo Cheng, Yan Ding, Ziyi Zhou, Shiqi Zhang, Danfei Xu, Ye Zhao
IEEE/ASME Transactions on Mechatronics, 2024
@article{zhao_survey_2024,
title = {A Survey of Optimization-Based Task and Motion Planning: From Classical to Learning Approaches},
author = {Zhigen Zhao and Shuo Cheng and Yan Ding and Ziyi Zhou and Shiqi Zhang and Danfei Xu and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10705419/},
doi = {10.1109/TMECH.2024.3452509},
year = {2024},
date = {2024-10-04},
urldate = {2024-10-04},
journal = {IEEE/ASME Transactions on Mechatronics},
pages = {1–27},
abstract = {Task and motion planning (TAMP) integrates high-level task planning and low-level motion planning to equip robots with the autonomy to effectively reason over long-horizon, dynamic tasks. Optimization-based TAMP focuses on hybrid optimization approaches that define goal conditions via objective functions and are capable of handling open-ended goals, robotic dynamics, and physical interaction between the robot and the environment. Therefore, optimization-based TAMP is particularly suited to solve highly complex, contact-rich locomotion and manipulation problems. This survey provides a comprehensive review on optimization-based TAMP, covering first, planning domain representations, including action description languages and temporal logic, second, individual solution strategies for components of TAMP, including AI planning and trajectory optimization (TO), and finally, the dynamic interplay between logic-based task planning and modelbased TO. A particular focus of this survey is to highlight the algorithm structures to efficiently solve TAMP, especially hierarchical and distributed approaches. In addition, the survey emphasizes the synergy between the classical methods and contemporary learning-based innovations, such as large language models. Furthermore, the future research directions for TAMP is discussed in this survey, highlighting both algorithmic and application-specific challenges.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Task and motion planning (TAMP) integrates high-level task planning and low-level motion planning to equip robots with the autonomy to effectively reason over long-horizon, dynamic tasks. Optimization-based TAMP focuses on hybrid optimization approaches that define goal conditions via objective functions and are capable of handling open-ended goals, robotic dynamics, and physical interaction between the robot and the environment. Therefore, optimization-based TAMP is particularly suited to solve highly complex, contact-rich locomotion and manipulation problems. This survey provides a comprehensive review on optimization-based TAMP, covering first, planning domain representations, including action description languages and temporal logic, second, individual solution strategies for components of TAMP, including AI planning and trajectory optimization (TO), and finally, the dynamic interplay between logic-based task planning and modelbased TO. A particular focus of this survey is to highlight the algorithm structures to efficiently solve TAMP, especially hierarchical and distributed approaches. In addition, the survey emphasizes the synergy between the classical methods and contemporary learning-based innovations, such as large language models. Furthermore, the future research directions for TAMP is discussed in this survey, highlighting both algorithmic and application-specific challenges.
Journal Article
Yunhai Han, Kelin Yu, Rahul Batra, Nathan Boyd, Chaitanya Mehta, Tuo Zhao, Yu She, Seth Hutchinson, Ye Zhao
IEEE/ASME Transactions on Mechatronics, 2024
@article{han_learning_2025,
title = {Learning Generalizable Vision-Tactile Robotic Grasping Strategy for Deformable Objects via Transformer},
author = {Yunhai Han and Kelin Yu and Rahul Batra and Nathan Boyd and Chaitanya Mehta and Tuo Zhao and Yu She and Seth Hutchinson and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10552075/
https://github.com/GTLIDAR/DeformableObjectsGrasping
https://www.youtube.com/watch?v=W7o8DsTivTk},
doi = {10.1109/TMECH.2024.3400789},
issn = {1083-4435, 1941-014X},
year = {2024},
date = {2024-06-07},
urldate = {2024-06-07},
journal = {IEEE/ASME Transactions on Mechatronics},
volume = {30},
number = {1},
pages = {554–566},
abstract = {Reliable robotic grasping, especially with deformable objects such as fruits, remains a challenging task due to underactuated contact interactions with a gripper, unknown object dynamics and geometries. In this study, we propose a transformer-based robotic grasping framework for rigid grippers that leverage tactile and visual information for safe object grasping. Specifically, the transformer models learn physical feature embeddings with sensor feedback through performing two predefined explorative actions (pinching and sliding) and predict a grasping outcome through a multilayer perceptron with a given grasping strength. Using these predictions, the gripper predicts a safe grasping strength via inference. Compared with convolutional recurrent networks, the transformer models can capture the long-term dependencies across the image sequences and process spatial–temporal features simultaneously. We first benchmark the transformer models on a public dataset for slip detection. Following that, we show that the transformer models outperform a CNN + LSTM model in terms of grasping accuracy and computational efficiency. We also collect a new fruit grasping dataset and conduct online grasping experiments using the proposed framework for both seen and unseen fruits. In addition, we extend our model to objects with different shapes and demonstrate the effectiveness of our pretrained model trained on our large-scale fruit dataset.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Reliable robotic grasping, especially with deformable objects such as fruits, remains a challenging task due to underactuated contact interactions with a gripper, unknown object dynamics and geometries. In this study, we propose a transformer-based robotic grasping framework for rigid grippers that leverage tactile and visual information for safe object grasping. Specifically, the transformer models learn physical feature embeddings with sensor feedback through performing two predefined explorative actions (pinching and sliding) and predict a grasping outcome through a multilayer perceptron with a given grasping strength. Using these predictions, the gripper predicts a safe grasping strength via inference. Compared with convolutional recurrent networks, the transformer models can capture the long-term dependencies across the image sequences and process spatial–temporal features simultaneously. We first benchmark the transformer models on a public dataset for slip detection. Following that, we show that the transformer models outperform a CNN + LSTM model in terms of grasping accuracy and computational efficiency. We also collect a new fruit grasping dataset and conduct online grasping experiments using the proposed framework for both seen and unseen fruits. In addition, we extend our model to objects with different shapes and demonstrate the effectiveness of our pretrained model trained on our large-scale fruit dataset.
Journal Article
Abdulaziz Shamsah, Zhaoyuan Gu, Jonas Warnke, Seth Hutchinson, Ye Zhao
IEEE Transactions on Robotics, 2023
@article{shamsah_integrated_2023,
title = {Integrated Task and Motion Planning for Safe Legged Navigation in Partially Observable Environments},
author = {Abdulaziz Shamsah and Zhaoyuan Gu and Jonas Warnke and Seth Hutchinson and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10215076/
https://www.youtube.com/watch?v=4nejt0X897E
https://github.com/GTLIDAR/safe-nav-locomotion},
doi = {10.1109/TRO.2023.3299524},
issn = {1552-3098, 1941-0468},
year = {2023},
date = {2023-12-01},
urldate = {2023-12-01},
journal = {IEEE Transactions on Robotics},
volume = {39},
number = {6},
pages = {4913–4934},
abstract = {This study proposes a hierarchically integrated framework for safe task and motion planning (TAMP) of bipedal locomotion in a partially observable environment with dynamic obstacles and uneven terrain. The high-level task planner employs linear temporal logic for a reactive game synthesis between the robot and its environment and provides a formal guarantee on navigation safety and task completion. To address environmental partial observability, a belief abstraction model is designed by partitioning the environment into multiple belief regions and employed at the high-level navigation planner to estimate the dynamic obstacles’ location. This additional location information of dynamic obstacles offered by belief abstraction enables less conservative long-horizon navigation actions beyond guaranteeing immediate collision avoidance. Accordingly, a synthesized action planner sends a set of locomotion actions to the middle-level motion planner while incorporating safe locomotion specifications extracted from safety theorems based on a reduced-order model (ROM) of the locomotion process. The motion planner employs the ROM to design safety criteria and a sampling algorithm to generate nonperiodic motion plans that accurately track high-level actions. At the low level, a foot placement controller based on an angular-momentum linear inverted pendulum model is implemented and integrated with an ankle-actuated passivity-based controller for full-body trajectory tracking. To address external perturbations, this study also investigates the safe sequential composition of the keyframe locomotion state and achieves robust transitions against external perturbations through reachability analysis. The overall TAMP framework is validated with extensive simulations and hardware experiments on bipedal walking robots Cassie and Digit designed by Agility Robotics.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study proposes a hierarchically integrated framework for safe task and motion planning (TAMP) of bipedal locomotion in a partially observable environment with dynamic obstacles and uneven terrain. The high-level task planner employs linear temporal logic for a reactive game synthesis between the robot and its environment and provides a formal guarantee on navigation safety and task completion. To address environmental partial observability, a belief abstraction model is designed by partitioning the environment into multiple belief regions and employed at the high-level navigation planner to estimate the dynamic obstacles’ location. This additional location information of dynamic obstacles offered by belief abstraction enables less conservative long-horizon navigation actions beyond guaranteeing immediate collision avoidance. Accordingly, a synthesized action planner sends a set of locomotion actions to the middle-level motion planner while incorporating safe locomotion specifications extracted from safety theorems based on a reduced-order model (ROM) of the locomotion process. The motion planner employs the ROM to design safety criteria and a sampling algorithm to generate nonperiodic motion plans that accurately track high-level actions. At the low level, a foot placement controller based on an angular-momentum linear inverted pendulum model is implemented and integrated with an ankle-actuated passivity-based controller for full-body trajectory tracking. To address external perturbations, this study also investigates the safe sequential composition of the keyframe locomotion state and achieves robust transitions against external perturbations through reachability analysis. The overall TAMP framework is validated with extensive simulations and hardware experiments on bipedal walking robots Cassie and Digit designed by Agility Robotics.
Journal Article
Lasitha Wijayarathne, Ziyi Zhou, Ye Zhao, Frank L. Hammond
IEEE Transactions on Robotics, 2023
@article{wijayarathne_real-time_2023,
title = {Real-Time Deformable-Contact-Aware Model Predictive Control for Force-Modulated Manipulation},
author = {Lasitha Wijayarathne and Ziyi Zhou and Ye Zhao and Frank L. Hammond},
url = {https://ieeexplore.ieee.org/document/10175188/
https://youtu.be/wuC0Zyr-ZKU
https://github.com/lasithagt/admm},
doi = {10.1109/TRO.2023.3286070},
issn = {1552-3098, 1941-0468},
year = {2023},
date = {2023-10-01},
urldate = {2023-10-01},
journal = {IEEE Transactions on Robotics},
volume = {39},
number = {5},
pages = {3549–3566},
abstract = {The force modulation of robotic manipulators has been extensively studied for several decades. However, it is not yet commonly used in safety-critical applications due to a lack of accurate interaction contact modeling and weak performance guarantees—a large proportion of them concerning the modulation of interaction forces. This study presents a high-level framework for simultaneous trajectory optimization and force control of the interaction between a manipulator and soft environments, which is prone to external disturbances. Sliding friction and normal contact force are taken into account. The dynamics of the soft contact model and the manipulator are simultaneously incorporated in a trajectory optimizer to generate desired motion and force profiles. A constrained optimization framework based on the alternative direction method of multipliers has been employed to efficiently generate real-time optimal control inputs and high-dimensional state trajectories in a model-predictive control fashion. The experimental validation of the model performance is conducted on a soft substrate with known material properties using a Cartesian space force control mode. Results show a comparison of ground truth and real-time model-based contact force and motion tracking for multiple Cartesian motions in the valid range of the friction model. It is shown that a contact-model-based motion planner can compensate for frictional forces and motion disturbances and improve the overall motion and force tracking accuracy. The proposed high-level planner has the potential to facilitate the automation of medical tasks involving the manipulation of compliant, delicate, and deformable tissues.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The force modulation of robotic manipulators has been extensively studied for several decades. However, it is not yet commonly used in safety-critical applications due to a lack of accurate interaction contact modeling and weak performance guarantees—a large proportion of them concerning the modulation of interaction forces. This study presents a high-level framework for simultaneous trajectory optimization and force control of the interaction between a manipulator and soft environments, which is prone to external disturbances. Sliding friction and normal contact force are taken into account. The dynamics of the soft contact model and the manipulator are simultaneously incorporated in a trajectory optimizer to generate desired motion and force profiles. A constrained optimization framework based on the alternative direction method of multipliers has been employed to efficiently generate real-time optimal control inputs and high-dimensional state trajectories in a model-predictive control fashion. The experimental validation of the model performance is conducted on a soft substrate with known material properties using a Cartesian space force control mode. Results show a comparison of ground truth and real-time model-based contact force and motion tracking for multiple Cartesian motions in the valid range of the friction model. It is shown that a contact-model-based motion planner can compensate for frictional forces and motion disturbances and improve the overall motion and force tracking accuracy. The proposed high-level planner has the potential to facilitate the automation of medical tasks involving the manipulation of compliant, delicate, and deformable tissues.
Jesse Jiang, Samuel Coogan, Ye Zhao
IEEE Open Journal of Control Systems, 2023
@article{jiang_abstraction-based_2023,
title = {Abstraction-Based Planning for Uncertainty-Aware Legged Navigation},
author = {Jesse Jiang and Samuel Coogan and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10184473/},
doi = {10.1109/OJCSYS.2023.3296000},
issn = {2694-085X},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {IEEE Open Journal of Control Systems},
volume = {2},
pages = {221–234},
abstract = {This article addresses the problem of temporal-logic-based planning for bipedal robots in uncertain environments. We first propose an Interval Markov Decision Process abstraction of bipedal locomotion (IMDP-BL). Motion perturbations from multiple sources of uncertainty are incorporated into our model using stacked Gaussian process learning in order to achieve formal guarantees on the behavior of the system. We consider tasks which can be specified using Linear Temporal Logic (LTL). Through a product IMDP construction combining the IMDP-BL of the bipedal robot and a Deterministic Rabin Automaton (DRA) of the specifications, we synthesize control policies which allow the robot to safely traverse the environment, iteratively learning the unknown dynamics until the specifications can be satisfied with satisfactory probability. We demonstrate our methods with simulation case studies.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This article addresses the problem of temporal-logic-based planning for bipedal robots in uncertain environments. We first propose an Interval Markov Decision Process abstraction of bipedal locomotion (IMDP-BL). Motion perturbations from multiple sources of uncertainty are incorporated into our model using stacked Gaussian process learning in order to achieve formal guarantees on the behavior of the system. We consider tasks which can be specified using Linear Temporal Logic (LTL). Through a product IMDP construction combining the IMDP-BL of the bipedal robot and a Deterministic Rabin Automaton (DRA) of the specifications, we synthesize control policies which allow the robot to safely traverse the environment, iteratively learning the unknown dynamics until the specifications can be satisfied with satisfactory probability. We demonstrate our methods with simulation case studies.
Ziyi Zhou, Bruce Wingo, Nathan Boyd, Seth Hutchinson, Ye Zhao
IEEE Robotics and Automation Letters, 2022
@article{zhou_momentum-aware_2022,
title = {Momentum-Aware Trajectory Optimization and Control for Agile Quadrupedal Locomotion},
author = {Ziyi Zhou and Bruce Wingo and Nathan Boyd and Seth Hutchinson and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/9804764/
https://www.youtube.com/watch?v=6M78cM0cgCM},
doi = {10.1109/LRA.2022.3185374},
issn = {2377-3766, 2377-3774},
year = {2022},
date = {2022-07-01},
urldate = {2022-07-01},
journal = {IEEE Robotics and Automation Letters},
volume = {7},
number = {3},
pages = {7755–7762},
abstract = {In this letter, we present a versatile hierarchical offline planning algorithm, along with an online control pipeline for agile quadrupedal locomotion. Our offline planner alternates between optimizing centroidal dynamics for a reduced-order model and whole-body trajectory optimization, with the aim of achieving dynamics consensus. Our novel momentum-inertia-aware centroidal optimization, which uses an equimomental ellipsoid parameterization, is able to generate highly acrobatic motions via “inertia shaping”. Our whole-body optimization approach significantly improves upon the quality of standard DDP-based approaches by iteratively exploiting feedback from the centroidal level. For online control, we have developed a novel convex model predictive control scheme through a linear transformation of the full centroidal dynamics. Our controller can efficiently optimize for both contact forces and joint accelerations in single optimization, enabling more straightforward tracking for momentum-rich motions compared to existing quadrupedal MPC controllers. We demonstrate the capability and generality of our trajectory planner on four different dynamic maneuvers. We then present one hardware experiment on the MIT Mini Cheetah platform to demonstrate the performance of the entire planning and control pipeline on a twisting jump maneuver.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this letter, we present a versatile hierarchical offline planning algorithm, along with an online control pipeline for agile quadrupedal locomotion. Our offline planner alternates between optimizing centroidal dynamics for a reduced-order model and whole-body trajectory optimization, with the aim of achieving dynamics consensus. Our novel momentum-inertia-aware centroidal optimization, which uses an equimomental ellipsoid parameterization, is able to generate highly acrobatic motions via “inertia shaping”. Our whole-body optimization approach significantly improves upon the quality of standard DDP-based approaches by iteratively exploiting feedback from the centroidal level. For online control, we have developed a novel convex model predictive control scheme through a linear transformation of the full centroidal dynamics. Our controller can efficiently optimize for both contact forces and joint accelerations in single optimization, enabling more straightforward tracking for momentum-rich motions compared to existing quadrupedal MPC controllers. We demonstrate the capability and generality of our trajectory planner on four different dynamic maneuvers. We then present one hardware experiment on the MIT Mini Cheetah platform to demonstrate the performance of the entire planning and control pipeline on a twisting jump maneuver.
Journal Article
Luke Drnach, John Z. Zhang, Ye Zhao
Frontiers in Robotics and AI, 2022
@article{drnach_mediating_2022,
title = {Mediating Between Contact Feasibility and Robustness of Trajectory Optimization Through Chance Complementarity Constraints},
author = {Luke Drnach and John Z. Zhang and Ye Zhao},
url = {https://www.frontiersin.org/articles/10.3389/frobt.2021.785925/full
https://github.com/GTLIDAR/ChanceConstrainedRobustCITO},
doi = {10.3389/frobt.2021.785925},
issn = {2296-9144},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {Frontiers in Robotics and AI},
volume = {8},
pages = {785925},
abstract = {As robots move from the laboratory into the real world, motion planning will need to account for model uncertainty and risk. For robot motions involving intermittent contact, planning for uncertainty in contact is especially important, as failure to successfully make and maintain contact can be catastrophic. Here, we model uncertainty in terrain geometry and friction characteristics, and combine a risk-sensitive objective with chance constraints to provide a trade-off between robustness to uncertainty and constraint satisfaction with an arbitrarily high feasibility guarantee. We evaluate our approach in two simple examples: a push-block system for benchmarking and a single-legged hopper. We demonstrate that chance constraints alone produce trajectories similar to those produced using strict complementarity constraints; however, when equipped with a robust objective, we show the chance constraints can mediate a trade-off between robustness to uncertainty and strict constraint satisfaction. Thus, our study may represent an important step towards reasoning about contact uncertainty in motion planning.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
As robots move from the laboratory into the real world, motion planning will need to account for model uncertainty and risk. For robot motions involving intermittent contact, planning for uncertainty in contact is especially important, as failure to successfully make and maintain contact can be catastrophic. Here, we model uncertainty in terrain geometry and friction characteristics, and combine a risk-sensitive objective with chance constraints to provide a trade-off between robustness to uncertainty and constraint satisfaction with an arbitrarily high feasibility guarantee. We evaluate our approach in two simple examples: a push-block system for benchmarking and a single-legged hopper. We demonstrate that chance constraints alone produce trajectories similar to those produced using strict complementarity constraints; however, when equipped with a robust objective, we show the chance constraints can mediate a trade-off between robustness to uncertainty and strict constraint satisfaction. Thus, our study may represent an important step towards reasoning about contact uncertainty in motion planning.
Journal Article
Jianwen Luo, Ye Zhao, Lecheng Ruan, Shixin Mao, Chenglong Fu
IEEE Transactions on Automation Science and Engineering, 2022
@article{luo_estimation_2022,
title = {Estimation of CoM and CoP Trajectories During Human Walking Based on a Wearable Visual Odometry Device},
author = {Jianwen Luo and Ye Zhao and Lecheng Ruan and Shixin Mao and Chenglong Fu},
url = {https://ieeexplore.ieee.org/document/9265451/},
doi = {10.1109/TASE.2020.3036530},
issn = {1545-5955, 1558-3783},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {IEEE Transactions on Automation Science and Engineering},
volume = {19},
number = {1},
pages = {396–409},
abstract = {Estimation of center of mass (CoM) and center of pressure (CoP) is critical for lower limb exoskeletons, prostheses, and legged robots. To meet the demand in these fields, this study presents a novel CoM and CoP estimation method for human walking through a wearable visual odometry (VO) device. This method is named VO-based estimation of CoM and CoP (VOECC). The methodology of VOECC is that the VO provides CoM trajectory estimation and the inherent walking dynamics model is exploited as prior knowledge for CoP trajectory estimation during human walking. Gait cycle is estimated based on the frequency analysis of the CoM trajectory, which is cropped into segments. Each segment mainly includes a half gait cycle. The segments are designed to be sliding to mitigate the disturbance of double-stance phase. For each segment, a quadratic programming (QP) problem is formulated to fit the CoM measurement with the theoretical walking dynamics model. The solution to this QP problem is an optimal gait parameters estimation, including CoP. Based on this solution, the human walking model with the CoM trajectory and CoP excursion is reconstructed. VOECC is evaluated experimentally where human walks on level ground and upstairs with VO device attached in front of the chest. The ground truth of CoM and CoP position is directly measured by the motion capture system and fully instrumented treadmill, respectively, and compared with the VOECC results. The proposed method is demonstrated to be effective in terms of wearable and extensible functionalities compared with the existing methods. Root-mean-squared errors between the CoP measured by fully instrumented treadmill and the CoP estimated by VOECC are evaluated and compared. This method has the potential to be extensible in lower limb rehabilitation, prosthetic, and legged locomotion fields. Note to Practitioners—This article addresses the problem of estimating center of mass (CoM) and center of pressure (CoP) trajectories using a minimum number of wearable sensors and reliable algorithms during human daily walking. Estimation of CoM and CoP trajectories is critical for lower limb exoskeletons, prostheses, and legged robots. In this study, a novel method named VO-based estimation of CoM and CoP (VOECC) is presented that utilizes a walking model as prior knowledge and integrates it with visual odometry data, which estimates the trajectory of the wearable visual device. Compared with the inertia measurement unit (IMU)-based method, VOECC only uses one wearable visual device and thus significantly reduces the cost and system complexity. In addition, the VOECC outperforms the motion capture system and force plate since it is not limited to space constraints and has the potential to be applicable for daily life locomotion tasks. VOECC is wearable and untethered and therefore can be directly amounted on lower limb exoskeletons, prostheses, and legged robots. In the experiments, motion capture system and force plates are used to measure CoM and CoP as ground truth to demonstrate the effectiveness of the proposed VOECC. Practical limitations include failure from fast turning and synchronization of multichannel sensors. These limitations will be addressed as our future research directions.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Estimation of center of mass (CoM) and center of pressure (CoP) is critical for lower limb exoskeletons, prostheses, and legged robots. To meet the demand in these fields, this study presents a novel CoM and CoP estimation method for human walking through a wearable visual odometry (VO) device. This method is named VO-based estimation of CoM and CoP (VOECC). The methodology of VOECC is that the VO provides CoM trajectory estimation and the inherent walking dynamics model is exploited as prior knowledge for CoP trajectory estimation during human walking. Gait cycle is estimated based on the frequency analysis of the CoM trajectory, which is cropped into segments. Each segment mainly includes a half gait cycle. The segments are designed to be sliding to mitigate the disturbance of double-stance phase. For each segment, a quadratic programming (QP) problem is formulated to fit the CoM measurement with the theoretical walking dynamics model. The solution to this QP problem is an optimal gait parameters estimation, including CoP. Based on this solution, the human walking model with the CoM trajectory and CoP excursion is reconstructed. VOECC is evaluated experimentally where human walks on level ground and upstairs with VO device attached in front of the chest. The ground truth of CoM and CoP position is directly measured by the motion capture system and fully instrumented treadmill, respectively, and compared with the VOECC results. The proposed method is demonstrated to be effective in terms of wearable and extensible functionalities compared with the existing methods. Root-mean-squared errors between the CoP measured by fully instrumented treadmill and the CoP estimated by VOECC are evaluated and compared. This method has the potential to be extensible in lower limb rehabilitation, prosthetic, and legged locomotion fields. Note to Practitioners—This article addresses the problem of estimating center of mass (CoM) and center of pressure (CoP) trajectories using a minimum number of wearable sensors and reliable algorithms during human daily walking. Estimation of CoM and CoP trajectories is critical for lower limb exoskeletons, prostheses, and legged robots. In this study, a novel method named VO-based estimation of CoM and CoP (VOECC) is presented that utilizes a walking model as prior knowledge and integrates it with visual odometry data, which estimates the trajectory of the wearable visual device. Compared with the inertia measurement unit (IMU)-based method, VOECC only uses one wearable visual device and thus significantly reduces the cost and system complexity. In addition, the VOECC outperforms the motion capture system and force plate since it is not limited to space constraints and has the potential to be applicable for daily life locomotion tasks. VOECC is wearable and untethered and therefore can be directly amounted on lower limb exoskeletons, prostheses, and legged robots. In the experiments, motion capture system and force plates are used to measure CoM and CoP as ground truth to demonstrate the effectiveness of the proposed VOECC. Practical limitations include failure from fast turning and synchronization of multichannel sensors. These limitations will be addressed as our future research directions.
Journal Article
Ye Zhao, Yinan Li, Luis Sentis, Ufuk Topcu, Jun Liu
The International Journal of Robotics Research, 2022
@article{zhao_reactive_2022,
title = {Reactive Task and Motion Planning for Robust Whole-Body Dynamic Locomotion in Constrained Environments},
author = {Ye Zhao and Yinan Li and Luis Sentis and Ufuk Topcu and Jun Liu},
url = {https://doi.org/10.1177/02783649221077714
https://www.youtube.com/watch?v=BdxYCmhRIMg},
doi = {https://doi.org/10.1177/02783649221077714},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {The International Journal of Robotics Research},
volume = {41},
number = {8},
pages = {812–847},
abstract = {Contact-based decision and planning methods are becoming increasingly important to endow higher levels of autonomy for legged robots. Formal synthesis methods derived from symbolic systems have great potential for reasoning about high-level locomotion decisions and achieving complex maneuvering behaviors with correctness guarantees. This study takes a first step toward formally devising an architecture composed of task planning and control of whole-body dynamic locomotion behaviors in constrained and dynamically changing environments. At the high level, we formulate a two-player temporal logic game between the multi-limb locomotion planner and its dynamic environment to synthesize a winning strategy that delivers symbolic locomotion actions. These locomotion actions satisfy the desired high-level task specifications expressed in a fragment of temporal logic. Those actions are sent to a robust finite transition system that synthesizes a locomotion controller that fulfills state reachability constraints. This controller is further executed via a low-level motion planner that generates feasible locomotion trajectories. We construct a set of dynamic locomotion models for legged robots to serve as a template library for handling diverse environmental events. We devise a replanning strategy that takes into consideration sudden environmental changes or large state disturbances to increase the robustness of the resulting locomotion behaviors. We formally prove the correctness of the layered locomotion framework guaranteeing a robust implementation by the motion planning layer. Simulations of reactive locomotion behaviors in diverse environments indicate that our framework has the potential to serve as a theoretical foundation for intelligent locomotion behaviors.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Contact-based decision and planning methods are becoming increasingly important to endow higher levels of autonomy for legged robots. Formal synthesis methods derived from symbolic systems have great potential for reasoning about high-level locomotion decisions and achieving complex maneuvering behaviors with correctness guarantees. This study takes a first step toward formally devising an architecture composed of task planning and control of whole-body dynamic locomotion behaviors in constrained and dynamically changing environments. At the high level, we formulate a two-player temporal logic game between the multi-limb locomotion planner and its dynamic environment to synthesize a winning strategy that delivers symbolic locomotion actions. These locomotion actions satisfy the desired high-level task specifications expressed in a fragment of temporal logic. Those actions are sent to a robust finite transition system that synthesizes a locomotion controller that fulfills state reachability constraints. This controller is further executed via a low-level motion planner that generates feasible locomotion trajectories. We construct a set of dynamic locomotion models for legged robots to serve as a template library for handling diverse environmental events. We devise a replanning strategy that takes into consideration sudden environmental changes or large state disturbances to increase the robustness of the resulting locomotion behaviors. We formally prove the correctness of the layered locomotion framework guaranteeing a robust implementation by the motion planning layer. Simulations of reactive locomotion behaviors in diverse environments indicate that our framework has the potential to serve as a theoretical foundation for intelligent locomotion behaviors.
Jesse Jiang, Ye Zhao, Samuel Coogan
IEEE Control Systems Letters, 2022
@article{jiang_safe_2022,
title = {Safe Learning for Uncertainty-Aware Planning via Interval MDP Abstraction},
author = {Jesse Jiang and Ye Zhao and Samuel Coogan},
url = {http://arxiv.org/abs/2202.01358
https://github.com/GTLIDAR/SafeLearningForPlanning},
doi = {10.1109/LCSYS.2022.3173993},
issn = {2475-1456},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {IEEE Control Systems Letters},
volume = {6},
pages = {2641–2646},
abstract = {We study the problem of refining satisfiability bounds for partially-known stochastic systems against planning specifications defined using syntactically co-safe Linear Temporal Logic (scLTL). We propose an abstraction-based approach that iteratively generates high-confidence Interval Markov Decision Process (IMDP) abstractions of the system from high-confidence bounds on the unknown component of the dynamics obtained via Gaussian process regression. In particular, we develop a synthesis strategy to sample the unknown dynamics by finding paths which avoid specification-violating states using a product IMDP. We further provide a heuristic to choose among various candidate paths to maximize the information gain. Finally, we propose an iterative algorithm to synthesize a satisfying control policy for the product IMDP system. We demonstrate our work with a case study on mobile robot navigation.},
note = {arXiv:2202.01358 [eess]},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
We study the problem of refining satisfiability bounds for partially-known stochastic systems against planning specifications defined using syntactically co-safe Linear Temporal Logic (scLTL). We propose an abstraction-based approach that iteratively generates high-confidence Interval Markov Decision Process (IMDP) abstractions of the system from high-confidence bounds on the unknown component of the dynamics obtained via Gaussian process regression. In particular, we develop a synthesis strategy to sample the unknown dynamics by finding paths which avoid specification-violating states using a product IMDP. We further provide a heuristic to choose among various candidate paths to maximize the information gain. Finally, we propose an iterative algorithm to synthesize a satisfying control policy for the product IMDP system. We demonstrate our work with a case study on mobile robot navigation.
Journal Article
Lixian Zhang, Ruixian Zhang, Tong Wu, Rui Weng, Minghao Han, Ye Zhao
IEEE Transactions on Neural Networks and Learning Systems, 2021
@article{zhang_safe_2021,
title = {Safe Reinforcement Learning With Stability Guarantee for Motion Planning of Autonomous Vehicles},
author = {Lixian Zhang and Ruixian Zhang and Tong Wu and Rui Weng and Minghao Han and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/9478933/},
doi = {10.1109/TNNLS.2021.3084685},
issn = {2162-237X, 2162-2388},
year = {2021},
date = {2021-12-01},
urldate = {2025-04-02},
journal = {IEEE Transactions on Neural Networks and Learning Systems},
volume = {32},
number = {12},
pages = {5435–5444},
abstract = {Reinforcement learning with safety constraints is promising for autonomous vehicles, of which various failures may result in disastrous losses. In general, a safe policy is trained by constrained optimization algorithms, in which the average constraint return as a function of states and actions should be lower than a predefined bound. However, most existing safe learning-based algorithms capture states via multiple high-precision sensors, which complicates the hardware systems and is power-consuming. This article is focused on safe motion planning with the stability guarantee for autonomous vehicles with limited size and power. To this end, the risk-identification method and the Lyapunov function are integrated with the well-known soft actor–critic (SAC) algorithm. By borrowing the concept of Lyapunov functions in the control theory, the learned policy can theoretically guarantee that the state trajectory always stays in a safe area. A novel risk-sensitive learning-based algorithm with the stability guarantee is proposed to train policies for the motion planning of autonomous vehicles. The learned policy is implemented on a differential drive vehicle in a simulation environment. The experimental results show that the proposed algorithm achieves a higher success rate than the SAC.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Reinforcement learning with safety constraints is promising for autonomous vehicles, of which various failures may result in disastrous losses. In general, a safe policy is trained by constrained optimization algorithms, in which the average constraint return as a function of states and actions should be lower than a predefined bound. However, most existing safe learning-based algorithms capture states via multiple high-precision sensors, which complicates the hardware systems and is power-consuming. This article is focused on safe motion planning with the stability guarantee for autonomous vehicles with limited size and power. To this end, the risk-identification method and the Lyapunov function are integrated with the well-known soft actor–critic (SAC) algorithm. By borrowing the concept of Lyapunov functions in the control theory, the learned policy can theoretically guarantee that the state trajectory always stays in a safe area. A novel risk-sensitive learning-based algorithm with the stability guarantee is proposed to train policies for the motion planning of autonomous vehicles. The learned policy is implemented on a differential drive vehicle in a simulation environment. The experimental results show that the proposed algorithm achieves a higher success rate than the SAC.
Journal Article
Hongwu Zhu, Dong Wang, Nathan Boyd, Ziyi Zhou, Lecheng Ruan, Aidong Zhang, Ning Ding, Ye Zhao, Jianwen Luo
Frontiers in Robotics and AI, 2021
@article{zhu_terrain-perception-free_2021,
title = {Terrain-Perception-Free Quadrupedal Spinning Locomotion on Versatile Terrains: Modeling, Analysis, and Experimental Validation},
author = {Hongwu Zhu and Dong Wang and Nathan Boyd and Ziyi Zhou and Lecheng Ruan and Aidong Zhang and Ning Ding and Ye Zhao and Jianwen Luo},
url = {https://www.frontiersin.org/articles/10.3389/frobt.2021.724138/full},
doi = {10.3389/frobt.2021.724138},
issn = {2296-9144},
year = {2021},
date = {2021-10-01},
urldate = {2025-04-02},
journal = {Frontiers in Robotics and AI},
volume = {8},
pages = {724138},
abstract = {Dynamic quadrupedal locomotion over rough terrains reveals remarkable progress over the last few decades. Small-scale quadruped robots are adequately flexible and adaptable to traverse uneven terrains along the sagittal direction, such as slopes and stairs. To accomplish autonomous locomotion navigation in complex environments, spinning is a fundamental yet indispensable functionality for legged robots. However, spinning behaviors of quadruped robots on uneven terrain often exhibit position drifts. Motivated by this problem, this study presents an algorithmic method to enable accurate spinning motions over uneven terrain and constrain the spinning radius of the center of mass (CoM) to be bounded within a small range to minimize the drift risks. A modified spherical foot kinematics representation is proposed to improve the foot kinematic model and rolling dynamics of the quadruped during locomotion. A CoM planner is proposed to generate a stable spinning motion based on projected stability margins. Accurate motion tracking is accomplished with linear quadratic regulator (LQR) to bind the position drift during the spinning movement. Experiments are conducted on a small-scale quadruped robot and the effectiveness of the proposed method is verified on versatile terrains including flat ground, stairs, and slopes.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Dynamic quadrupedal locomotion over rough terrains reveals remarkable progress over the last few decades. Small-scale quadruped robots are adequately flexible and adaptable to traverse uneven terrains along the sagittal direction, such as slopes and stairs. To accomplish autonomous locomotion navigation in complex environments, spinning is a fundamental yet indispensable functionality for legged robots. However, spinning behaviors of quadruped robots on uneven terrain often exhibit position drifts. Motivated by this problem, this study presents an algorithmic method to enable accurate spinning motions over uneven terrain and constrain the spinning radius of the center of mass (CoM) to be bounded within a small range to minimize the drift risks. A modified spherical foot kinematics representation is proposed to improve the foot kinematic model and rolling dynamics of the quadruped during locomotion. A CoM planner is proposed to generate a stable spinning motion based on projected stability margins. Accurate motion tracking is accomplished with linear quadratic regulator (LQR) to bind the position drift during the spinning movement. Experiments are conducted on a small-scale quadruped robot and the effectiveness of the proposed method is verified on versatile terrains including flat ground, stairs, and slopes.
Jianwen Luo, Zelin Gong, Yao Su, Lecheng Ruan, Ye Zhao, H. Harry Asada, Chenglong Fu
IEEE Robotics and Automation Letters, 2021
@article{luo_modeling_2021,
title = {Modeling and Balance Control of Supernumerary Robotic Limb for Overhead Tasks},
author = {Jianwen Luo and Zelin Gong and Yao Su and Lecheng Ruan and Ye Zhao and H. Harry Asada and Chenglong Fu},
url = {https://ieeexplore.ieee.org/document/9384151/},
doi = {10.1109/LRA.2021.3067850},
issn = {2377-3766, 2377-3774},
year = {2021},
date = {2021-04-01},
urldate = {2021-04-01},
journal = {IEEE Robotics and Automation Letters},
volume = {6},
number = {2},
pages = {4125–4132},
abstract = {Overhead manipulation tasks often require collaborations between two operators, which becomes challenging in confined spaces such as in a compartment. Supernumerary Robotic Limb (SuperLimb), as a promising wearable robotic solution, can provide assistance in terms of broader workspace, wider manipulation functionalities and safer working conditions. However, the safety concerns of human-centered SuperLimb interaction mechanisms are rarely studied to date, particularly regarding human standing balance. This study proposes a balance controller by which one individual operator can accomplish overhead tasks with the assistance of SuperLimb via tunable interaction force and supporting force regulation. The SuperLimb-human interaction is modeled and a dynamics control method based on QR decomposition (also known as QR factorization, in which a matrix is factorized into an orthogonal matrix and an upper triangular matrix) is adopted to decouple joint torques of the SuperLimb and the interaction forces. Therefore, the supporting forces can be regulated independently to guarantee the operator-SuperLimb interaction forces in a safe region. Force plate is used for measuring the CoP position as an evaluation method of the standing balance. The critical horizontal push force is learned through experiment to guide the balance controller. This method is implemented on a SuperLimb prototype worn on the operator's back, to provide necessary supporting forces on overhead object while allowing the operator to move freely underneath.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Overhead manipulation tasks often require collaborations between two operators, which becomes challenging in confined spaces such as in a compartment. Supernumerary Robotic Limb (SuperLimb), as a promising wearable robotic solution, can provide assistance in terms of broader workspace, wider manipulation functionalities and safer working conditions. However, the safety concerns of human-centered SuperLimb interaction mechanisms are rarely studied to date, particularly regarding human standing balance. This study proposes a balance controller by which one individual operator can accomplish overhead tasks with the assistance of SuperLimb via tunable interaction force and supporting force regulation. The SuperLimb-human interaction is modeled and a dynamics control method based on QR decomposition (also known as QR factorization, in which a matrix is factorized into an orthogonal matrix and an upper triangular matrix) is adopted to decouple joint torques of the SuperLimb and the interaction forces. Therefore, the supporting forces can be regulated independently to guarantee the operator-SuperLimb interaction forces in a safe region. Force plate is used for measuring the CoP position as an evaluation method of the standing balance. The critical horizontal push force is learned through experiment to guide the balance controller. This method is implemented on a SuperLimb prototype worn on the operator's back, to provide necessary supporting forces on overhead object while allowing the operator to move freely underneath.
Journal Article
Luke Drnach, Ye Zhao
IEEE Robotics and Automation Letters, 2021
@article{drnach_robust_2021,
title = {Robust Trajectory Optimization Over Uncertain Terrain With Stochastic Complementarity},
author = {Luke Drnach and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/9343725/
https://youtu.be/IdHXPEV1iRw
https://github.com/GTLIDAR/RobustContactERM},
doi = {10.1109/LRA.2021.3056064},
issn = {2377-3766, 2377-3774},
year = {2021},
date = {2021-04-01},
urldate = {2021-04-01},
journal = {IEEE Robotics and Automation Letters},
volume = {6},
number = {2},
pages = {1168–1175},
abstract = {Trajectory optimization with contact-rich behaviors has recently gained attention for generating diverse locomotion behaviors without pre-specified ground contact sequences. However, these approaches rely on precise models of robot dynamics and the terrain and are susceptible to uncertainty. Recent works have attempted to handle uncertainties in the system model, but few have investigated uncertainty in contact dynamics. In this letter, we model uncertainty stemming from the terrain and design corresponding risk-sensitive objectives for contact-implicit trajectory optimization. In particular, we parameterize uncertainties from the terrain contact distance and friction coefficients using probability distributions and propose a corresponding expected residual minimization cost approach. We evaluate our method in three simple robotic examples, including a legged hopping robot, and we benchmark one of our examples in simulation against a robust worst-case solution. We show that our risk-sensitive method produces contact-averse trajectories that are robust to terrain perturbations. Moreover, we demonstrate that the resulting trajectories converge to those generated by a traditional, non-robust method as the terrain model becomes more certain. Our study marks an important step towards a fully robust, contact-implicit approach suitable for deploying robots on real-world terrain.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Trajectory optimization with contact-rich behaviors has recently gained attention for generating diverse locomotion behaviors without pre-specified ground contact sequences. However, these approaches rely on precise models of robot dynamics and the terrain and are susceptible to uncertainty. Recent works have attempted to handle uncertainties in the system model, but few have investigated uncertainty in contact dynamics. In this letter, we model uncertainty stemming from the terrain and design corresponding risk-sensitive objectives for contact-implicit trajectory optimization. In particular, we parameterize uncertainties from the terrain contact distance and friction coefficients using probability distributions and propose a corresponding expected residual minimization cost approach. We evaluate our method in three simple robotic examples, including a legged hopping robot, and we benchmark one of our examples in simulation against a robust worst-case solution. We show that our risk-sensitive method produces contact-averse trajectories that are robust to terrain perturbations. Moreover, we demonstrate that the resulting trajectories converge to those generated by a traditional, non-robust method as the terrain model becomes more certain. Our study marks an important step towards a fully robust, contact-implicit approach suitable for deploying robots on real-world terrain.
Journal Article
Zhigen Zhao, Ziyi Zhou, Michael Park, Ye Zhao
IEEE Access, 2021
@article{zhao_sydebo_2021,
title = {SyDeBO: Symbolic-Decision-Embedded Bilevel Optimization for Long-Horizon Manipulation in Dynamic Environments},
author = {Zhigen Zhao and Ziyi Zhou and Michael Park and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/9537786/},
doi = {10.1109/ACCESS.2021.3112879},
issn = {2169-3536},
year = {2021},
date = {2021-01-01},
urldate = {2025-04-02},
journal = {IEEE Access},
volume = {9},
pages = {128817–128826},
abstract = {This study proposes a Task and Motion Planning (TAMP) method with symbolic decisions embedded in a bilevel optimization. This TAMP method exploits the discrete structure of sequential manipulation for long-horizon and versatile tasks in dynamic environments. At the symbolic planning level, we propose a scalable decision-making method for long-horizon manipulation tasks using the Planning Domain Definition Language (PDDL) with causal graph decomposition. At the motion planning level, we devise a trajectory optimization (TO) approach based on the Differential Dynamic Programming (DDP) and Alternating Direction Method of Multipliers (ADMM), suitable for solving constrained, large-scale nonlinear optimization in a distributed manner. Distinct from conventional geometric motion planners, our approach generates highly dynamic manipulation motions by incorporating the full robot and object dynamics. Furthermore, in lieu of a hierarchical planning approach, we solve a holistically integrated bilevel optimization problem involving costs from both the low-level TO and the high-level search. Simulation and experimental results demonstrate dynamic manipulation for long-horizon object sorting tasks in clutter and on a moving conveyor belt.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study proposes a Task and Motion Planning (TAMP) method with symbolic decisions embedded in a bilevel optimization. This TAMP method exploits the discrete structure of sequential manipulation for long-horizon and versatile tasks in dynamic environments. At the symbolic planning level, we propose a scalable decision-making method for long-horizon manipulation tasks using the Planning Domain Definition Language (PDDL) with causal graph decomposition. At the motion planning level, we devise a trajectory optimization (TO) approach based on the Differential Dynamic Programming (DDP) and Alternating Direction Method of Multipliers (ADMM), suitable for solving constrained, large-scale nonlinear optimization in a distributed manner. Distinct from conventional geometric motion planners, our approach generates highly dynamic manipulation motions by incorporating the full robot and object dynamics. Furthermore, in lieu of a hierarchical planning approach, we solve a holistically integrated bilevel optimization problem involving costs from both the low-level TO and the high-level search. Simulation and experimental results demonstrate dynamic manipulation for long-horizon object sorting tasks in clutter and on a moving conveyor belt.
Ye Zhao, Yan Gu
International Journal of Intelligent Robotics and Applications, 2020
@article{zhao_non-periodic_2020,
title = {A non-periodic planning and control framework of dynamic legged locomotion},
author = {Ye Zhao and Yan Gu},
url = {http://link.springer.com/10.1007/s41315-020-00122-7},
doi = {10.1007/s41315-020-00122-7},
issn = {2366-5971, 2366-598X},
year = {2020},
date = {2020-03-01},
urldate = {2025-04-08},
journal = {International Journal of Intelligent Robotics and Applications},
volume = {4},
number = {1},
pages = {95–108},
abstract = {This study proposes an integrated planning and control framework for achieving three-dimensional robust and dynamic legged locomotion over uneven terrain. The proposed framework is composed of three hierarchical layers. The high-level layer is a state-space motion planner designing highly dynamic locomotion behaviors based on a reduced-order robot model. This motion planner incorporates two robust bundles, named as invariant and recoverability bundles, which quantify analytical state-space deviations for robust planning design. The low-level layer is a modelbased trajectory tracking controller capable of robustly realizing the planned locomotion behaviors. This controller is synthesized based on full-order hybrid dynamic modeling, model-based state feedback control, and Lyapunov stability analysis. The planning and control layers are concatenated by a middle-level trajectory generator that produces nominal behaviors for a full-order robot model. The proposed framework is validated through flat and uneven terrain walking simulations of a three-dimensional bipedal robot.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study proposes an integrated planning and control framework for achieving three-dimensional robust and dynamic legged locomotion over uneven terrain. The proposed framework is composed of three hierarchical layers. The high-level layer is a state-space motion planner designing highly dynamic locomotion behaviors based on a reduced-order robot model. This motion planner incorporates two robust bundles, named as invariant and recoverability bundles, which quantify analytical state-space deviations for robust planning design. The low-level layer is a modelbased trajectory tracking controller capable of robustly realizing the planned locomotion behaviors. This controller is synthesized based on full-order hybrid dynamic modeling, model-based state feedback control, and Lyapunov stability analysis. The planning and control layers are concatenated by a middle-level trajectory generator that produces nominal behaviors for a full-order robot model. The proposed framework is validated through flat and uneven terrain walking simulations of a three-dimensional bipedal robot.
Minghao Han, Ruixian Zhang, Lixian Zhang, Ye Zhao, Wei Pan
IEEE/CAA Journal of Automatica Sinica, 2020
@article{han_asynchronous_2020,
title = {Asynchronous observer design for switched linear systems: a tube-based approach},
author = {Minghao Han and Ruixian Zhang and Lixian Zhang and Ye Zhao and Wei Pan},
url = {https://ieeexplore.ieee.org/document/8945485/},
doi = {10.1109/JAS.2019.1911822},
issn = {2329-9266, 2329-9274},
year = {2020},
date = {2020-01-01},
urldate = {2025-04-02},
journal = {IEEE/CAA Journal of Automatica Sinica},
volume = {7},
number = {1},
pages = {70–81},
abstract = {This paper proposes a tube-based method for the asynchronous observation problem of discrete-time switched linear systems in the presence of amplitude-bounded disturbances. Sufficient stability conditions of the nominal observer error system under mode-dependent persistent dwell-time (MPDT) switching are first established. Taking the disturbances into account, a novel asynchronous MPDT robust positive invariant (RPI) set and an asynchronous MPDT generalized RPI (GRPI) set are determined for the difference system between the nominal and disturbed observer error systems. Further, the global uniform asymptotical stability of the observer error system is established in the sense of converging to the asynchronous MPDT GRPI set, i.e., the cross section of the tube of the observer error system. Finally, the proposed results are validated on a space robot manipulator example.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper proposes a tube-based method for the asynchronous observation problem of discrete-time switched linear systems in the presence of amplitude-bounded disturbances. Sufficient stability conditions of the nominal observer error system under mode-dependent persistent dwell-time (MPDT) switching are first established. Taking the disturbances into account, a novel asynchronous MPDT robust positive invariant (RPI) set and an asynchronous MPDT generalized RPI (GRPI) set are determined for the difference system between the nominal and disturbed observer error systems. Further, the global uniform asymptotical stability of the observer error system is established in the sense of converging to the asynchronous MPDT GRPI set, i.e., the cross section of the tube of the observer error system. Finally, the proposed results are validated on a space robot manipulator example.
Jianwen Luo, Yao Su, Lecheng Ruan, Ye Zhao, Donghyun Kim, Luis Sentis, Chenglong Fu
Robotica, 2019
@article{luo_robust_2019,
title = {Robust Bipedal Locomotion Based on a Hierarchical Control Structure},
author = {Jianwen Luo and Yao Su and Lecheng Ruan and Ye Zhao and Donghyun Kim and Luis Sentis and Chenglong Fu},
url = {https://www.cambridge.org/core/product/identifier/S0263574719000237/type/journal_article},
doi = {10.1017/S0263574719000237},
issn = {0263-5747, 1469-8668},
year = {2019},
date = {2019-10-01},
urldate = {2025-04-02},
journal = {Robotica},
volume = {37},
number = {10},
pages = {1750–1767},
abstract = {To improve biped locomotion’s robustness to internal and external disturbances, this study proposes a hierarchical structure with three control levels. At the high level, a foothold sequence is generated so that the Center of Mass (CoM) trajectory tracks a planned path. The planning procedure is simplified by selecting the midpoint between two consecutive Center of Pressure (CoP) points as the feature point. At the middle level, a novel robust hybrid controller is devised to drive perturbed system states back to the nominal trajectory within finite cycles without chattering. The novelty lies in that the hybrid controller is not subject to linear CoM dynamic constraints. The hybrid controller consists of two sub-controllers: an oscillation controller and a smoothing controller. For the oscillation controller, the desired CoM height is specified as a sine-shaped function, avoiding a new attractive limit cycle. However, this controller results in the inevitable chattering because of discontinuities. A smoothing controller provides continuous properties and thus can inhibit the chattering problem, but has a smaller region of attraction compared with the oscillation controller. A hybrid controller merges the two controllers for a smooth transition. At the low level, the desired CoM motion is defined as tasks and embedded in a whole body operational space (WBOS) controller to compute the joint torques analytically. The novelty of the low-level controller lies in that within the WBOS framework, CoM motion is not subject to fixed CoM dynamics and thus can be generalized.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
To improve biped locomotion’s robustness to internal and external disturbances, this study proposes a hierarchical structure with three control levels. At the high level, a foothold sequence is generated so that the Center of Mass (CoM) trajectory tracks a planned path. The planning procedure is simplified by selecting the midpoint between two consecutive Center of Pressure (CoP) points as the feature point. At the middle level, a novel robust hybrid controller is devised to drive perturbed system states back to the nominal trajectory within finite cycles without chattering. The novelty lies in that the hybrid controller is not subject to linear CoM dynamic constraints. The hybrid controller consists of two sub-controllers: an oscillation controller and a smoothing controller. For the oscillation controller, the desired CoM height is specified as a sine-shaped function, avoiding a new attractive limit cycle. However, this controller results in the inevitable chattering because of discontinuities. A smoothing controller provides continuous properties and thus can inhibit the chattering problem, but has a smaller region of attraction compared with the oscillation controller. A hybrid controller merges the two controllers for a smooth transition. At the low level, the desired CoM motion is defined as tasks and embedded in a whole body operational space (WBOS) controller to compute the joint torques analytically. The novelty of the low-level controller lies in that within the WBOS framework, CoM motion is not subject to fixed CoM dynamics and thus can be generalized.
Jianwen Luo, Shuguo Wang, Ye Zhao, Yili Fu
Intelligent Service Robotics, 2018
@article{luo_variable_2018,
title = {Variable stiffness control of series elastic actuated biped locomotion},
author = {Jianwen Luo and Shuguo Wang and Ye Zhao and Yili Fu},
url = {http://link.springer.com/10.1007/s11370-018-0248-y},
doi = {10.1007/s11370-018-0248-y},
issn = {1861-2776, 1861-2784},
year = {2018},
date = {2018-07-01},
urldate = {2025-04-08},
journal = {Intelligent Service Robotics},
volume = {11},
number = {3},
pages = {225–235},
abstract = {This study investigates the problem of dynamic walking impact on a biped robot. Two online variable stiffness control algorithms, i.e., torque balance algorithm (TBA) and surface fitting algorithm (SFA), are proposed based on virtual spring leg to achieve compliant performance. These two algorithms target on solving the high nonlinearity commonly existing in legged robot actuators. A planar biped robot experiment platform is designed for testing the proposed variable stiffness control. The experiments compare the performance of TBA and SFA and verify that applying the variable stiffness control of a virtual spring leg is capable of effectively absorbing unforeseen ground impacts and thus improving stability and safety of walking biped robots.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study investigates the problem of dynamic walking impact on a biped robot. Two online variable stiffness control algorithms, i.e., torque balance algorithm (TBA) and surface fitting algorithm (SFA), are proposed based on virtual spring leg to achieve compliant performance. These two algorithms target on solving the high nonlinearity commonly existing in legged robot actuators. A planar biped robot experiment platform is designed for testing the proposed variable stiffness control. The experiments compare the performance of TBA and SFA and verify that applying the variable stiffness control of a virtual spring leg is capable of effectively absorbing unforeseen ground impacts and thus improving stability and safety of walking biped robots.
Ye Zhao, Nicholas Paine, Steven Jens Jorgensen, Luis Sentis
IEEE Transactions on Industrial Electronics, 2018
@article{zhao_impedance_2018,
title = {Impedance Control and Performance Measure of Series Elastic Actuators},
author = {Ye Zhao and Nicholas Paine and Steven Jens Jorgensen and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/8016601/
https://www.google.com/url?q=https%3A%2F%2Fgithub.com%2FYeZhao%2Fseries-elastic-actuation-impedance-control&sa=D&sntz=1&usg=AFQjCNGTlaLzxcN0SS4QH336cUrQF6O0sw
https://www.youtube.com/watch?v=biIdlcAMPyE},
doi = {10.1109/TIE.2017.2745407},
issn = {0278-0046, 1557-9948},
year = {2018},
date = {2018-03-01},
urldate = {2018-03-01},
journal = {IEEE Transactions on Industrial Electronics},
volume = {65},
number = {3},
pages = {2817–2827},
abstract = {Series elastic actuators (SEAs) have become prevalent in torque-controlled robots in recent years to achieve compliant interactions with environments and humans. However, designing optimal impedance controllers and characterizing impedance performance for SEAs with time delays and filtering are still underexplored problems. This article addresses the controller design problem by devising a critically damped gain design method for a class of SEA cascaded control architectures, which is composed of outer impedance and inner torque feedback loops. We indicate that the proposed gain design criterion solves optimal controller gains by maximizing phasemargin-based stability. Meanwhile, we observe a tradeoff between impedance and torque controller gains and analyze their interdependence in terms of closed-loop stability and overall impedance performance. Via the proposed controller design criterion, we adopt frequency-domain methods to thoroughly analyze the effects of time delays, filtering, and load inertia on SEA impedance performance. A novel impedance performance metric, defined as “Z-region,” is proposed to simultaneously quantify achievable impedance magnitude range (i.e., Z-width) and frequency range (i.e., Z-depth). Maximizing the Z-region enables SEA-equipped robots to achieve a wide variety of Cartesian impedance tasks without alternating the control structure. Simulations and experimental implementations are performed to validate the proposed method and performance metric.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Series elastic actuators (SEAs) have become prevalent in torque-controlled robots in recent years to achieve compliant interactions with environments and humans. However, designing optimal impedance controllers and characterizing impedance performance for SEAs with time delays and filtering are still underexplored problems. This article addresses the controller design problem by devising a critically damped gain design method for a class of SEA cascaded control architectures, which is composed of outer impedance and inner torque feedback loops. We indicate that the proposed gain design criterion solves optimal controller gains by maximizing phasemargin-based stability. Meanwhile, we observe a tradeoff between impedance and torque controller gains and analyze their interdependence in terms of closed-loop stability and overall impedance performance. Via the proposed controller design criterion, we adopt frequency-domain methods to thoroughly analyze the effects of time delays, filtering, and load inertia on SEA impedance performance. A novel impedance performance metric, defined as “Z-region,” is proposed to simultaneously quantify achievable impedance magnitude range (i.e., Z-width) and frequency range (i.e., Z-depth). Maximizing the Z-region enables SEA-equipped robots to achieve a wide variety of Cartesian impedance tasks without alternating the control structure. Simulations and experimental implementations are performed to validate the proposed method and performance metric.
Journal Article
Ye Zhao, Benito R Fernandez, Luis Sentis
The International Journal of Robotics Research, 2017
@article{zhao_robust_2017,
title = {Robust optimal planning and control of non-periodic bipedal locomotion with a centroidal momentum model},
author = {Ye Zhao and Benito R Fernandez and Luis Sentis},
url = {https://journals.sagepub.com/doi/10.1177/0278364917730602
https://www.youtube.com/watch?v=eSqQS4z7EYA
https://www.google.com/url?q=https%3A%2F%2Fgithub.com%2FYeZhao%2Fphase-space-planner-locomotion&sa=D&sntz=1&usg=AFQjCNHxP-Zg2DyQ0HRKNoKL5F1Q12gvBg},
doi = {10.1177/0278364917730602},
issn = {0278-3649, 1741-3176},
year = {2017},
date = {2017-09-01},
urldate = {2017-09-01},
journal = {The International Journal of Robotics Research},
volume = {36},
number = {11},
pages = {1211–1242},
abstract = {This study presents a theoretical method for planning and controlling agile bipedal locomotion based on robustly tracking a set of non-periodic keyframe states. Based on centroidal momentum dynamics, we formulate a hybrid phase-space planning and control method that includes the following key components: (i) a step transition solver that enables dynamically tracking non-periodic keyframe states over various types of terrain; (ii) a robust hybrid automaton to effectively formulate planning and control algorithms; (iii) a steering direction model to control the robot’s heading; (iv) a phase-space metric to measure distance to the planned locomotion manifolds; and (v) a hybrid control method based on the previous distance metric to produce robust dynamic locomotion under external disturbances. Compared with other locomotion methodologies, we have a large focus on non-periodic gait generation and robustness metrics to deal with disturbances. This focus enables the proposed control method to track non-periodic keyframe states robustly over various challenging terrains and under external disturbances, as illustrated through several simulations.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study presents a theoretical method for planning and controlling agile bipedal locomotion based on robustly tracking a set of non-periodic keyframe states. Based on centroidal momentum dynamics, we formulate a hybrid phase-space planning and control method that includes the following key components: (i) a step transition solver that enables dynamically tracking non-periodic keyframe states over various types of terrain; (ii) a robust hybrid automaton to effectively formulate planning and control algorithms; (iii) a steering direction model to control the robot’s heading; (iv) a phase-space metric to measure distance to the planned locomotion manifolds; and (v) a hybrid control method based on the previous distance metric to produce robust dynamic locomotion under external disturbances. Compared with other locomotion methodologies, we have a large focus on non-periodic gait generation and robustness metrics to deal with disturbances. This focus enables the proposed control method to track non-periodic keyframe states robustly over various challenging terrains and under external disturbances, as illustrated through several simulations.
Journal Article
Donghyun Kim, Ye Zhao, Gray Thomas, Benito R. Fernandez, Luis Sentis
IEEE Transactions on Robotics, 2016
@article{kim_stabilizing_2016,
title = {Stabilizing Series-Elastic Point-Foot Bipeds Using Whole-Body Operational Space Control},
author = {Donghyun Kim and Ye Zhao and Gray Thomas and Benito R. Fernandez and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/7736085/
https://www.google.com/url?q=https%3A%2F%2Farxiv.org%2Fabs%2F1501.02855&sa=D&sntz=1&usg=AFQjCNFa3MAQXRfT2_v966oMeKfIQJDSBg
https://www.youtube.com/watch?v=DLfmw0yVvrk},
doi = {10.1109/TRO.2016.2597314},
issn = {1552-3098, 1941-0468},
year = {2016},
date = {2016-12-01},
urldate = {2016-12-01},
journal = {IEEE Transactions on Robotics},
volume = {32},
number = {6},
pages = {1362–1379},
abstract = {Whole-body operational space controllers (WBOSCs) are versatile and well suited for simultaneously controlling motion and force behaviors, which can enable sophisticated modes of locomotion and balance. In this paper, we formulate a WBOSC for point-foot bipeds with series-elastic actuators (SEA) and experiment with it using a teen-size SEA biped robot. Our main contributions are on devising a WBOSC strategy for point-foot bipedal robots, 2) formulating planning algorithms for achieving unsupported dynamic balancing on our point-foot biped robot and testing them using a WBOSC, and 3) formulating force feedback control of the internal forces—corresponding to the subset of contact forces that do not generate robot motions—to regulate contact interactions with the complex environment. We experimentally validate the efficacy of our new whole-body control and planning strategies via balancing over a disjointed terrain and attaining dynamic balance through continuous stepping without a mechanical support.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Whole-body operational space controllers (WBOSCs) are versatile and well suited for simultaneously controlling motion and force behaviors, which can enable sophisticated modes of locomotion and balance. In this paper, we formulate a WBOSC for point-foot bipeds with series-elastic actuators (SEA) and experiment with it using a teen-size SEA biped robot. Our main contributions are on devising a WBOSC strategy for point-foot bipedal robots, 2) formulating planning algorithms for achieving unsupported dynamic balancing on our point-foot biped robot and testing them using a WBOSC, and 3) formulating force feedback control of the internal forces—corresponding to the subset of contact forces that do not generate robot motions—to regulate contact interactions with the complex environment. We experimentally validate the efficacy of our new whole-body control and planning strategies via balancing over a disjointed terrain and attaining dynamic balance through continuous stepping without a mechanical support.
Ye Zhao, Nicholas Paine, Kwan Suk Kim, Luis Sentis
IEEE Transactions on Industrial Electronics, 2015
@article{zhao_stability_2015,
title = {Stability and Performance Limits of Latency-Prone Distributed Feedback Controllers},
author = {Ye Zhao and Nicholas Paine and Kwan Suk Kim and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/7131513/
https://www.youtube.com/watch?v=4dXS0gT3ra0
https://www.youtube.com/watch?v=vDiEG-p7Pio
https://www.google.com/url?q=https%3A%2F%2Farxiv.org%2Fpdf%2F1501.02854.pdf&sa=D&sntz=1&usg=AFQjCNE3RoQ65CyQ8mmPVKOJzpmm6H-nUg},
doi = {10.1109/TIE.2015.2448513},
issn = {0278-0046, 1557-9948},
year = {2015},
date = {2015-11-01},
urldate = {2015-11-01},
journal = {IEEE Transactions on Industrial Electronics},
volume = {62},
number = {11},
pages = {7151–7162},
abstract = {Robotic systems are increasingly relying on distributed feedback controllers to tackle complex sensing and decision problems, such as those found in highly articulated human-centered robots. These demands come at the cost of a growing computational burden and, as a result, larger controller latencies. To maximize robustness to mechanical disturbances by maximizing control feedback gains, this paper emphasizes the necessity for compromise between high- and low-level feedback control efforts in distributed controllers. Specifically, the effect of distributed impedance controllers is studied, where damping feedback effort is executed in close proximity to the control plant and stiffness feedback effort is executed in a latency-prone centralized control process. A central observation is that the stability of high-impedance distributed controllers is very sensitive to damping feedback delay but much less to stiffness feedback delay. This study pursues a detailed analysis of this observation that leads to a physical understanding of the disparity. Then, a practical controller breakdown gain rule is derived to aim at enabling control designers to consider the benefits of implementing their control applications in a distributed fashion. These considerations are further validated through the analysis, simulation, and experimental testing on high-performance actuators and on an omnidirectional mobile base.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Robotic systems are increasingly relying on distributed feedback controllers to tackle complex sensing and decision problems, such as those found in highly articulated human-centered robots. These demands come at the cost of a growing computational burden and, as a result, larger controller latencies. To maximize robustness to mechanical disturbances by maximizing control feedback gains, this paper emphasizes the necessity for compromise between high- and low-level feedback control efforts in distributed controllers. Specifically, the effect of distributed impedance controllers is studied, where damping feedback effort is executed in close proximity to the control plant and stiffness feedback effort is executed in a latency-prone centralized control process. A central observation is that the stability of high-impedance distributed controllers is very sensitive to damping feedback delay but much less to stiffness feedback delay. This study pursues a detailed analysis of this observation that leads to a physical understanding of the disparity. Then, a practical controller breakdown gain rule is derived to aim at enabling control designers to consider the benefits of implementing their control applications in a distributed fashion. These considerations are further validated through the analysis, simulation, and experimental testing on high-performance actuators and on an omnidirectional mobile base.
Weichao Sun, Ye Zhao, Jinfu Li, Lixian Zhang, Huijun Gao
IEEE Transactions on Industrial Electronics, 2012
@article{weichao_sun_active_2012,
title = {Active Suspension Control With Frequency Band Constraints and Actuator Input Delay},
author = {Weichao Sun and Ye Zhao and Jinfu Li and Lixian Zhang and Huijun Gao},
url = {http://ieeexplore.ieee.org/document/5740340/},
doi = {10.1109/TIE.2011.2134057},
issn = {0278-0046, 1557-9948},
year = {2012},
date = {2012-01-01},
urldate = {2025-04-02},
journal = {IEEE Transactions on Industrial Electronics},
volume = {59},
number = {1},
pages = {530–537},
abstract = {This paper investigates the problem of vehicle active suspension control with frequency band constraints and actuator input delay. First, the mathematical model of suspension systems is established, and the problem of suspension control with finitefrequency constraints is formulated to match the characteristics of the human body. Then, the finite-frequency method is developed to deal with the problem of suspension control with actuator input delay, based on the generalized Kalman–Yakubovich–Popov lemma. Compared with the traditional entire-frequency approach for active suspension systems, the finite-frequency approach proposed in this paper achieves better disturbance attenuation performance for the chosen frequency range while the constraints required by real situation are guaranteed in the controller design. The effectiveness and merits of the proposed method are verified by a number of simulations with several types of road disturbances.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper investigates the problem of vehicle active suspension control with frequency band constraints and actuator input delay. First, the mathematical model of suspension systems is established, and the problem of suspension control with finitefrequency constraints is formulated to match the characteristics of the human body. Then, the finite-frequency method is developed to deal with the problem of suspension control with actuator input delay, based on the generalized Kalman–Yakubovich–Popov lemma. Compared with the traditional entire-frequency approach for active suspension systems, the finite-frequency approach proposed in this paper achieves better disturbance attenuation performance for the chosen frequency range while the constraints required by real situation are guaranteed in the controller design. The effectiveness and merits of the proposed method are verified by a number of simulations with several types of road disturbances.
Journal Article
Lixian Zhang, Naigang Cui, Ming Liu, Ye Zhao
IEEE Transactions on Circuits and Systems I: Regular Papers, 2011
@article{zhang_asynchronous_2011,
title = {Asynchronous Filtering of Discrete-Time Switched Linear Systems With Average Dwell Time},
author = {Lixian Zhang and Naigang Cui and Ming Liu and Ye Zhao},
url = {http://ieeexplore.ieee.org/document/5672394/},
doi = {10.1109/TCSI.2010.2092151},
issn = {1549-8328, 1558-0806},
year = {2011},
date = {2011-05-01},
urldate = {2011-05-01},
journal = {IEEE Transactions on Circuits and Systems I: Regular Papers},
volume = {58},
number = {5},
pages = {1109–1118},
abstract = {Switched dynamical systems can be found in many practical electronic circuits, such as various kinds of power converters, chaos generators, etc. This paper is concerned with the filter design problem for a class of switched system with average dwell time switching. Mode-dependent full-order filters are designed taking a more practical phenomenon, the asynchronous switching into account, where “asynchronous” means that the switching of the filters to be designed has a lag to the switching of the system modes. New results on the stability and l2-gain analyses for the systems are first given where the Lyapunov-like functions during the running time of subsystems are allowed to increase. In light of the proposed Lyapunov-like functions, the desired mode-dependent filters can be designed in that the unmatched filters are allowed to perform in the interval of the asynchronous switching before the matched ones are applied. In sense, the problem of asynchronous filtering for the underlying systems in linear cases is formulated and the conditions of the existence of admissible asynchronous filters are obtained. Two examples are provided to show the potential of the developed results.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Switched dynamical systems can be found in many practical electronic circuits, such as various kinds of power converters, chaos generators, etc. This paper is concerned with the filter design problem for a class of switched system with average dwell time switching. Mode-dependent full-order filters are designed taking a more practical phenomenon, the asynchronous switching into account, where “asynchronous” means that the switching of the filters to be designed has a lag to the switching of the system modes. New results on the stability and l2-gain analyses for the systems are first given where the Lyapunov-like functions during the running time of subsystems are allowed to increase. In light of the proposed Lyapunov-like functions, the desired mode-dependent filters can be designed in that the unmatched filters are allowed to perform in the interval of the asynchronous switching before the matched ones are applied. In sense, the problem of asynchronous filtering for the underlying systems in linear cases is formulated and the conditions of the existence of admissible asynchronous filters are obtained. Two examples are provided to show the potential of the developed results.
Journal Article
Weichao Sun, Jinfu Li, Ye Zhao, Huijun Gao
Mechatronics, 2011
@article{sun_vibration_2011,
title = {Vibration control for active seat suspension systems via dynamic output feedback with limited frequency characteristic},
author = {Weichao Sun and Jinfu Li and Ye Zhao and Huijun Gao},
url = {https://linkinghub.elsevier.com/retrieve/pii/S0957415810001923},
doi = {10.1016/j.mechatronics.2010.11.001},
issn = {09574158},
year = {2011},
date = {2011-02-01},
urldate = {2025-04-02},
journal = {Mechatronics},
volume = {21},
number = {1},
pages = {250–260},
abstract = {This paper investigates the problem of H1 control for active seat suspension systems via dynamic output feedback control. A vertical vibration model of human body is introduced in order to make the modeling of seat suspension systems more precise. Meantime, different from the existing H1 control methods which conduct disturbance attenuation within the entire frequency domain, this paper addresses the problem of H1 control for active seat suspension systems in finite frequency domain to match the characteristics of the human body. By using the generalized Kalman–Yakubovich–Popov (KYP) lemma, the H1 norm from the disturbance to the controlled output is decreased over the chosen frequency band between which the human body is extremely sensitive to the vibration, to improve the ride comfort. Considering a practical situation of active seat suspension systems, a dynamic output feedback controller of order equal to the plant is designed, where an effective multiplier expansion is used to convert the controller design to a convex optimization problem. Compared with the entire frequency approach for active seat suspension systems, the finite frequency approach achieves better disturbance attenuation for the concerned frequency range, while the performance constraint is guaranteed in the controller design, which is verified by a practical example with certain and random road disturbances.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper investigates the problem of H1 control for active seat suspension systems via dynamic output feedback control. A vertical vibration model of human body is introduced in order to make the modeling of seat suspension systems more precise. Meantime, different from the existing H1 control methods which conduct disturbance attenuation within the entire frequency domain, this paper addresses the problem of H1 control for active seat suspension systems in finite frequency domain to match the characteristics of the human body. By using the generalized Kalman–Yakubovich–Popov (KYP) lemma, the H1 norm from the disturbance to the controlled output is decreased over the chosen frequency band between which the human body is extremely sensitive to the vibration, to improve the ride comfort. Considering a practical situation of active seat suspension systems, a dynamic output feedback controller of order equal to the plant is designed, where an effective multiplier expansion is used to convert the controller design to a convex optimization problem. Compared with the entire frequency approach for active seat suspension systems, the finite frequency approach achieves better disturbance attenuation for the concerned frequency range, while the performance constraint is guaranteed in the controller design, which is verified by a practical example with certain and random road disturbances.
Journal Article
Ye Zhao, Lixian Zhang, Shen Shen, Huijun Gao
IEEE Transactions on Neural Networks, 2011
@article{ye_zhao_robust_2011,
title = {Robust Stability Criterion for Discrete-Time Uncertain Markovian Jumping Neural Networks With Defective Statistics of Modes Transitions},
author = {Ye Zhao and Lixian Zhang and Shen Shen and Huijun Gao},
url = {http://ieeexplore.ieee.org/document/5648729/},
doi = {10.1109/TNN.2010.2093151},
year = {2011},
date = {2011-01-01},
urldate = {2025-04-02},
journal = {IEEE Transactions on Neural Networks},
volume = {22},
pages = {164–170},
abstract = {This brief is concerned with the robust stability problem for a class of discrete-time uncertain Markovian jumping neural networks with defective statistics of modes transitions. The parameter uncertainties are considered to be norm-bounded, and the stochastic perturbations are described in terms of Brownian motion. Defective statistics means that the transition probabilities of the multimode neural networks are not exactly known, as assumed usually. The scenario is more practical, and such defective transition probabilities comprise three types: known, uncertain, and unknown. By invoking the property of the transition probability matrix and the convexity of uncertain domains, a sufficient stability criterion for the underlying system is derived. Furthermore, a monotonicity is observed concerning the maximum value of a given scalar, which bounds the stochastic perturbation that the system can tolerate as the level of the defectiveness varies. Numerical examples are given to verify the effectiveness of the developed results.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This brief is concerned with the robust stability problem for a class of discrete-time uncertain Markovian jumping neural networks with defective statistics of modes transitions. The parameter uncertainties are considered to be norm-bounded, and the stochastic perturbations are described in terms of Brownian motion. Defective statistics means that the transition probabilities of the multimode neural networks are not exactly known, as assumed usually. The scenario is more practical, and such defective transition probabilities comprise three types: known, uncertain, and unknown. By invoking the property of the transition probability matrix and the convexity of uncertain domains, a sufficient stability criterion for the underlying system is derived. Furthermore, a monotonicity is observed concerning the maximum value of a given scalar, which bounds the stochastic perturbation that the system can tolerate as the level of the defectiveness varies. Numerical examples are given to verify the effectiveness of the developed results.
Proceedings Articles
Proceedings Article
Max Asselmeier, Dhruv Ahuja, Abdel Zaro, Ahmad Abuaish, Ye Zhao, Patricio A. Vela
IEEE International Conference on Robotics and Automation (ICRA), 2025
@inproceedings{asselmeier_dynamic_2024,
title = {Dynamic Gap: Safe Gap-based Navigation in Dynamic Environments},
author = {Max Asselmeier and Dhruv Ahuja and Abdel Zaro and Ahmad Abuaish and Ye Zhao and Patricio A. Vela},
url = {http://arxiv.org/abs/2210.05022},
year = {2025},
date = {2025-05-19},
urldate = {2025-05-19},
booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
publisher = {IEEE},
abstract = {This paper extends the family of gap-based local planners to unknown dynamic environments through generating provable collision-free properties for hierarchical navigation systems. Existing perception-informed local planners that operate in dynamic environments rely on emergent or empirical robustness for collision avoidance as opposed to performing formal analysis of dynamic obstacles. In addition to this, the obstacle tracking that is performed in these existent planners is often achieved with respect to a global inertial frame, subjecting such tracking estimates to transformation errors from odometry drift. The proposed local planner, dynamic gap, shifts the tracking paradigm to modeling how the free space, represented as gaps, evolves over time. Gap crossing and closing conditions are developed to aid in determining the feasibility of passage through gaps, and a breadth of simulation benchmarking is performed against other navigation planners in the literature where the proposed dynamic gap planner achieves the highest success rate out of all planners tested in all environments.},
keywords = {},
pubstate = {forthcoming},
tppubtype = {inproceedings}
}
This paper extends the family of gap-based local planners to unknown dynamic environments through generating provable collision-free properties for hierarchical navigation systems. Existing perception-informed local planners that operate in dynamic environments rely on emergent or empirical robustness for collision avoidance as opposed to performing formal analysis of dynamic obstacles. In addition to this, the obstacle tracking that is performed in these existent planners is often achieved with respect to a global inertial frame, subjecting such tracking estimates to transformation errors from odometry drift. The proposed local planner, dynamic gap, shifts the tracking paradigm to modeling how the free space, represented as gaps, evolves over time. Gap crossing and closing conditions are developed to aid in determining the feasibility of passage through gaps, and a breadth of simulation benchmarking is performed against other navigation planners in the literature where the proposed dynamic gap planner achieves the highest success rate out of all planners tested in all environments.
Proceedings Article
Abdulaziz Shamsah, Jesse Jiang, Ziwon Yoon, Samuel Coogan, Ye Zhao
2025 IEEE International Conference on Robotics and Automation (ICRA), 2025
@inproceedings{shamsah_terrain-aware_2024,
title = {Terrain-Aware Model Predictive Control of Heterogeneous Bipedal and Aerial Robot Coordination for Search and Rescue Tasks},
author = {Abdulaziz Shamsah and Jesse Jiang and Ziwon Yoon and Samuel Coogan and Ye Zhao},
url = {http://arxiv.org/abs/2409.15174
https://youtu.be/yWMuhZYh1HI},
year = {2025},
date = {2025-05-19},
urldate = {2024-09-01},
booktitle = {2025 IEEE International Conference on Robotics and Automation (ICRA)},
publisher = {IEEE},
abstract = {Humanoid robots offer significant advantages for search and rescue tasks, thanks to their capability to traverse rough terrains and perform transportation tasks. In this study, we present a task and motion planning framework for search and rescue operations using a heterogeneous robot team composed of humanoids and aerial robots. We propose a terrainaware Model Predictive Controller (MPC) that incorporates terrain elevation gradients learned using Gaussian processes (GP). This terrain-aware MPC generates safe navigation paths for the bipedal robots to traverse rough terrain while minimizing terrain slopes, and it directs the quadrotors to perform aerial search and mapping tasks. The rescue subjects’ locations are estimated by a target belief GP, which is updated online during the map exploration. A high-level planner for task allocation is designed by encoding the navigation tasks using syntactically cosafe Linear Temporal Logic (scLTL), and a consensus-based algorithm is designed for task assignment of individual robots. We evaluate the efficacy of our planning framework in simulation in an uncertain environment with various terrains and random rescue subject placements.},
keywords = {},
pubstate = {forthcoming},
tppubtype = {inproceedings}
}
Humanoid robots offer significant advantages for search and rescue tasks, thanks to their capability to traverse rough terrains and perform transportation tasks. In this study, we present a task and motion planning framework for search and rescue operations using a heterogeneous robot team composed of humanoids and aerial robots. We propose a terrainaware Model Predictive Controller (MPC) that incorporates terrain elevation gradients learned using Gaussian processes (GP). This terrain-aware MPC generates safe navigation paths for the bipedal robots to traverse rough terrain while minimizing terrain slopes, and it directs the quadrotors to perform aerial search and mapping tasks. The rescue subjects’ locations are estimated by a target belief GP, which is updated online during the map exploration. A high-level planner for task allocation is designed by encoding the navigation tasks using syntactically cosafe Linear Temporal Logic (scLTL), and a consensus-based algorithm is designed for task assignment of individual robots. We evaluate the efficacy of our planning framework in simulation in an uncertain environment with various terrains and random rescue subject placements.
Proceedings Article
Xuan Lin, Jiming Ren, Samuel Coogan, Ye Zhao
2025 IEEE International Conference on Robotics and Automation (ICRA), 2025
@inproceedings{lin2024optimization,
title = {Optimization-based Task and Motion Planning under Signal Temporal Logic Specifications using Logic Network Flow},
author = {Xuan Lin and Jiming Ren and Samuel Coogan and Ye Zhao},
url = {https://arxiv.org/abs/2409.19168},
year = {2025},
date = {2025-05-19},
urldate = {2025-05-19},
booktitle = {2025 IEEE International Conference on Robotics and Automation (ICRA)},
publisher = {IEEE},
abstract = {This paper proposes an optimization-based task and motion planning framework, named ``Logic Network Flow", to integrate signal temporal logic (STL) specifications into efficient mixed-binary linear programmings. In this framework, temporal predicates are encoded as polyhedron constraints on each edge of the network flow, instead of as constraints between the nodes as in the traditional Logic Tree formulation. Synthesized with Dynamic Network Flows, Logic Network Flows render a tighter convex relaxation compared to Logic Trees derived from these STL specifications. Our formulation is evaluated on several multi-robot motion planning case studies. Empirical results demonstrate that our formulation outperforms Logic Tree formulation in terms of computation time for several planning problems. As the problem size scales up, our method still discovers better lower and upper bounds by exploring fewer number of nodes during the branch-and-bound process, although this comes at the cost of increased computational load for each node when exploring branches.},
keywords = {},
pubstate = {forthcoming},
tppubtype = {inproceedings}
}
This paper proposes an optimization-based task and motion planning framework, named ``Logic Network Flow", to integrate signal temporal logic (STL) specifications into efficient mixed-binary linear programmings. In this framework, temporal predicates are encoded as polyhedron constraints on each edge of the network flow, instead of as constraints between the nodes as in the traditional Logic Tree formulation. Synthesized with Dynamic Network Flows, Logic Network Flows render a tighter convex relaxation compared to Logic Trees derived from these STL specifications. Our formulation is evaluated on several multi-robot motion planning case studies. Empirical results demonstrate that our formulation outperforms Logic Tree formulation in terms of computation time for several planning problems. As the problem size scales up, our method still discovers better lower and upper bounds by exploring fewer number of nodes during the branch-and-bound process, although this comes at the cost of increased computational load for each node when exploring branches.
Proceedings Article
Jiming Ren, Haris Miller, Karen M. Feigh, Samuel Coogan, Ye Zhao
2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024
@inproceedings{ren_ltl-d_2024,
title = {LTL-D*: Incrementally Optimal Replanning for Feasible and Infeasible Tasks in Linear Temporal Logic Specifications},
author = {Jiming Ren and Haris Miller and Karen M. Feigh and Samuel Coogan and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10802218/
https://youtu.be/Fu-FrcdluCI},
doi = {10.1109/IROS58592.2024.10802218},
isbn = {979-8-3503-7770-5},
year = {2024},
date = {2024-10-01},
urldate = {2024-10-01},
booktitle = {2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages = {4495–4502},
publisher = {IEEE},
address = {Abu Dhabi, United Arab Emirates},
abstract = {This paper presents an incremental replanning algorithm, dubbed LTL-D*, for temporal-logic-based task planning in a dynamically changing environment. Unexpected changes in the environment may lead to failures in satisfying a task specification in the form of a Linear Temporal Logic (LTL). In this study, the considered failures are categorized into two classes: (i) the desired LTL specification can be satisfied via replanning, and (ii) the desired LTL specification is infeasible to meet strictly and can only be satisfied in a “relaxed” fashion. To address these failures, the proposed algorithm finds an optimal replanning solution that minimally violates desired task specifications. In particular, our approach leverages the D* Lite algorithm and employs a distance metric within the synthesized automaton to quantify the degree of the task violation and then replan incrementally. This ensures plan optimality and reduces planning time, especially when frequent replanning is required. Our approach is implemented in a robot navigation simulation to demonstrate a significant improvement in the computational efficiency for replanning by two orders of magnitude.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper presents an incremental replanning algorithm, dubbed LTL-D*, for temporal-logic-based task planning in a dynamically changing environment. Unexpected changes in the environment may lead to failures in satisfying a task specification in the form of a Linear Temporal Logic (LTL). In this study, the considered failures are categorized into two classes: (i) the desired LTL specification can be satisfied via replanning, and (ii) the desired LTL specification is infeasible to meet strictly and can only be satisfied in a “relaxed” fashion. To address these failures, the proposed algorithm finds an optimal replanning solution that minimally violates desired task specifications. In particular, our approach leverages the D* Lite algorithm and employs a distance metric within the synthesized automaton to quantify the degree of the task violation and then replan incrementally. This ensures plan optimality and reduces planning time, especially when frequent replanning is required. Our approach is implemented in a robot navigation simulation to demonstrate a significant improvement in the computational efficiency for replanning by two orders of magnitude.
Proceedings Article
Kasidit Muenprasitivej, Jesse Jiang, Abdulaziz Shamsah, Samuel Coogan, Ye Zhao
2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024
@inproceedings{muenprasitivej_bipedal_2024,
title = {Bipedal Safe Navigation over Uncertain Rough Terrain: Unifying Terrain Mapping and Locomotion Stability},
author = {Kasidit Muenprasitivej and Jesse Jiang and Abdulaziz Shamsah and Samuel Coogan and Ye Zhao},
url = {https://doi.org/10.1109/IROS58592.2024.10802816
http://arxiv.org/abs/2403.16356
https://youtu.be/27hOcBNKAvU},
doi = {https://doi.org/10.1109/IROS58592.2024.10802816},
year = {2024},
date = {2024-10-01},
urldate = {2024-10-01},
booktitle = {2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages = {11264-11271},
publisher = {IEEE},
address = {Abu Dhabi, United Arab Emirates},
abstract = {We study the problem of bipedal robot navigation in complex environments with uncertain and rough terrain. In particular, we consider a scenario in which the robot is expected to reach a desired goal location by traversing an environment with uncertain terrain elevation. Such terrain uncertainties induce not only untraversable regions but also robot motion perturbations. Thus, the problems of terrain mapping and locomotion stability are intertwined. We evaluate three different kernels for Gaussian process (GP) regression to learn the terrain elevation. We also learn the motion deviation resulting from both the terrain as well as the discrepancy between the reduced-order Prismatic Inverted Pendulum Model used for planning and the full-order locomotion dynamics. We propose a hierarchical locomotion-dynamics-aware sampling-based navigation planner. The global navigation planner plans a series of local waypoints to reach the desired goal locations while respecting locomotion stability constraints. Then, a local navigation planner is used to generate a sequence of dynamically feasible footsteps to reach local waypoints. We develop a novel trajectory evaluation metric to minimize motion deviation and maximize information gain of the terrain elevation map. We evaluate the efficacy of our planning framework on Digit bipedal robot simulation in MuJoCo.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We study the problem of bipedal robot navigation in complex environments with uncertain and rough terrain. In particular, we consider a scenario in which the robot is expected to reach a desired goal location by traversing an environment with uncertain terrain elevation. Such terrain uncertainties induce not only untraversable regions but also robot motion perturbations. Thus, the problems of terrain mapping and locomotion stability are intertwined. We evaluate three different kernels for Gaussian process (GP) regression to learn the terrain elevation. We also learn the motion deviation resulting from both the terrain as well as the discrepancy between the reduced-order Prismatic Inverted Pendulum Model used for planning and the full-order locomotion dynamics. We propose a hierarchical locomotion-dynamics-aware sampling-based navigation planner. The global navigation planner plans a series of local waypoints to reach the desired goal locations while respecting locomotion stability constraints. Then, a local navigation planner is used to generate a sequence of dynamically feasible footsteps to reach local waypoints. We develop a novel trajectory evaluation metric to minimize motion deviation and maximize information gain of the terrain elevation map. We evaluate the efficacy of our planning framework on Digit bipedal robot simulation in MuJoCo.
Real-time Model Predictive Control with Zonotope-Based Neural Networks for Bipedal Social Navigation
Proceedings Article
Abdulaziz Shamsah, Krishanu Agarwal, Shreyas Kousik, Ye Zhao
2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024
@inproceedings{shamsah_real-time_2024,
title = {Real-time Model Predictive Control with Zonotope-Based Neural Networks for Bipedal Social Navigation},
author = {Abdulaziz Shamsah and Krishanu Agarwal and Shreyas Kousik and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10801435/
https://www.youtube.com/watch?v=w5cg66GQPUQ},
doi = {10.1109/IROS58592.2024.10801435},
isbn = {979-8-3503-7770-5},
year = {2024},
date = {2024-10-01},
urldate = {2024-10-01},
booktitle = {2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages = {13741–13748},
publisher = {IEEE},
address = {Abu Dhabi, United Arab Emirates},
abstract = {This study addresses the challenge of bipedal navigation in a dynamic human-crowded environment, a research area that remains largely underexplored in the field of legged navigation. We propose two cascaded zonotope-based neural networks: a Pedestrian Prediction Network (PPN) for pedestrians’ future trajectory prediction and an Ego-agent Social Network (ESN) for ego-agent social path planning. Representing future paths as zonotopes allows for efficient reachability-based planning and collision checking. The ESN is then integrated with a Model Predictive Controller (ESN-MPC) for footstep planning for our bipedal robot Digit designed by Agility Robotics. ESN-MPC solves for a collision-free optimal trajectory by optimizing through the gradients of ESN. ESNMPC optimal trajectory is sent to the low-level controller for full-order simulation of Digit. The overall proposed framework is validated with extensive simulations on randomly generated initial settings with varying human crowd densities.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This study addresses the challenge of bipedal navigation in a dynamic human-crowded environment, a research area that remains largely underexplored in the field of legged navigation. We propose two cascaded zonotope-based neural networks: a Pedestrian Prediction Network (PPN) for pedestrians’ future trajectory prediction and an Ego-agent Social Network (ESN) for ego-agent social path planning. Representing future paths as zonotopes allows for efficient reachability-based planning and collision checking. The ESN is then integrated with a Model Predictive Controller (ESN-MPC) for footstep planning for our bipedal robot Digit designed by Agility Robotics. ESN-MPC solves for a collision-free optimal trajectory by optimizing through the gradients of ESN. ESNMPC optimal trajectory is sent to the low-level controller for full-order simulation of Digit. The overall proposed framework is validated with extensive simulations on randomly generated initial settings with varying human crowd densities.
Proceedings Article
Jesse Jiang, Ye Zhao, Samuel Coogan
2024 American Control Conference (ACC), 2024
@inproceedings{jiang_local-global_2024,
title = {Local-Global Interval MDPs for Efficient Motion Planning with Learnable Uncertainty},
author = {Jesse Jiang and Ye Zhao and Samuel Coogan},
url = {https://ieeexplore.ieee.org/document/10644660/},
doi = {10.23919/ACC60939.2024.10644660},
isbn = {979-8-3503-8265-5},
year = {2024},
date = {2024-07-01},
urldate = {2025-04-02},
booktitle = {2024 American Control Conference (ACC)},
pages = {1343–1349},
publisher = {IEEE},
address = {Toronto, ON, Canada},
abstract = {We study the problem of computationally efficient control synthesis for Interval Markov Decision Processes (IMDPs), that is, MDPs with interval uncertainty on the transition probabilities, against tasks specified in linear temporal logic. To address the scalability challenge when synthesizing this control policy in a holistic way, we propose decomposing the monolithic global IMDP into a collection of interconnected local IMDPs. We focus on the problem of robotic motion planning. Specifically, we assume a setting in which the transition probabilities can be learned and their interval uncertainty reduced by observing the dynamics of the system at runtime. This creates an objective of exploration to ensure that the planning task can be completed with sufficient probability of success. We perform decoupled exploration and learning on the local IMDPs and then combine local control policies to guarantee global task satisfaction. In a simulation-based case study, we show that, compared to existing approaches, our proposed decomposition leads to faster learning and satisfaction of the planning task and provides a feasible controller when other methods are infeasible.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We study the problem of computationally efficient control synthesis for Interval Markov Decision Processes (IMDPs), that is, MDPs with interval uncertainty on the transition probabilities, against tasks specified in linear temporal logic. To address the scalability challenge when synthesizing this control policy in a holistic way, we propose decomposing the monolithic global IMDP into a collection of interconnected local IMDPs. We focus on the problem of robotic motion planning. Specifically, we assume a setting in which the transition probabilities can be learned and their interval uncertainty reduced by observing the dynamics of the system at runtime. This creates an objective of exploration to ensure that the planning task can be completed with sufficient probability of success. We perform decoupled exploration and learning on the local IMDPs and then combine local control policies to guarantee global task satisfaction. In a simulation-based case study, we show that, compared to existing approaches, our proposed decomposition leads to faster learning and satisfaction of the planning task and provides a feasible controller when other methods are infeasible.
Proceedings Article
Feiyang Wu, Zhaoyuan Gu, Hanran Wu, Anqi Wu, Ye Zhao
2024 IEEE International Conference on Robotics and Automation (ICRA), 2024
@inproceedings{wu_infer_2024,
title = {Infer and Adapt: Bipedal Locomotion Reward Learning from Demonstrations via Inverse Reinforcement Learning},
author = {Feiyang Wu and Zhaoyuan Gu and Hanran Wu and Anqi Wu and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10611685/
https://youtu.be/WAviIdSmGM0?si=5tUX1swHDCjHGkEB},
doi = {10.1109/ICRA57147.2024.10611685},
isbn = {979-8-3503-8457-4},
year = {2024},
date = {2024-05-01},
urldate = {2024-05-01},
booktitle = {2024 IEEE International Conference on Robotics and Automation (ICRA)},
pages = {16243–16250},
publisher = {IEEE},
address = {Yokohama, Japan},
abstract = {Enabling bipedal walking robots to learn how to maneuver over highly uneven, dynamically changing terrains is challenging due to the complexity of robot dynamics and interacted environments. Recent advancements in learning from demonstrations have shown promising results for robot learning in complex environments. While imitation learning of expert policies has been well-explored, the study of learning expert reward functions is largely under-explored in legged locomotion. This paper brings state-of-the-art Inverse Reinforcement Learning (IRL) techniques to solving bipedal locomotion problems over complex terrains. We propose algorithms for learning expert reward functions, and we subsequently analyze the learned functions. Through nonlinear function approximation, we uncover meaningful insights into the expert’s locomotion strategies. Furthermore, we empirically demonstrate that training a bipedal locomotion policy with the inferred reward functions enhances its walking performance on unseen terrains, highlighting the adaptability offered by reward learning.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Enabling bipedal walking robots to learn how to maneuver over highly uneven, dynamically changing terrains is challenging due to the complexity of robot dynamics and interacted environments. Recent advancements in learning from demonstrations have shown promising results for robot learning in complex environments. While imitation learning of expert policies has been well-explored, the study of learning expert reward functions is largely under-explored in legged locomotion. This paper brings state-of-the-art Inverse Reinforcement Learning (IRL) techniques to solving bipedal locomotion problems over complex terrains. We propose algorithms for learning expert reward functions, and we subsequently analyze the learned functions. Through nonlinear function approximation, we uncover meaningful insights into the expert’s locomotion strategies. Furthermore, we empirically demonstrate that training a bipedal locomotion policy with the inferred reward functions enhances its walking performance on unseen terrains, highlighting the adaptability offered by reward learning.
Proceedings Article
Max Asselmeier, Jane Ivanova, Ziyi Zhou, Patricio A. Vela, Ye Zhao
2024 IEEE International Conference on Robotics and Automation (ICRA), 2024
@inproceedings{asselmeier_hierarchical_2024,
title = {Hierarchical Experience-informed Navigation for Multi-modal Quadrupedal Rebar Grid Traversal},
author = {Max Asselmeier and Jane Ivanova and Ziyi Zhou and Patricio A. Vela and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10610248/
https://youtu.be/NHK-VPDyDm0},
doi = {10.1109/ICRA57147.2024.10610248},
isbn = {979-8-3503-8457-4},
year = {2024},
date = {2024-05-01},
urldate = {2024-05-01},
booktitle = {2024 IEEE International Conference on Robotics and Automation (ICRA)},
pages = {8065–8072},
publisher = {IEEE},
address = {Yokohama, Japan},
abstract = {This study focuses on a layered, experience-based, multi-modal contact planning framework for agile quadrupedal locomotion over a constrained rebar environment. To this end, our hierarchical planner incorporates locomotion-specific modules into the high-level contact sequence planner and solves kinodynamically-aware trajectory optimization as the low-level motion planner. Through quantitative analysis of the experience accumulation process and experimental validation of the kinodynamic feasibility of the generated locomotion trajectories, we demonstrate that the experience planning heuristic offers an effective way of providing candidate footholds for a legged contact planner. Additionally, we introduce a guiding torso path heuristic at the global planning level to enhance the navigation success rate in the presence of environmental obstacles. Our results indicate that the torso-path guided experience accumulation requires significantly fewer offline trials to successfully reach the goal compared to regular experience accumulation. Finally, our planning framework is validated in both dynamics simulations and real hardware implementations on a quadrupedal robot provided by Skymul Inc.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This study focuses on a layered, experience-based, multi-modal contact planning framework for agile quadrupedal locomotion over a constrained rebar environment. To this end, our hierarchical planner incorporates locomotion-specific modules into the high-level contact sequence planner and solves kinodynamically-aware trajectory optimization as the low-level motion planner. Through quantitative analysis of the experience accumulation process and experimental validation of the kinodynamic feasibility of the generated locomotion trajectories, we demonstrate that the experience planning heuristic offers an effective way of providing candidate footholds for a legged contact planner. Additionally, we introduce a guiding torso path heuristic at the global planning level to enhance the navigation success rate in the presence of environmental obstacles. Our results indicate that the torso-path guided experience accumulation requires significantly fewer offline trials to successfully reach the goal compared to regular experience accumulation. Finally, our planning framework is validated in both dynamics simulations and real hardware implementations on a quadrupedal robot provided by Skymul Inc.
Proceedings Article
Zhaoyuan Gu, Rongming Guo, William Yates, Yipu Chen, Yuntian Zhao, Ye Zhao
2024 IEEE International Conference on Robotics and Automation (ICRA), 2024
@inproceedings{gu_walking-by-logic_2024,
title = {Walking-by-Logic: Signal Temporal Logic-Guided Model Predictive Control for Bipedal Locomotion Resilient to External Perturbations},
author = {Zhaoyuan Gu and Rongming Guo and William Yates and Yipu Chen and Yuntian Zhao and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10610811/
https://youtu.be/ALPZ2QX-rZo?si=UtJupBGi7x2_80vL},
doi = {10.1109/ICRA57147.2024.10610811},
isbn = {979-8-3503-8457-4},
year = {2024},
date = {2024-05-01},
urldate = {2024-05-01},
booktitle = {2024 IEEE International Conference on Robotics and Automation (ICRA)},
pages = {1121–1127},
publisher = {IEEE},
address = {Yokohama, Japan},
abstract = {This study proposes a novel planning framework based on a model predictive control formulation that incorporates signal temporal logic (STL) specifications for task completion guarantees and robustness quantification. This marks the first-ever study to apply STL-guided trajectory optimization for bipedal locomotion push recovery, where the robot experiences unexpected disturbances. Existing recovery strategies often struggle with complex task logic reasoning and locomotion robustness evaluation, making them susceptible to failures caused by inappropriate recovery strategies or insufficient robustness. To address this issue, the STL-guided framework generates optimal and safe recovery trajectories that simultaneously satisfy the task specification and maximize the locomotion robustness. Our framework outperforms a stateof-the-art locomotion controller in a high-fidelity dynamic simulation, especially in scenarios involving crossed-leg maneuvers. Furthermore, it demonstrates versatility in tasks such as locomotion on stepping stones, where the robot must select from a set of disjointed footholds to maneuver successfully.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This study proposes a novel planning framework based on a model predictive control formulation that incorporates signal temporal logic (STL) specifications for task completion guarantees and robustness quantification. This marks the first-ever study to apply STL-guided trajectory optimization for bipedal locomotion push recovery, where the robot experiences unexpected disturbances. Existing recovery strategies often struggle with complex task logic reasoning and locomotion robustness evaluation, making them susceptible to failures caused by inappropriate recovery strategies or insufficient robustness. To address this issue, the STL-guided framework generates optimal and safe recovery trajectories that simultaneously satisfy the task specification and maximize the locomotion robustness. Our framework outperforms a stateof-the-art locomotion controller in a high-fidelity dynamic simulation, especially in scenarios involving crossed-leg maneuvers. Furthermore, it demonstrates versatility in tasks such as locomotion on stepping stones, where the robot must select from a set of disjointed footholds to maneuver successfully.
Proceedings Article
Kelin Yu, Yunhai Han, Qixian Wang, Vaibhav Saxena, Danfei Xu, Ye Zhao
Conference on Robot Learning (CoRL), 2024
@inproceedings{yu_mimictouch_2024,
title = {MimicTouch: Leveraging Multi-modal Human Tactile Demonstrations for Contact-rich Manipulation},
author = {Kelin Yu and Yunhai Han and Qixian Wang and Vaibhav Saxena and Danfei Xu and Ye Zhao},
url = {http://arxiv.org/abs/2310.16917
https://sites.google.com/view/mimictouch/%E9%A6%96%E9%A1%B5},
doi = {10.48550/arXiv.2310.16917},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
booktitle = {Conference on Robot Learning (CoRL)},
abstract = {Tactile sensing is critical to fine-grained, contact-rich manipulation tasks, such as insertion and assembly. Prior research has shown the possibility of learning tactile-guided policy from teleoperated demonstration data. However, to provide the demonstration, human demonstrators often rely on visual feedback to control the robot. This creates a gap between the sensing modality used for controlling the robot (visual) and the modality of interest (tactile). To bridge this gap, we introduce “MimicTouch”, a novel framework for learning policies directly from demonstrations provided by human users with their hands. The key innovations are i) a human tactile data collection system which collects multimodal tactile dataset for learning human’s tactile-guided control strategy, ii) an imitation learning-based framework for learning human’s tactile-guided control strategy through such data, and iii) an online residual RL framework to bridge the embodiment gap between the human hand and the robot gripper. Through comprehensive experiments, we highlight the efficacy of utilizing human’s tactile-guided control strategy to resolve contact-rich manipulation tasks. The project website is at https://sites.google.com/view/MimicTouch.},
note = {arXiv:2310.16917 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Tactile sensing is critical to fine-grained, contact-rich manipulation tasks, such as insertion and assembly. Prior research has shown the possibility of learning tactile-guided policy from teleoperated demonstration data. However, to provide the demonstration, human demonstrators often rely on visual feedback to control the robot. This creates a gap between the sensing modality used for controlling the robot (visual) and the modality of interest (tactile). To bridge this gap, we introduce “MimicTouch”, a novel framework for learning policies directly from demonstrations provided by human users with their hands. The key innovations are i) a human tactile data collection system which collects multimodal tactile dataset for learning human’s tactile-guided control strategy, ii) an imitation learning-based framework for learning human’s tactile-guided control strategy through such data, and iii) an online residual RL framework to bridge the embodiment gap between the human hand and the robot gripper. Through comprehensive experiments, we highlight the efficacy of utilizing human’s tactile-guided control strategy to resolve contact-rich manipulation tasks. The project website is at https://sites.google.com/view/MimicTouch.
Proceedings Article
Shuai Wang, Jingfan Zhang, Weiyi Kong, Chong Zhang, Jie Lai, Dongsheng Zhang, Chunyan Wang, Ke Chen, Zhaoyuan Gu, Ye Zhao, Ke Zhang, Yu Zheng
2023 21st International Conference on Advanced Robotics (ICAR), 2023
@inproceedings{wang_hybrid_2023,
title = {Hybrid Stepping Motion Generation for Wheeled-Bipedal Robots Without Roll Joints on Legs},
author = {Shuai Wang and Jingfan Zhang and Weiyi Kong and Chong Zhang and Jie Lai and Dongsheng Zhang and Chunyan Wang and Ke Chen and Zhaoyuan Gu and Ye Zhao and Ke Zhang and Yu Zheng},
url = {https://ieeexplore.ieee.org/document/10406411/},
doi = {10.1109/ICAR58858.2023.10406411},
isbn = {979-8-3503-4229-1},
year = {2023},
date = {2023-12-01},
urldate = {2025-04-08},
booktitle = {2023 21st International Conference on Advanced Robotics (ICAR)},
pages = {332–339},
publisher = {IEEE},
address = {Abu Dhabi, United Arab Emirates},
abstract = {Wheeled-bipedal robots without roll joints on legs, such as Handle by Boston Dynamics and Ascento by ETH, have drawn increasing attention due to their superior motion agility but pose unique challenges to motion generation. So far, there is little to no research on how to enable these robots to step forward with their legs. In this study, we will explore hybrid stepping locomotion strategies via a two-phase design procedure. During the single-leg support phase, a twomass variable height inverted pendulum model will be used for stepping locomotion generation and control. As for the double-leg support phase, given the difficulty of modeling contact sliding, friction, and collision, a model-free reinforcement learning approach is employed to leverage the rich data for reliable motion generation. Experiments on our own developed wheeled-bipedal robot Ollie demonstrate that the robot is capable of stepping forward with varied stepping frequencies. Stepping with yaw rotation and tests in different scenarios show the efficacy and robustness of the hybrid stepping motion generation method.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Wheeled-bipedal robots without roll joints on legs, such as Handle by Boston Dynamics and Ascento by ETH, have drawn increasing attention due to their superior motion agility but pose unique challenges to motion generation. So far, there is little to no research on how to enable these robots to step forward with their legs. In this study, we will explore hybrid stepping locomotion strategies via a two-phase design procedure. During the single-leg support phase, a twomass variable height inverted pendulum model will be used for stepping locomotion generation and control. As for the double-leg support phase, given the difficulty of modeling contact sliding, friction, and collision, a model-free reinforcement learning approach is employed to leverage the rich data for reliable motion generation. Experiments on our own developed wheeled-bipedal robot Ollie demonstrate that the robot is capable of stepping forward with varied stepping frequencies. Stepping with yaw rotation and tests in different scenarios show the efficacy and robustness of the hybrid stepping motion generation method.
Proceedings Article
Shiyu Feng, Ziyi Zhou, Justin S. Smith, Max Asselmeier, Ye Zhao, Patricio A. Vela
2023 IEEE International Conference on Robotics and Automation (ICRA), 2023
@inproceedings{feng_gpf-bg_2023,
title = {GPF-BG: A Hierarchical Vision-Based Planning Framework for Safe Quadrupedal Navigation},
author = {Shiyu Feng and Ziyi Zhou and Justin S. Smith and Max Asselmeier and Ye Zhao and Patricio A. Vela},
url = {https://ieeexplore.ieee.org/document/10160804/
https://youtu.be/avUnefrbhY8},
doi = {10.1109/ICRA48891.2023.10160804},
isbn = {979-8-3503-2365-8},
year = {2023},
date = {2023-05-01},
urldate = {2023-05-01},
booktitle = {2023 IEEE International Conference on Robotics and Automation (ICRA)},
pages = {1968–1975},
publisher = {IEEE},
address = {London, United Kingdom},
abstract = {Safe quadrupedal navigation through unknown environments is a challenging problem. This paper proposes a hierarchical vision-based planning framework (GPF-BG) integrating our previous Global Path Follower (GPF) navigation system and a gap-based local planner using Be´zier curves, so called B´ezier gap (BG). This BG-based trajectory synthesis can generate smooth trajectories and guarantee safety for point-mass robots. With an empirical robot geometry extension and stabilized perception space, safety and robustness are significantly improved for quadrupedal navigation. Both simulation and real experiments are conducted to evaluate safe navigation performance of the proposed framework and multiple benchmarking frameworks. The proposed GPF-BG demonstrates the best safety results among all experiments.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Safe quadrupedal navigation through unknown environments is a challenging problem. This paper proposes a hierarchical vision-based planning framework (GPF-BG) integrating our previous Global Path Follower (GPF) navigation system and a gap-based local planner using Be´zier curves, so called B´ezier gap (BG). This BG-based trajectory synthesis can generate smooth trajectories and guarantee safety for point-mass robots. With an empirical robot geometry extension and stabilized perception space, safety and robustness are significantly improved for quadrupedal navigation. Both simulation and real experiments are conducted to evaluate safe navigation performance of the proposed framework and multiple benchmarking frameworks. The proposed GPF-BG demonstrates the best safety results among all experiments.
Proceedings Article
Ted Tyler, Vaibhav Malhotra, Adam Montague, Zhigen Zhao, Frank L. Hammond III, Ye Zhao
Modeling, Estimation, and Control Conference (MECC), 2023
@inproceedings{tyler_integrating_2023,
title = {Integrating Reconfigurable Foot Design, Multi-modal Contact Sensing, and Terrain Classification for Bipedal Locomotion},
author = {Ted Tyler and Vaibhav Malhotra and Adam Montague and Zhigen Zhao and Frank L. Hammond III and Ye Zhao},
url = {http://arxiv.org/abs/2304.09370},
doi = {10.48550/arXiv.2304.09370},
year = {2023},
date = {2023-04-01},
urldate = {2025-04-02},
booktitle = {Modeling, Estimation, and Control Conference (MECC)},
abstract = {The ability of bipedal robots to adapt to diverse and unstructured terrain conditions is crucial for their deployment in real-world environments. To this end, we present a novel, bio-inspired robot foot design with stabilizing tarsal segments and a multifarious sensor suite involving acoustic, capacitive, tactile, temperature, and acceleration sensors. A real-time signal processing and terrain classification system is developed and evaluated. The sensed terrain information is used to control actuated segments of the foot, leading to improved ground contact and stability. The proposed framework highlights the potential of the sensor-integrated adaptive foot for intelligent and adaptive locomotion.},
note = {arXiv:2304.09370 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The ability of bipedal robots to adapt to diverse and unstructured terrain conditions is crucial for their deployment in real-world environments. To this end, we present a novel, bio-inspired robot foot design with stabilizing tarsal segments and a multifarious sensor suite involving acoustic, capacitive, tactile, temperature, and acceleration sensors. A real-time signal processing and terrain classification system is developed and evaluated. The sensed terrain information is used to control actuated segments of the foot, leading to improved ground contact and stability. The proposed framework highlights the potential of the sensor-integrated adaptive foot for intelligent and adaptive locomotion.
Proceedings Article
Yunhai Han, Madie Xie, Ye Zhao, Harish Ravichandar
Conference on Robot Learning (CoRL), 2023
@inproceedings{han_utility_2023,
title = {On the Utility of Koopman Operator Theory in Learning Dexterous Manipulation Skills},
author = {Yunhai Han and Madie Xie and Ye Zhao and Harish Ravichandar},
url = {https://proceedings.mlr.press/v229/han23a.html},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
booktitle = {Conference on Robot Learning (CoRL)},
abstract = {Despite impressive dexterous manipulation capabilities enabled by learning-based approaches, we are yet to witness widespread adoption beyond well-resourced laboratories. This is likely due to practical limitations, such as significant computational burden, inscrutable learned behaviors, sensitivity to initialization, and the considerable technical expertise required for implementation. In this work, we investigate the utility of Koopman operator theory in alleviating these limitations. Koopman operators are simple yet powerful control-theoretic structures to represent complex nonlinear dynamics as linear systems in higher dimensions. Motivated by the fact that complex nonlinear dynamics underlie dexterous manipulation, we develop a Koopman operator-based imitation learning framework to learn the desired motions of both the robotic hand and the object simultaneously. We show that Koopman operators are surprisingly effective for dexterous manipulation and offer a number of unique benefits. Notably, policies can be learned analytically, drastically reducing computation burden and eliminating sensitivity to initialization and the need for painstaking hyperparameter optimization. Our experiments reveal that a Koopman operator-based approach can perform comparably to state-of-the-art imitation learning algorithms in terms of success rate and sample efficiency, while being an order of magnitude faster. Policy videos can be viewed at https://sites.google.com/view/kodex-corl.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Despite impressive dexterous manipulation capabilities enabled by learning-based approaches, we are yet to witness widespread adoption beyond well-resourced laboratories. This is likely due to practical limitations, such as significant computational burden, inscrutable learned behaviors, sensitivity to initialization, and the considerable technical expertise required for implementation. In this work, we investigate the utility of Koopman operator theory in alleviating these limitations. Koopman operators are simple yet powerful control-theoretic structures to represent complex nonlinear dynamics as linear systems in higher dimensions. Motivated by the fact that complex nonlinear dynamics underlie dexterous manipulation, we develop a Koopman operator-based imitation learning framework to learn the desired motions of both the robotic hand and the object simultaneously. We show that Koopman operators are surprisingly effective for dexterous manipulation and offer a number of unique benefits. Notably, policies can be learned analytically, drastically reducing computation burden and eliminating sensitivity to initialization and the need for painstaking hyperparameter optimization. Our experiments reveal that a Koopman operator-based approach can perform comparably to state-of-the-art imitation learning algorithms in terms of success rate and sample efficiency, while being an order of magnitude faster. Policy videos can be viewed at https://sites.google.com/view/kodex-corl.
Proceedings Article
Hongyi Chen, Shiyu Feng, Ye Zhao, Changliu Liu, Patricio A. Vela
2022 IEEE 61st Conference on Decision and Control (CDC), 2022
@inproceedings{chen_safe_2022,
title = {Safe Hierarchical Navigation in Crowded Dynamic Uncertain Environments},
author = {Hongyi Chen and Shiyu Feng and Ye Zhao and Changliu Liu and Patricio A. Vela},
url = {https://ieeexplore.ieee.org/document/9992674/},
doi = {10.1109/CDC51059.2022.9992674},
isbn = {978-1-6654-6761-2},
year = {2022},
date = {2022-12-01},
urldate = {2025-04-02},
booktitle = {2022 IEEE 61st Conference on Decision and Control (CDC)},
pages = {1174–1181},
publisher = {IEEE},
address = {Cancun, Mexico},
abstract = {This paper describes a hierarchical solution consisting of a multi-phase planner and a low-level safe controller to jointly solve the safe navigation problem in crowded, dynamic, and uncertain environments. The planner employs dynamic gap analysis and trajectory optimization to achieve collision avoidance with respect to the predicted trajectories of dynamic agents within the sensing and planning horizon and with robustness to agent uncertainty. To address uncertainty over the planning horizon and real-time safety, a fast reactive safe set algorithm (SSA) is adopted, which monitors and modifies the unsafe control during trajectory tracking. Compared to other existing methods, our approach offers theoretical guarantees of safety and achieves collision-free navigation with higher probability in uncertain environments, as demonstrated in scenarios with 20 and 50 dynamic agents.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper describes a hierarchical solution consisting of a multi-phase planner and a low-level safe controller to jointly solve the safe navigation problem in crowded, dynamic, and uncertain environments. The planner employs dynamic gap analysis and trajectory optimization to achieve collision avoidance with respect to the predicted trajectories of dynamic agents within the sensing and planning horizon and with robustness to agent uncertainty. To address uncertainty over the planning horizon and real-time safety, a fast reactive safe set algorithm (SSA) is adopted, which monitors and modifies the unsafe control during trajectory tracking. Compared to other existing methods, our approach offers theoretical guarantees of safety and achieves collision-free navigation with higher probability in uncertain environments, as demonstrated in scenarios with 20 and 50 dynamic agents.
Proceedings Article
Michael E. Cao, Xinpei Ni, Jonas Warnke, Yunhai Han, Samuel Coogan, Ye Zhao
2022 IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR), 2022
@inproceedings{cao_leveraging_2022,
title = {Leveraging Heterogeneous Capabilities in Multi-Agent Systems for Environmental Conflict Resolution},
author = {Michael E. Cao and Xinpei Ni and Jonas Warnke and Yunhai Han and Samuel Coogan and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/10018728/
https://www.youtube.com/watch?v=6Lci6Ozi23M},
doi = {10.1109/SSRR56537.2022.10018728},
isbn = {978-1-6654-5680-7},
year = {2022},
date = {2022-11-01},
urldate = {2022-11-01},
booktitle = {2022 IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR)},
pages = {94–101},
publisher = {IEEE},
address = {Sevilla, Spain},
abstract = {In this paper, we introduce a high-level controller synthesis framework that enables teams of heterogeneous agents to assist each other in resolving environmental conflicts that appear at runtime. This conflict resolution method is built upon temporal-logic-based reactive synthesis to guarantee safety and task completion under specific environment assumptions. In heterogeneous multi-agent systems, every agent is expected to complete its own tasks in service of a global team objective. However, at runtime, an agent may encounter un-modeled obstacles (e.g., doors or walls) that prevent it from achieving its own task. To address this problem, we employ the capabilities of other heterogeneous agents to resolve the obstacle. A controller framework is proposed to redirect agents with the capability of resolving the appropriate obstacles to the required target when such a situation is detected. Three case studies involving a bipedal robot Digit and a quadcopter are used to evaluate the controller performance in action. Additionally, we implement the proposed framework on a physical multi-agent robotic system to demonstrate its viability for real world applications.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we introduce a high-level controller synthesis framework that enables teams of heterogeneous agents to assist each other in resolving environmental conflicts that appear at runtime. This conflict resolution method is built upon temporal-logic-based reactive synthesis to guarantee safety and task completion under specific environment assumptions. In heterogeneous multi-agent systems, every agent is expected to complete its own tasks in service of a global team objective. However, at runtime, an agent may encounter un-modeled obstacles (e.g., doors or walls) that prevent it from achieving its own task. To address this problem, we employ the capabilities of other heterogeneous agents to resolve the obstacle. A controller framework is proposed to redirect agents with the capability of resolving the appropriate obstacles to the required target when such a situation is detected. Three case studies involving a bipedal robot Digit and a quadcopter are used to evaluate the controller performance in action. Additionally, we implement the proposed framework on a physical multi-agent robotic system to demonstrate its viability for real world applications.
Ziyi Zhou, Dong Jae Lee, Yuki Yoshinaga, Stephen Balakirsky, Dejun Guo, Ye Zhao
IEEE International Conference on Automation Science and Engineering (CASE), 2022
@inproceedings{zhou_reactive_2022,
title = {Reactive Task Allocation and Planning for Quadrupedal and Wheeled Robot Teaming},
author = {Ziyi Zhou and Dong Jae Lee and Yuki Yoshinaga and Stephen Balakirsky and Dejun Guo and Ye Zhao},
url = {http://arxiv.org/abs/2110.08436
https://www.youtube.com/watch?v=xtjLYctN03Y
https://github.com/GTLIDAR/ltl_multi_agent},
doi = {10.48550/arXiv.2110.08436},
year = {2022},
date = {2022-06-01},
urldate = {2022-06-01},
booktitle = {IEEE International Conference on Automation Science and Engineering (CASE)},
publisher = {IEEE},
abstract = {This paper takes the first step towards a reactive, hierarchical multi-robot task allocation and planning framework given a global Linear Temporal Logic specification. The capabilities of both quadrupedal and wheeled robots are leveraged via a heterogeneous team to accomplish a variety of navigation and delivery tasks. However, when deployed in the real world, all robots can be susceptible to different types of disturbances, including but not limited to locomotion failures, human interventions, and obstructions from the environment. To address these disturbances, we propose task-level local and global reallocation strategies to efficiently generate updated action-state sequences online while guaranteeing the completion of the original task. These task reallocation approaches eliminate reconstructing the entire plan or resynthesizing a new task. To integrate the task planner with low-level inputs, a Behavior Tree execution layer monitors different types of disturbances and employs the reallocation methods to make corresponding recovery strategies. To evaluate this planning framework, dynamic simulations are conducted in a realistic hospital environment with a heterogeneous robot team consisting of quadrupeds and wheeled robots for delivery tasks.},
note = {arXiv:2110.08436 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper takes the first step towards a reactive, hierarchical multi-robot task allocation and planning framework given a global Linear Temporal Logic specification. The capabilities of both quadrupedal and wheeled robots are leveraged via a heterogeneous team to accomplish a variety of navigation and delivery tasks. However, when deployed in the real world, all robots can be susceptible to different types of disturbances, including but not limited to locomotion failures, human interventions, and obstructions from the environment. To address these disturbances, we propose task-level local and global reallocation strategies to efficiently generate updated action-state sequences online while guaranteeing the completion of the original task. These task reallocation approaches eliminate reconstructing the entire plan or resynthesizing a new task. To integrate the task planner with low-level inputs, a Behavior Tree execution layer monitors different types of disturbances and employs the reallocation methods to make corresponding recovery strategies. To evaluate this planning framework, dynamic simulations are conducted in a realistic hospital environment with a heterogeneous robot team consisting of quadrupeds and wheeled robots for delivery tasks.
Proceedings Article
Zhigen Zhao, Zhigen Zhao, Simiao Zuo, Tuo Zhao, Ye Zhao
Annual Conference on Learning for Dynamics and Control (L4DC), 2022
@inproceedings{zhao_adversarially_2022,
title = {Adversarially Regularized Policy Learning Guided by Trajectory Optimization},
author = {Zhigen Zhao and Zhigen Zhao and Simiao Zuo and Tuo Zhao and Ye Zhao},
url = {https://proceedings.mlr.press/v168/zhao22b.html},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
booktitle = {Annual Conference on Learning for Dynamics and Control (L4DC)},
abstract = {Recent advancement in combining trajectory optimization with function approximation (especially neural networks) shows promise in learning complex control policies for diverse tasks in robot systems. Despite their great flexibility, the large neural networks for parameterizing control policies impose significant challenges. The learned neural control policies are often overcomplex and non-smooth, which can easily cause unexpected or diverging robot motions. Therefore, they often yield poor generalization performance in practice. To address this issue, we propose adversarially regularized policy learning guided by trajectory optimization (VERONICA) for learning smooth control policies. Specifically, our proposed approach controls the smoothness (local Lipschitz continuity) of the neural control policies by stabilizing the output control with respect to the worst-case perturbation to the input state. Our experiments on robot manipulation show that our proposed approach not only improves the sample efficiency of neural policy learning but also enhances the robustness of the policy against various types of disturbances, including sensor noise, environmental uncertainty, and model mismatch.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Recent advancement in combining trajectory optimization with function approximation (especially neural networks) shows promise in learning complex control policies for diverse tasks in robot systems. Despite their great flexibility, the large neural networks for parameterizing control policies impose significant challenges. The learned neural control policies are often overcomplex and non-smooth, which can easily cause unexpected or diverging robot motions. Therefore, they often yield poor generalization performance in practice. To address this issue, we propose adversarially regularized policy learning guided by trajectory optimization (VERONICA) for learning smooth control policies. Specifically, our proposed approach controls the smoothness (local Lipschitz continuity) of the neural control policies by stabilizing the output control with respect to the worst-case perturbation to the input state. Our experiments on robot manipulation show that our proposed approach not only improves the sample efficiency of neural policy learning but also enhances the robustness of the policy against various types of disturbances, including sensor noise, environmental uncertainty, and model mismatch.
Proceedings Article
Sergio Aguilera, Muhammad Ali Murtaza, Ye Zhao, Seth Hutchinson
2021 IEEE International Conference on Robotics and Automation (ICRA), 2021
@inproceedings{aguilera_mass_2021,
title = {Mass Estimation of a Moving Object Through Minimal Manipulation Interaction},
author = {Sergio Aguilera and Muhammad Ali Murtaza and Ye Zhao and Seth Hutchinson},
url = {https://ieeexplore.ieee.org/document/9560975/},
doi = {10.1109/ICRA48506.2021.9560975},
isbn = {978-1-7281-9077-8},
year = {2021},
date = {2021-05-01},
urldate = {2025-04-02},
booktitle = {2021 IEEE International Conference on Robotics and Automation (ICRA)},
pages = {6461–6467},
publisher = {IEEE},
address = {Xi'an, China},
abstract = {In this paper, we study the problem of dynamic interaction between a robot and an unknown object (e.g., catching a ball, or handing off an object during locomotion). In particular, we propose a method for estimating the inertial parameters of an object during dynamic interaction, while minimally altering the trajectory of the object – a minimal interaction approach. Our method combines trajectory estimation (e.g., using standard methods from computer vision) with a model-based estimator that exploits the robot’s known dynamic model. We first develop the method for a generalized threedimensional problem, and then evaluate the method for the case of an object moving along a linear trajectory. We present experimental results obtained using a KUKA iiwa 7 interacting with rolling balls of varying mass. Our experiments demonstrate that the mass of the objects can be accurately estimated at the moment of impact when accurate object trajectory estimates are available, and that significant improvement can be obtained by incorporating force measurements at the contact point while following the object.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we study the problem of dynamic interaction between a robot and an unknown object (e.g., catching a ball, or handing off an object during locomotion). In particular, we propose a method for estimating the inertial parameters of an object during dynamic interaction, while minimally altering the trajectory of the object – a minimal interaction approach. Our method combines trajectory estimation (e.g., using standard methods from computer vision) with a model-based estimator that exploits the robot’s known dynamic model. We first develop the method for a generalized threedimensional problem, and then evaluate the method for the case of an object moving along a linear trajectory. We present experimental results obtained using a KUKA iiwa 7 interacting with rolling balls of varying mass. Our experiments demonstrate that the mass of the objects can be accurately estimated at the moment of impact when accurate object trajectory estimates are available, and that significant improvement can be obtained by incorporating force measurements at the contact point while following the object.
Proceedings Article
Jonas Warnke, Abdulaziz Shamsah, Yingke Li, Ye Zhao
2020 59th IEEE Conference on Decision and Control (CDC), 2020
@inproceedings{warnke_towards_2020,
title = {Towards Safe Locomotion Navigation in Partially Observable Environments with Uneven Terrain},
author = {Jonas Warnke and Abdulaziz Shamsah and Yingke Li and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/9304219/
https://www.youtube.com/watch?v=q2qkb7nJ9-Y&t=1s
https://github.com/GTLIDAR/safe-nav-locomotion},
doi = {10.1109/CDC42340.2020.9304219},
isbn = {978-1-7281-7447-1},
year = {2020},
date = {2020-12-01},
urldate = {2020-12-01},
booktitle = {2020 59th IEEE Conference on Decision and Control (CDC)},
pages = {958–965},
publisher = {IEEE},
address = {Jeju, Korea (South)},
abstract = {This study proposes an integrated task and motion planning method for dynamic locomotion in partially observable environments with multi-level safety guarantees. This layered planning framework is composed of a high-level symbolic task planner and a low-level phase-space motion planner. A belief abstraction at the task planning level enables belief estimation of dynamic obstacle locations and guarantees navigation safety with collision avoidance. The high-level task planner, i.e., a two-level navigation planner, employs linear temporal logic for a reactive game synthesis between the robot and its environment while incorporating low-level safe keyframe policies into formal task specification design. The synthesized task planner commands a series of locomotion actions including walking step length, step height, and heading angle changes, to the underlying keyframe decision-maker, which further determines the robot center-of-mass apex velocity keyframe. The low-level phase-space planner uses a reduced-order locomotion model to generate non-periodic trajectories meeting balancing safety criteria for straight and steering walking. These criteria are characterized by constraints on locomotion keyframe states, and are used to define keyframe transition policies via viability kernels. Simulation results of a Cassie bipedal robot designed by Agility Robotics demonstrate locomotion maneuvering in a three-dimensional, partially observable environment consisting of dynamic obstacles and uneven terrain.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This study proposes an integrated task and motion planning method for dynamic locomotion in partially observable environments with multi-level safety guarantees. This layered planning framework is composed of a high-level symbolic task planner and a low-level phase-space motion planner. A belief abstraction at the task planning level enables belief estimation of dynamic obstacle locations and guarantees navigation safety with collision avoidance. The high-level task planner, i.e., a two-level navigation planner, employs linear temporal logic for a reactive game synthesis between the robot and its environment while incorporating low-level safe keyframe policies into formal task specification design. The synthesized task planner commands a series of locomotion actions including walking step length, step height, and heading angle changes, to the underlying keyframe decision-maker, which further determines the robot center-of-mass apex velocity keyframe. The low-level phase-space planner uses a reduced-order locomotion model to generate non-periodic trajectories meeting balancing safety criteria for straight and steering walking. These criteria are characterized by constraints on locomotion keyframe states, and are used to define keyframe transition policies via viability kernels. Simulation results of a Cassie bipedal robot designed by Agility Robotics demonstrate locomotion maneuvering in a three-dimensional, partially observable environment consisting of dynamic obstacles and uneven terrain.
Proceedings Article
Lasitha Wijayarathne, Qie Sima, Ziyi Zhou, Ye Zhao, Frank L. Hammond
2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020
@inproceedings{wijayarathne_simultaneous_2020,
title = {Simultaneous Trajectory Optimization and Force Control with Soft Contact Mechanics},
author = {Lasitha Wijayarathne and Qie Sima and Ziyi Zhou and Ye Zhao and Frank L. Hammond},
url = {https://ieeexplore.ieee.org/document/9341030/},
doi = {10.1109/IROS45743.2020.9341030},
isbn = {978-1-7281-6212-6},
year = {2020},
date = {2020-10-01},
urldate = {2025-04-02},
booktitle = {2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages = {3164–3171},
publisher = {IEEE},
address = {Las Vegas, NV, USA},
abstract = {Force modulation of robotic manipulators has been extensively studied for several decades but is not yet commonly used in safety-critical applications due to a lack of accurate interaction contact modeling and weak performance guarantees - a large proportion of them concerning the modulation of interaction forces. This study presents a highlevel framework for simultaneous trajectory optimization and force control of the interaction between manipulator and soft environments. Sliding friction and normal contact force are taken into account. The dynamics of the soft contact model and the manipulator dynamics are simultaneously incorporated in a trajectory optimizer to generate desired motion and force profiles. A constrained optimization framework based on Differential Dynamic Programming and Alternative Direction Method of Multipliers has been employed to generate optimal control inputs and high-dimensional state trajectories. Experimental validation of the model performance is conducted on a soft substrate with known material properties using a Cartesian space force control mode. Results show a comparison of ground truth and predicted model based contact force states for multiple Cartesian motions and the validity range of the friction model. The proposed high-level planning has the potential to be leveraged for medical tasks involving manipulation of compliant, delicate, and deformable tissues.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Force modulation of robotic manipulators has been extensively studied for several decades but is not yet commonly used in safety-critical applications due to a lack of accurate interaction contact modeling and weak performance guarantees - a large proportion of them concerning the modulation of interaction forces. This study presents a highlevel framework for simultaneous trajectory optimization and force control of the interaction between manipulator and soft environments. Sliding friction and normal contact force are taken into account. The dynamics of the soft contact model and the manipulator dynamics are simultaneously incorporated in a trajectory optimizer to generate desired motion and force profiles. A constrained optimization framework based on Differential Dynamic Programming and Alternative Direction Method of Multipliers has been employed to generate optimal control inputs and high-dimensional state trajectories. Experimental validation of the model performance is conducted on a soft substrate with known material properties using a Cartesian space force control mode. Results show a comparison of ground truth and predicted model based contact force states for multiple Cartesian motions and the validity range of the friction model. The proposed high-level planning has the potential to be leveraged for medical tasks involving manipulation of compliant, delicate, and deformable tissues.
Proceedings Article
Marko Mihalec, Ye Zhao, Jingang Yi
2020 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), 2020
@inproceedings{mihalec_recoverability_2020,
title = {Recoverability Estimation and Control for an Inverted Pendulum Walker Model Under Foot Slip},
author = {Marko Mihalec and Ye Zhao and Jingang Yi},
url = {https://ieeexplore.ieee.org/document/9159043/},
doi = {10.1109/AIM43001.2020.9159043},
isbn = {978-1-7281-6794-7},
year = {2020},
date = {2020-07-01},
urldate = {2025-04-02},
booktitle = {2020 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM)},
pages = {771–776},
publisher = {IEEE},
address = {Boston, MA, USA},
abstract = {Locomotion on low-friction surfaces is one of the most challenging problems for bipedal walking. When a stance foot moves and slips on the ground surface, the walker tries to determine whether it is feasible to avoid falling and continue walking. This study uses a simplified two-mass linear inverted pendulum model to analyze the biped dynamics under foot-slip conditions while maintaining closed-form solutions. Using the model, we analytically calculate safe, recoverable, and falling sets to determine whether the walker is able to recover towards a stable position or the fall is inevitable. We present a set of configurations which partition state space and determine the recoverability of the walker. A simple center-of-mass controller is introduced to re-gain the stability by allowing the walker to recover from fall-prone configurations. One attractive property of the developed closed-form expressions lies in feasibility for real-time implementation as a basis for a high-level robust slip recovery controller.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Locomotion on low-friction surfaces is one of the most challenging problems for bipedal walking. When a stance foot moves and slips on the ground surface, the walker tries to determine whether it is feasible to avoid falling and continue walking. This study uses a simplified two-mass linear inverted pendulum model to analyze the biped dynamics under foot-slip conditions while maintaining closed-form solutions. Using the model, we analytically calculate safe, recoverable, and falling sets to determine whether the walker is able to recover towards a stable position or the fall is inevitable. We present a set of configurations which partition state space and determine the recoverability of the walker. A simple center-of-mass controller is introduced to re-gain the stability by allowing the walker to recover from fall-prone configurations. One attractive property of the developed closed-form expressions lies in feasibility for real-time implementation as a basis for a high-level robust slip recovery controller.
Proceedings Article
Ziyi Zhou, Ye Zhao
2020 American Control Conference (ACC), 2020
@inproceedings{zhou_accelerated_2020,
title = {Accelerated ADMM based Trajectory Optimization for Legged Locomotion with Coupled Rigid Body Dynamics},
author = {Ziyi Zhou and Ye Zhao},
url = {https://ieeexplore.ieee.org/document/9147887/
https://www.youtube.com/watch?v=BP3YILbidN0&t
},
doi = {10.23919/ACC45564.2020.9147887},
isbn = {978-1-5386-8266-1},
year = {2020},
date = {2020-07-01},
urldate = {2020-07-01},
booktitle = {2020 American Control Conference (ACC)},
pages = {5082–5089},
publisher = {IEEE},
address = {Denver, CO, USA},
abstract = {Trajectory optimization is becoming increasingly powerful in addressing motion planning problems of underactuated robotic systems. Numerous prior studies solve such a class of large non-convex optimal control problems in a hierarchical fashion. However, numerical accuracy issues are prone to occur when one uses a full-order model to track reference trajectories generated from a reduced-order model. This study investigates an approach of Alternating Direction Method of Multipliers (ADMM) and proposes a new splitting scheme for legged locomotion problems. Rigid body dynamics constraints and other general constraints such as box and cone constraints are decomposed to multiple sub-problems in a principled manner. The resulting multi-block ADMM framework enables us to leverage the efficiency of an unconstrained optimization method–Differential Dynamical Programming–to iteratively solve the optimizations using centroidal and wholebody models. Furthermore, we propose a Stage-wise Accelerated ADMM with over-relaxation and varying-penalty schemes to improve the overall convergence rate. We evaluate and validate the performance of the proposed ADMM algorithm on a car-parking example and a bipedal locomotion problem over rough terrains.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Trajectory optimization is becoming increasingly powerful in addressing motion planning problems of underactuated robotic systems. Numerous prior studies solve such a class of large non-convex optimal control problems in a hierarchical fashion. However, numerical accuracy issues are prone to occur when one uses a full-order model to track reference trajectories generated from a reduced-order model. This study investigates an approach of Alternating Direction Method of Multipliers (ADMM) and proposes a new splitting scheme for legged locomotion problems. Rigid body dynamics constraints and other general constraints such as box and cone constraints are decomposed to multiple sub-problems in a principled manner. The resulting multi-block ADMM framework enables us to leverage the efficiency of an unconstrained optimization method–Differential Dynamical Programming–to iteratively solve the optimizations using centroidal and wholebody models. Furthermore, we propose a Stage-wise Accelerated ADMM with over-relaxation and varying-penalty schemes to improve the overall convergence rate. We evaluate and validate the performance of the proposed ADMM algorithm on a car-parking example and a bipedal locomotion problem over rough terrains.
Sutej Kulgod, Wentao Chen, Junda Huang, Ye Zhao, Nikolay Atanasov
2020 American Control Conference (ACC), 2020
@inproceedings{kulgod_temporal_2020,
title = {Temporal Logic Guided Locomotion Planning and Control in Cluttered Environments},
author = {Sutej Kulgod and Wentao Chen and Junda Huang and Ye Zhao and Nikolay Atanasov},
url = {https://ieeexplore.ieee.org/document/9147621/},
doi = {10.23919/ACC45564.2020.9147621},
isbn = {978-1-5386-8266-1},
year = {2020},
date = {2020-07-01},
urldate = {2020-07-01},
booktitle = {2020 American Control Conference (ACC)},
pages = {5425–5432},
publisher = {IEEE},
address = {Denver, CO, USA},
abstract = {We present planning and control techniques for non-periodic locomotion tasks specified by temporal logic in rough cluttered terrains. Our planning approach is based on a discrete set of motion primitives for the center of mass (CoM) of a general bipedal robot model. A deterministic shortest path problem is solved over the Büchi automaton of the temporal logic task specification, composed with the graph of CoM keyframe states generated by the motion primitives. A lowlevel controller based on quadratic programming is proposed to track the resulting CoM and foot trajectories. We demonstrate dynamically stable, non-periodic locomotion of a kneed compass gait bipedal robot satisfying complex task specifications.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We present planning and control techniques for non-periodic locomotion tasks specified by temporal logic in rough cluttered terrains. Our planning approach is based on a discrete set of motion primitives for the center of mass (CoM) of a general bipedal robot model. A deterministic shortest path problem is solved over the Büchi automaton of the temporal logic task specification, composed with the graph of CoM keyframe states generated by the motion primitives. A lowlevel controller based on quadratic programming is proposed to track the resulting CoM and foot trajectories. We demonstrate dynamically stable, non-periodic locomotion of a kneed compass gait bipedal robot satisfying complex task specifications.
Xiaofeng Guo, Bryan Blaise, Jennifer Molnar, Jeremiah Coholich, Shantanu Padte, Ye Zhao, Frank L. Hammond
2020 3rd IEEE International Conference on Soft Robotics (RoboSoft), 2020
@inproceedings{guo_soft_2020,
title = {Soft Foot Sensor Design and Terrain Classification for Dynamic Legged Locomotion},
author = {Xiaofeng Guo and Bryan Blaise and Jennifer Molnar and Jeremiah Coholich and Shantanu Padte and Ye Zhao and Frank L. Hammond},
url = {https://ieeexplore.ieee.org/document/9115990/
https://www.youtube.com/watch?v=X_rr4-1K_Yw
},
doi = {10.1109/RoboSoft48309.2020.9115990},
isbn = {978-1-7281-6570-7},
year = {2020},
date = {2020-05-01},
urldate = {2020-05-01},
booktitle = {2020 3rd IEEE International Conference on Soft Robotics (RoboSoft)},
pages = {550–557},
publisher = {IEEE},
address = {New Haven, CT, USA},
abstract = {Dynamic legged locomotion is being explored as a means to maneuver on rugged and unstructured terrains. However, limited foot contact sensing capabilities often prohibit bipedal robots from being deployed on complex terrains. Locomotion over cluttered outdoor environments requires the contacting foot to be aware of terrain geometries, stiffness, and granular media properties. To achieve this, we designed a new soft contact pad integrated with a variety of embedded sensors, including tactile, acoustic, capacitive, and temperature sensors, as well as an accelerometer. In addition, we devised a terrain classification algorithm based on features extracted from those sensors and various real-world terrains. The classifier uses these features as inputs and classifies various terrains via Random Forests and a memory function. Our cross-validation tests demonstrate that the proposed classification algorithm achieves an accuracy of about 96.5%, manifesting the applicability of this foot sensing device to bipedal locomotion over diverse terrains.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Dynamic legged locomotion is being explored as a means to maneuver on rugged and unstructured terrains. However, limited foot contact sensing capabilities often prohibit bipedal robots from being deployed on complex terrains. Locomotion over cluttered outdoor environments requires the contacting foot to be aware of terrain geometries, stiffness, and granular media properties. To achieve this, we designed a new soft contact pad integrated with a variety of embedded sensors, including tactile, acoustic, capacitive, and temperature sensors, as well as an accelerometer. In addition, we devised a terrain classification algorithm based on features extracted from those sensors and various real-world terrains. The classifier uses these features as inputs and classifies various terrains via Random Forests and a memory function. Our cross-validation tests demonstrate that the proposed classification algorithm achieves an accuracy of about 96.5%, manifesting the applicability of this foot sensing device to bipedal locomotion over diverse terrains.
Proceedings Article
Akash Harapanahalli, Emil Muly, Hogan Welch, Timothy Brumfiel, Zhengyang Weng, Luke Drnach, Dong Jae Lee, Ye Zhao
IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), 2020
@inproceedings{harapanahalli_towards_2020,
title = {Towards a Biomimetic and Dexterous Robot Avatar: Design, Control, and Kinematics Considerations},
author = {Akash Harapanahalli and Emil Muly and Hogan Welch and Timothy Brumfiel and Zhengyang Weng and Luke Drnach and Dong Jae Lee and Ye Zhao},
url = {https://akashhara.com/files/Athena_Abstract.pdf},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM)},
note = {First Place in the Hardware, Devices & Robotics Track of the 2021 GaTech VIP Innovation Competition},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Proceedings Article
Jianwen Luo, Ye Zhao, Donghyun Kim, Oussama Khatib, Luis Sentis
2017 IEEE International Conference on Robotics and Biomimetics (ROBIO), 2017
@inproceedings{luo_locomotion_2017,
title = {Locomotion control of three dimensional passive-foot biped robot based on whole body operational space framework},
author = {Jianwen Luo and Ye Zhao and Donghyun Kim and Oussama Khatib and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/8324642/},
doi = {10.1109/ROBIO.2017.8324642},
isbn = {978-1-5386-3742-5},
year = {2017},
date = {2017-12-01},
urldate = {2025-04-02},
booktitle = {2017 IEEE International Conference on Robotics and Biomimetics (ROBIO)},
pages = {1577–1582},
publisher = {IEEE},
address = {Macau},
abstract = {This study presents a whole body operational space (WBOS) framework for controlling three dimensional passive-foot biped robot. Stability of WBOS controller is analyzed, and a foot placement planner is proposed. In many cases, the WBOS controller generates torque commands to execute the trajectories planned by high-level planners at every control loop. The planners design trajectories by sensing the locomotion behaviors over a long horizon. Instead, our planner updates a step location every control loop by estimating the center-of-mass (CoM) state to achieve robust balancing. The robustness is enhanced because contact events vary the stance leg switching time from a given nominal step frequency. Via this new foot placement planner, the locomotion robustness to unknown terrains is improved. Dynamically stable walking are tested on flat and unknown terrains, and under push recovery by using a real-time dynamic simulation.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This study presents a whole body operational space (WBOS) framework for controlling three dimensional passive-foot biped robot. Stability of WBOS controller is analyzed, and a foot placement planner is proposed. In many cases, the WBOS controller generates torque commands to execute the trajectories planned by high-level planners at every control loop. The planners design trajectories by sensing the locomotion behaviors over a long horizon. Instead, our planner updates a step location every control loop by estimating the center-of-mass (CoM) state to achieve robust balancing. The robustness is enhanced because contact events vary the stance leg switching time from a given nominal step frequency. Via this new foot placement planner, the locomotion robustness to unknown terrains is improved. Dynamically stable walking are tested on flat and unknown terrains, and under push recovery by using a real-time dynamic simulation.
Proceedings Article
Ye Zhao, Ufuk Topcu, Luis Sentis
2016 IEEE 55th Conference on Decision and Control (CDC), 2016
@inproceedings{zhao_high-level_2016,
title = {High-level planner synthesis for whole-body locomotion in unstructured environments},
author = {Ye Zhao and Ufuk Topcu and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/7799278/
https://www.youtube.com/watch?v=urp7xu8vx3s
https://drive.google.com/file/d/0B_7VcYBOhr8uWlpGb3FqcGpMeWs/view},
doi = {10.1109/CDC.2016.7799278},
isbn = {978-1-5090-1837-6},
year = {2016},
date = {2016-12-01},
urldate = {2016-12-01},
booktitle = {2016 IEEE 55th Conference on Decision and Control (CDC)},
pages = {6557–6564},
publisher = {IEEE},
address = {Las Vegas, NV, USA},
abstract = {Contact-based decision and planning methods are increasingly being sought for task execution in humanoid robots. However, formal methods from the verification and synthesis communities have not been yet incorporated into the motion planning sequence for complex mobility behaviors in humanoid robots. This study takes a step toward formally synthesizing high-level reactive task planners for whole-body locomotion in unstructured environments. We formulate a twoplayer temporal logic game between the contact planner and its possibly adversarial environment. The resulting discrete planner satisfies the given task specifications expressed in a fragment of temporal logic. The resulting commands are executed by a low-level 3D phase-space motion planner algorithm. We devise various low-level locomotion modes based on centroidal momentum dynamics. Provable correctness of the low-level execution of the synthesized discrete task planner is guaranteed through the so-called simulation relations. Simulations of dynamic locomotion in unstructured environments support the effectiveness of the hierarchical planner protocol.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Contact-based decision and planning methods are increasingly being sought for task execution in humanoid robots. However, formal methods from the verification and synthesis communities have not been yet incorporated into the motion planning sequence for complex mobility behaviors in humanoid robots. This study takes a step toward formally synthesizing high-level reactive task planners for whole-body locomotion in unstructured environments. We formulate a twoplayer temporal logic game between the contact planner and its possibly adversarial environment. The resulting discrete planner satisfies the given task specifications expressed in a fragment of temporal logic. The resulting commands are executed by a low-level 3D phase-space motion planner algorithm. We devise various low-level locomotion modes based on centroidal momentum dynamics. Provable correctness of the low-level execution of the synthesized discrete task planner is guaranteed through the so-called simulation relations. Simulations of dynamic locomotion in unstructured environments support the effectiveness of the hierarchical planner protocol.
Proceedings Article
Ye Zhao, Luis Sentis
2016 IEEE-RAS 16th International Conference on Humanoid Robots (Humanoids), 2016
@inproceedings{zhao_passivity_2016,
title = {Passivity of time-delayed whole-body operational space control with series elastic actuation},
author = {Ye Zhao and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/7803436/
https://drive.google.com/open?id=0B_7VcYBOhr8uWHNybTRCbTJKenM},
doi = {10.1109/HUMANOIDS.2016.7803436},
isbn = {978-1-5090-4718-5},
year = {2016},
date = {2016-11-01},
urldate = {2016-11-01},
booktitle = {2016 IEEE-RAS 16th International Conference on Humanoid Robots (Humanoids)},
pages = {1290–1297},
publisher = {IEEE},
address = {Cancun, Mexico},
abstract = {Whole-Body Control has been extensively used to achieve humanoid robot force and motion tasks simultaneously during recent years. However, most existing results have not incorporated low-level actuator dynamics and time delays yet. In this study, we propose a novel time-delayed WholeBody Operational Space control (WBOSC) with series elastic actuator (SEA) dynamics. This type of controller generalizes our previously proposed distributed control structure to multiinput and multi-output free floating humanoid robotic systems. Namely, Cartesian stiffness control is adopted to design the WBOSC at the centralized level while motor damping control is implemented at the embedded level to remedy the stability deterioration caused by time delays. Additionally, embeddedlevel torque feedback control is formulated and physically interpreted as a shaping of the motor inertia. To ensure passivity, we separate the overall system into two subsystems interconnected in a feedback configuration. By the LyapunovKrasovskii functional technique, we propose a delay-dependent passivity criterion of the closed-loop system in the form of linear matrix inequalities (LMIs), and solve for the allowable maximum time delays via the passivity criterion. Numerical simulations of a dynamic locomotion process are used to validate the proposed passivity criterion and the WBOSC framework.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Whole-Body Control has been extensively used to achieve humanoid robot force and motion tasks simultaneously during recent years. However, most existing results have not incorporated low-level actuator dynamics and time delays yet. In this study, we propose a novel time-delayed WholeBody Operational Space control (WBOSC) with series elastic actuator (SEA) dynamics. This type of controller generalizes our previously proposed distributed control structure to multiinput and multi-output free floating humanoid robotic systems. Namely, Cartesian stiffness control is adopted to design the WBOSC at the centralized level while motor damping control is implemented at the embedded level to remedy the stability deterioration caused by time delays. Additionally, embeddedlevel torque feedback control is formulated and physically interpreted as a shaping of the motor inertia. To ensure passivity, we separate the overall system into two subsystems interconnected in a feedback configuration. By the LyapunovKrasovskii functional technique, we propose a delay-dependent passivity criterion of the closed-loop system in the form of linear matrix inequalities (LMIs), and solve for the allowable maximum time delays via the passivity criterion. Numerical simulations of a dynamic locomotion process are used to validate the proposed passivity criterion and the WBOSC framework.
Proceedings Article
Ye Zhao, Benito Fernandez, Luis Sentis
Robotics: Science and Systems XII, 2016
@inproceedings{zhao_robust_2016,
title = {Robust Phase-Space Planning for Agile Legged Locomotion over Various Terrain Topologies},
author = {Ye Zhao and Benito Fernandez and Luis Sentis},
url = {http://www.roboticsproceedings.org/rss12/p10.pdf
https://www.youtube.com/watch?v=F8uTHsqn1dc
http://www.google.com/url?q=http%3A%2F%2Frss2016.engin.umich.edu%2Fvideos%2Fslides%2Fp10.mp4&sa=D&sntz=1&usg=AFQjCNFZtBGp5GGkklaI3jYaOHGw-XWSiA},
doi = {10.15607/RSS.2016.XII.010},
isbn = {978-0-9923747-2-3},
year = {2016},
date = {2016-01-01},
urldate = {2016-01-01},
booktitle = {Robotics: Science and Systems XII},
publisher = {Robotics: Science and Systems Foundation},
abstract = {In this study, we present a framework for phasespace planning and control of agile bipedal locomotion while robustly tracking a set of non-periodic keyframes. By using a reduced-order model, we formulate a hybrid planning framework where the center-of-mass motion is constrained to a general surface manifold. This framework also proposes phase-space bundles to characterize robustness and a robust hybrid automaton to effectively design planning algorithms. A newly defined phasespace locomotion manifold is used as a Riemannian metric to measure the distance between the disturbed state and the planned manifold. Based on this metric, a dynamic programming based hybrid controller is introduced to produce robust locomotions. The robustness of the proposed framework is validated by using simulations of rough terrain locomotion recovery from external disturbances. Additionally, the agility of this framework is demonstrated by using simulations of the dynamic locomotion over random rough terrains.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this study, we present a framework for phasespace planning and control of agile bipedal locomotion while robustly tracking a set of non-periodic keyframes. By using a reduced-order model, we formulate a hybrid planning framework where the center-of-mass motion is constrained to a general surface manifold. This framework also proposes phase-space bundles to characterize robustness and a robust hybrid automaton to effectively design planning algorithms. A newly defined phasespace locomotion manifold is used as a Riemannian metric to measure the distance between the disturbed state and the planned manifold. Based on this metric, a dynamic programming based hybrid controller is introduced to produce robust locomotions. The robustness of the proposed framework is validated by using simulations of rough terrain locomotion recovery from external disturbances. Additionally, the agility of this framework is demonstrated by using simulations of the dynamic locomotion over random rough terrains.
Proceedings Article
Ye Zhao, Donghyun Kim, Gray Thomas, Luis Sentis
Proceedings of the 18th International Conference on Hybrid Systems: Computation and Control, 2015
@inproceedings{zhao_hybrid_2015,
title = {Hybrid multi-contact dynamics for wedge jumping locomotion behaviors},
author = {Ye Zhao and Donghyun Kim and Gray Thomas and Luis Sentis},
url = {https://dl.acm.org/doi/10.1145/2728606.2728645},
doi = {10.1145/2728606.2728645},
isbn = {978-1-4503-3433-4},
year = {2015},
date = {2015-04-01},
urldate = {2025-04-02},
booktitle = {Proceedings of the 18th International Conference on Hybrid Systems: Computation and Control},
pages = {293–294},
publisher = {ACM},
address = {Seattle Washington},
abstract = {Legged robots naturally exhibit continuous and discrete dynamics when maneuvering over level-ground and uneven terrains. In recent years, numerous studies have focused on locomotion hybrid dynamics. However, locomotion on more challenging terrains such as split wedges in Figure 1 has rarely been explored, let alone its hybrid dynamics. In this study, we specifically focus on a two-phase hybrid automaton formulation for this highly steep wedge locomotion. This automaton incorporates both multi-contact and flight single contact phase motions. To dynamically balance and jump upwards on this wedge, an aperiodic phase space planning is used for trajectory generations. Three control strategies are employed simultaneously: internal force control, linear and angular momentum control. Finally, simulation results are shown to verify our strategy’s effectiveness.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Legged robots naturally exhibit continuous and discrete dynamics when maneuvering over level-ground and uneven terrains. In recent years, numerous studies have focused on locomotion hybrid dynamics. However, locomotion on more challenging terrains such as split wedges in Figure 1 has rarely been explored, let alone its hybrid dynamics. In this study, we specifically focus on a two-phase hybrid automaton formulation for this highly steep wedge locomotion. This automaton incorporates both multi-contact and flight single contact phase motions. To dynamically balance and jump upwards on this wedge, an aperiodic phase space planning is used for trajectory generations. Three control strategies are employed simultaneously: internal force control, linear and angular momentum control. Finally, simulation results are shown to verify our strategy’s effectiveness.
Ye Zhao, Nicholas Paine, Luis Sentis
2014 IEEE-RAS International Conference on Humanoid Robots, 2014
@inproceedings{zhao_feedback_2014,
title = {Feedback parameter selection for impedance control of series elastic actuators},
author = {Ye Zhao and Nicholas Paine and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/7041485/},
doi = {10.1109/HUMANOIDS.2014.7041485},
isbn = {978-1-4799-7174-9},
year = {2014},
date = {2014-11-01},
urldate = {2025-04-03},
booktitle = {2014 IEEE-RAS International Conference on Humanoid Robots},
pages = {999–1006},
publisher = {IEEE},
address = {Madrid, Spain},
abstract = {The interest of series elastic actuators (SEAs) for legged robots has recently increased to achieve compliant interactions and efficient gaits. However, control of legged robots with SEAs is difficult due to the need to design controllers that take into account both torque and impedance feedback loops. The work presented here addresses this issue by proposing a critically-damped fourth order system gain selection criterion for a cascaded SEA control structure with inner torque and outer impedance feedback loops. Velocity filtering and feedback delays are taken into consideration for stability and impedance performance analysis. We observe and analyze the interdependence between torque and impedance feedback gains to achieve the desired closed loop performance. Based on this analysis we derive a simple gain design criterion to maximize the tracking performance of SEAs. Our final goal is to maximize the output impedance capabilities of SEAs in order to fulfill a wide range of application needs. In contrast to low impedance design studies, we focus here specifically on achieving the highest possible impedance gains of SEAs. Finally, experiments using our UTSEA are conducted to verify our proposed approach. This study serves as a stepping stone towards utilizing and designing humanoid robots with SEA actuators for mobile behaviors and interaction with cluttered and unknown environments.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The interest of series elastic actuators (SEAs) for legged robots has recently increased to achieve compliant interactions and efficient gaits. However, control of legged robots with SEAs is difficult due to the need to design controllers that take into account both torque and impedance feedback loops. The work presented here addresses this issue by proposing a critically-damped fourth order system gain selection criterion for a cascaded SEA control structure with inner torque and outer impedance feedback loops. Velocity filtering and feedback delays are taken into consideration for stability and impedance performance analysis. We observe and analyze the interdependence between torque and impedance feedback gains to achieve the desired closed loop performance. Based on this analysis we derive a simple gain design criterion to maximize the tracking performance of SEAs. Our final goal is to maximize the output impedance capabilities of SEAs in order to fulfill a wide range of application needs. In contrast to low impedance design studies, we focus here specifically on achieving the highest possible impedance gains of SEAs. Finally, experiments using our UTSEA are conducted to verify our proposed approach. This study serves as a stepping stone towards utilizing and designing humanoid robots with SEA actuators for mobile behaviors and interaction with cluttered and unknown environments.
Proceedings Article
Donghyun Kim, Ye Zhao, Gray Thomas, Luis Sentis
Dynamic Systems and Control Conference, 2014
@inproceedings{kim_empirical_2014,
title = {Empirical Modifications to a Phase Space Planner Which Compensates for Low Stiffness Actuation in a Planar, Point-Foot, Biped Robot},
author = {Donghyun Kim and Ye Zhao and Gray Thomas and Luis Sentis},
url = {https://asmedigitalcollection.asme.org/DSCC/proceedings/DSCC2014/46186/San%20Antonio,%20Texas,%20USA/228413},
doi = {10.1115/DSCC2014-6033},
isbn = {978-0-7918-4618-6},
year = {2014},
date = {2014-10-01},
urldate = {2014-10-01},
booktitle = {Dynamic Systems and Control Conference},
pages = {V001T11A001},
publisher = {American Society of Mechanical Engineers},
address = {San Antonio, Texas, USA},
abstract = {This paper presents an extensive experimental study of the first steps of the Hume robot. Hume is an adult sized, 20 kg, series-elastic, point-foot biped robot capable of very fast leg movements. In this study, Hume is constrained to planar motion by a linkage mechanism. We present our application of phase space planning to one, two, and three step walking, the last one over an obstacle. In the implementation, we modified the original theory and added ad-hoc adjustments since the robot could not follow the original theory’s planned walking trajectories despite their theoretical stability. We present a good correlation between the phase space plans and our various experiments, and an analysis of the robot’s final behavior. Overall the planner and ad-hoc modifications allowed us to execute very smooth gaits even over non-flat surfaces but at the same time demonstrated the shortcomings of open loop techniques.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper presents an extensive experimental study of the first steps of the Hume robot. Hume is an adult sized, 20 kg, series-elastic, point-foot biped robot capable of very fast leg movements. In this study, Hume is constrained to planar motion by a linkage mechanism. We present our application of phase space planning to one, two, and three step walking, the last one over an obstacle. In the implementation, we modified the original theory and added ad-hoc adjustments since the robot could not follow the original theory’s planned walking trajectories despite their theoretical stability. We present a good correlation between the phase space plans and our various experiments, and an analysis of the robot’s final behavior. Overall the planner and ad-hoc modifications allowed us to execute very smooth gaits even over non-flat surfaces but at the same time demonstrated the shortcomings of open loop techniques.
Proceedings Article
Ye Zhao, Nicholas Paine, Luis Sentis
Dynamic Systems and Control Conference, 2014
@inproceedings{zhao_sensitivity_2014,
title = {Sensitivity Comparison to Loop Latencies Between Damping Versus Stiffness Feedback Control Action in Distributed Controllers},
author = {Ye Zhao and Nicholas Paine and Luis Sentis},
url = {https://asmedigitalcollection.asme.org/DSCC/proceedings/DSCC2014/46186/San%20Antonio,%20Texas,%20USA/228296},
doi = {10.1115/DSCC2014-6207},
isbn = {978-0-7918-4618-6},
year = {2014},
date = {2014-10-01},
urldate = {2014-10-01},
booktitle = {Dynamic Systems and Control Conference},
pages = {V001T11A003},
publisher = {American Society of Mechanical Engineers},
address = {San Antonio, Texas, USA},
abstract = {This paper studies the effects of damping and stiffness feedback loop latencies on closed-loop system stability and performance. Phase margin stability analysis, step response performance and tracking accuracy are respectively simulated for a rigid actuator with impedance control. Both system stability and tracking performance are more sensitive to damping feedback than stiffness feedback latencies. Several comparative tests are simulated and experimentally implemented on a real-world actuator to verify our conclusion. This discrepancy in sensitivity motivates the necessity of implementing embedded damping, in which damping feedback is implemented locally at the low level joint controller. A direct benefit of this distributed impedance control strategy is the enhancement of closed-loop system stability. Using this strategy, feedback effort and thus closed-loop actuator impedance may be increased beyond the levels possible for a monolithic impedance controller. High impedance is desirable to minimize tracking error in the presence of disturbances. Specially, trajectory tracking accuracy is tested by a fast swing and a slow stance motion of a knee joint emulating NASA-JSC’s Valkyrie legged robot. When damping latencies are lowered beyond stiffness latencies, gravitational disturbance is rejected, thus demonstrating the accurate tracking performance enabled by a distributed impedance controller.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper studies the effects of damping and stiffness feedback loop latencies on closed-loop system stability and performance. Phase margin stability analysis, step response performance and tracking accuracy are respectively simulated for a rigid actuator with impedance control. Both system stability and tracking performance are more sensitive to damping feedback than stiffness feedback latencies. Several comparative tests are simulated and experimentally implemented on a real-world actuator to verify our conclusion. This discrepancy in sensitivity motivates the necessity of implementing embedded damping, in which damping feedback is implemented locally at the low level joint controller. A direct benefit of this distributed impedance control strategy is the enhancement of closed-loop system stability. Using this strategy, feedback effort and thus closed-loop actuator impedance may be increased beyond the levels possible for a monolithic impedance controller. High impedance is desirable to minimize tracking error in the presence of disturbances. Specially, trajectory tracking accuracy is tested by a fast swing and a slow stance motion of a knee joint emulating NASA-JSC’s Valkyrie legged robot. When damping latencies are lowered beyond stiffness latencies, gravitational disturbance is rejected, thus demonstrating the accurate tracking performance enabled by a distributed impedance controller.
Proceedings Article
Ye Zhao, Donghyun Kim, Benito R Fernandez, Luis Sentis
2013 13th IEEE-RAS International Conference on Humanoid Robots (Humanoids), 2013
@inproceedings{zhao_phase_2013,
title = {Phase space planning and robust control for data-driven locomotion behaviors},
author = {Ye Zhao and Donghyun Kim and Benito R Fernandez and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/7029959/},
doi = {10.1109/HUMANOIDS.2013.7029959},
isbn = {978-1-4799-2617-6 978-1-4799-2619-0},
year = {2013},
date = {2013-10-01},
urldate = {2013-10-01},
booktitle = {2013 13th IEEE-RAS International Conference on Humanoid Robots (Humanoids)},
pages = {80–87},
publisher = {IEEE},
address = {Atlanta, GA},
abstract = {We utilize here regression tools to plan dynamic locomotion in the Phase Space of the robot’s center of mass behavior and state feedback controllers to accomplish the desired plans. In real robotic systems, simplified locomotion models and disturbances in the control processes result in deviations from the actual closed loop dynamics with respect to the desired locomotion trajectories. To tackle these challenges, we propose here the use of two control strategies: (1) support vector regression to approximate complex nonlinear center of mass dynamics and plan the feet contact transitions, and (2) sliding mode control to track feet trajectories given the contact timing and location plans. First, support vector regression is utilized to learn a data set obtained through numerical simulation, providing an analytical approximation of the center of mass behavior. To approximate Phase Plane curves, which are characterized by vertical tangents and loop or cyclic behaviors, we use implicit functions for regression as opposed to explicit methods. Based on the proposed regression approximations of the dynamics, we develop contact transition plans and apply robust controllers to converge to the desired feet trajectories. In particular, state feedback controllers might be more convenient than time based controllers in terms of robustness to disturbances. Overall, our methods are capable of learning complex center of mass trajectories and might benefit from the use of robust control techniques. Various case studies are analyzed to validate the effectiveness of the methods including single and multi step planning in a numerical simulation, and swing leg trajectory control on our Hume bipedal robot.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We utilize here regression tools to plan dynamic locomotion in the Phase Space of the robot’s center of mass behavior and state feedback controllers to accomplish the desired plans. In real robotic systems, simplified locomotion models and disturbances in the control processes result in deviations from the actual closed loop dynamics with respect to the desired locomotion trajectories. To tackle these challenges, we propose here the use of two control strategies: (1) support vector regression to approximate complex nonlinear center of mass dynamics and plan the feet contact transitions, and (2) sliding mode control to track feet trajectories given the contact timing and location plans. First, support vector regression is utilized to learn a data set obtained through numerical simulation, providing an analytical approximation of the center of mass behavior. To approximate Phase Plane curves, which are characterized by vertical tangents and loop or cyclic behaviors, we use implicit functions for regression as opposed to explicit methods. Based on the proposed regression approximations of the dynamics, we develop contact transition plans and apply robust controllers to converge to the desired feet trajectories. In particular, state feedback controllers might be more convenient than time based controllers in terms of robustness to disturbances. Overall, our methods are capable of learning complex center of mass trajectories and might benefit from the use of robust control techniques. Various case studies are analyzed to validate the effectiveness of the methods including single and multi step planning in a numerical simulation, and swing leg trajectory control on our Hume bipedal robot.
Ye Zhao, Luis Sentis
2012 12th IEEE-RAS International Conference on Humanoid Robots (Humanoids 2012), 2012
@inproceedings{zhao_three_2012,
title = {A three dimensional foot placement planner for locomotion in very rough terrains},
author = {Ye Zhao and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/6651600/},
doi = {10.1109/HUMANOIDS.2012.6651600},
isbn = {978-1-4673-1369-8},
year = {2012},
date = {2012-11-01},
urldate = {2025-04-03},
booktitle = {2012 12th IEEE-RAS International Conference on Humanoid Robots (Humanoids 2012)},
pages = {726–733},
publisher = {IEEE},
address = {Osaka, Japan},
abstract = {Maneuvering through 3D structures nimbly is pivotal to the advancement of legged locomotion. However, few methods have been developed that can generate 3D gaits in those terrains and fewer if none can be generalized to control dynamic maneuvers. In this study, foot placement planning for dynamic locomotion traversing irregular terrains is explored in three dimensional space. Given boundary values of the center of mass’ apexes during the gait, sagittal and lateral phase-plane trajectories are predicted based on multi-contact and inverted pendulum dynamics. To deal with the nonlinear dynamics of the contact motions and their dimensionality, we plan a geometric surface of motion beforehand and rely on numerical integration to solve the models. In particular, we combine multi-contact and prismatic inverted pendulum models to resolve feet transitions between steps, allowing to produce trajectory patterns similar to those observed in human locomotion. Our contributions lay in the following points: (1) the introduction of non planar surfaces to characterize the center of mass’ geometric behavior; (2) an automatic gait planner that simultaneously resolves sagittal and lateral feet placements; (3) the introduction of multi-contact dynamics to smoothly transition between steps in the rough terrains.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Maneuvering through 3D structures nimbly is pivotal to the advancement of legged locomotion. However, few methods have been developed that can generate 3D gaits in those terrains and fewer if none can be generalized to control dynamic maneuvers. In this study, foot placement planning for dynamic locomotion traversing irregular terrains is explored in three dimensional space. Given boundary values of the center of mass’ apexes during the gait, sagittal and lateral phase-plane trajectories are predicted based on multi-contact and inverted pendulum dynamics. To deal with the nonlinear dynamics of the contact motions and their dimensionality, we plan a geometric surface of motion beforehand and rely on numerical integration to solve the models. In particular, we combine multi-contact and prismatic inverted pendulum models to resolve feet transitions between steps, allowing to produce trajectory patterns similar to those observed in human locomotion. Our contributions lay in the following points: (1) the introduction of non planar surfaces to characterize the center of mass’ geometric behavior; (2) an automatic gait planner that simultaneously resolves sagittal and lateral feet placements; (3) the introduction of multi-contact dynamics to smoothly transition between steps in the rough terrains.
Proceedings Article
Lixian Zhang, Yu Leng, Lingjie Chen, Ye Zhao
Proceedings of the IFAC 18th world congress, 2011
@inproceedings{zhang_brl_2011,
title = {A BRL for A Class of Discrete-time Markov Jump Linear System with Piecewise-Constant TPs},
author = {Lixian Zhang and Yu Leng and Lingjie Chen and Ye Zhao},
url = {https://linkinghub.elsevier.com/retrieve/pii/S147466701645007X},
doi = {10.3182/20110828-6-IT-1002.01884},
year = {2011},
date = {2011-01-01},
urldate = {2025-04-03},
booktitle = {Proceedings of the IFAC 18th world congress},
volume = {44},
pages = {8699–8704},
abstract = {This paper aims to obtain a Bounded Real Lemma (BRL) for a class of Markov jump linear systems (MJLSs) with time-varying transition probabilities (TPs) in discrete-time domain. The time-varying character of TPs is considered as piecewise-constant and the variation of TP matrices is subject to average dwell time (ADT) switching, i.e., the number of switches in a finite interval is bounded and the average time between two consecutive switchings of TP matrices is not less than a constant. Combining the Lyapunov function approach and the linear matrix inequality technique, a BRL for the underlying system is derived in order to check whether the corresponding system is stochastically stable and has a guaranteed H∞ noiseattenuation performance index scheduled based on the variation of TP matrices. A numerical example is provided to demonstrate that the method obtained in this paper is a significant improvement over previous one.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper aims to obtain a Bounded Real Lemma (BRL) for a class of Markov jump linear systems (MJLSs) with time-varying transition probabilities (TPs) in discrete-time domain. The time-varying character of TPs is considered as piecewise-constant and the variation of TP matrices is subject to average dwell time (ADT) switching, i.e., the number of switches in a finite interval is bounded and the average time between two consecutive switchings of TP matrices is not less than a constant. Combining the Lyapunov function approach and the linear matrix inequality technique, a BRL for the underlying system is derived in order to check whether the corresponding system is stochastically stable and has a guaranteed H∞ noiseattenuation performance index scheduled based on the variation of TP matrices. A numerical example is provided to demonstrate that the method obtained in this paper is a significant improvement over previous one.
Proceedings Article
Huijun Gao, Jia You, Peng Shi, Lixian Zhang, Ye Zhao
Proceedings of the IFAC 18th world congress, 2011
@inproceedings{gao_stabilization_2011,
title = {Stabilization of Continuous-Time Markov Jump Linear Systems with Defective Statistics of Modes Transitions},
author = {Huijun Gao and Jia You and Peng Shi and Lixian Zhang and Ye Zhao},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1474667016450068},
doi = {10.3182/20110828-6-IT-1002.01710},
year = {2011},
date = {2011-01-01},
urldate = {2025-04-03},
booktitle = {Proceedings of the IFAC 18th world congress},
volume = {44},
pages = {8693–8698},
abstract = {This paper concerns the stabilization problem of a class of Markov jump linear system (MJLS) with defective statistics of modes transitions in the continuous-time domain. Differing from the recent separate studies on the so-called uncertain transition probabilities (TPs) and partially unknown TPs, the defective statistics about modes transitions in this study take the two situations into account in a composite way. The scenario is more practicable in that it divides the TPs into three sets: known, uncertain and unknown. The necessary and sufficient conditions for the stability and stabilization of the underlying system are obtained by fully using the properties of the transition rate matrix (TRM) and the convexity of uncertain domains. The monotonicity, in concern of the existence of the admissible stabilizing controller, is observed when the unknown elements become uncertain and the intervals of the uncertain ones become tighter. Numerical examples are provided to verify the theoretical findings.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper concerns the stabilization problem of a class of Markov jump linear system (MJLS) with defective statistics of modes transitions in the continuous-time domain. Differing from the recent separate studies on the so-called uncertain transition probabilities (TPs) and partially unknown TPs, the defective statistics about modes transitions in this study take the two situations into account in a composite way. The scenario is more practicable in that it divides the TPs into three sets: known, uncertain and unknown. The necessary and sufficient conditions for the stability and stabilization of the underlying system are obtained by fully using the properties of the transition rate matrix (TRM) and the convexity of uncertain domains. The monotonicity, in concern of the existence of the admissible stabilizing controller, is observed when the unknown elements become uncertain and the intervals of the uncertain ones become tighter. Numerical examples are provided to verify the theoretical findings.
On Stability of Markovian Jumping Neural Networks with both unknown and uncertain transition probabilities
Proceedings Article
Ye Zhao, Lixian Zhang
International Conference on Computational Intelligence and Vehicular System, 2010
@inproceedings{ye_zhao_stability_2010,
title = {On Stability of Markovian Jumping Neural Networks with both unknown and uncertain transition probabilities},
author = {Ye Zhao and Lixian Zhang},
year = {2010},
date = {2010-01-01},
booktitle = {International Conference on Computational Intelligence and Vehicular System},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
PhD Theses
Abdulaziz Shamsah
2024
@phdthesis{Shamsah2024,
title = {Safe Bipedal Locomotion and Navigation in Uncertain Environments},
author = {Abdulaziz Shamsah },
url = {https://hdl.handle.net/1853/76948},
year = {2024},
date = {2024-12-04},
urldate = {2024-12-04},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
PhD Thesis
Ye Zhao
2016
@phdthesis{zhao_planning_2016,
title = {A Planning and Control Framework for Humanoid Systems: Robust, Optimal, and Real-time Performance},
author = {Ye Zhao},
url = {https://repositories.lib.utexas.edu/items/554e9fce-99a2-41c7-9f5d-2d8a83701af7},
year = {2016},
date = {2016-01-01},
keywords = {},
pubstate = {published},
tppubtype = {phdthesis}
}
Masters Theses
Nathan Boyd
2022
@mastersthesis{boyd_task_2022,
title = {Task and Motion Planning with Behavior Trees for Locomotion and Manipulation},
author = {Nathan Boyd},
url = {https://repository.gatech.edu/entities/publication/acd241ac-adf7-4b6c-b328-c218c326f702},
doi = {https://hdl.handle.net/1853/72488},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
Masters Thesis
Jonas Warnke
2021
@mastersthesis{warnke_safety-guaranteed_2021,
title = {Safety-guaranteed Task Planning for Bipedal Navigation In Partially Observable Environments},
author = {Jonas Warnke},
url = {https://repository.gatech.edu/entities/publication/70cd064b-1c20-43fa-b541-7e90666cf700},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
Book Sections
Smarter Cyber Physical Systems: Enabling Methodologies and Applications
Book Section
Jesse Jiang, Ye Zhao, Samuel Coogan
Specification-Guided Safe Learning for Robotic Systems, 2024
@incollection{jesse_jiang_specification-guided_2024,
title = {Smarter Cyber Physical Systems: Enabling Methodologies and Applications},
author = {Jesse Jiang and Ye Zhao and Samuel Coogan},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
booktitle = {Specification-Guided Safe Learning for Robotic Systems},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Book Section
Ye Zhao, Luis Sentis
Stability, Control and Application of Time-delay Systems, 2019
@incollection{zhao_distributed_2019,
title = {Distributed impedance control of latency-prone robotic systems with series elastic actuation},
author = {Ye Zhao and Luis Sentis},
url = {https://linkinghub.elsevier.com/retrieve/pii/B9780128149287000020},
doi = {10.1016/B978-0-12-814928-7.00002-0},
isbn = {978-0-12-814928-7},
year = {2019},
date = {2019-01-01},
urldate = {2025-04-03},
booktitle = {Stability, Control and Application of Time-delay Systems},
pages = {23–51},
publisher = {Elsevier},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
Workshops
Socially Acceptable Bipedal Navigation: A Signal-Temporal-Logic- Driven Approach for Safe Locomotion
Workshop
Abdulaziz Shamsah, Ye Zhao
IROS Workshop on Social Robot Navigation: Advances and Evaluation, 2023
@workshop{shamsah_socially_2023,
title = {Socially Acceptable Bipedal Navigation: A Signal-Temporal-Logic- Driven Approach for Safe Locomotion},
author = {Abdulaziz Shamsah and Ye Zhao},
url = {http://arxiv.org/abs/2310.09969},
doi = {10.48550/arXiv.2310.09969},
year = {2023},
date = {2023-10-01},
urldate = {2023-10-01},
booktitle = {IROS Workshop on Social Robot Navigation: Advances and Evaluation},
publisher = {arXiv},
abstract = {Social navigation for bipedal robots remains relatively unexplored due to the highly complex, nonlinear dynamics of bipedal locomotion. This study presents a preliminary exploration of social navigation for bipedal robots in a human crowded environment. We propose a social path planner that ensures the locomotion safety of the bipedal robot while navigating under a social norm. The proposed planner leverages a conditional variational autoencoder architecture and learns from human crowd datasets to produce a socially acceptable path plan. Robot-specific locomotion safety is formally enforced by incorporating signal temporal logic specifications during the learning process. We demonstrate the integration of the social path planner with a model predictive controller and a lowlevel passivity controller to enable comprehensive full-body joint control of Digit in a dynamic simulation.},
note = {arXiv:2310.09969 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Social navigation for bipedal robots remains relatively unexplored due to the highly complex, nonlinear dynamics of bipedal locomotion. This study presents a preliminary exploration of social navigation for bipedal robots in a human crowded environment. We propose a social path planner that ensures the locomotion safety of the bipedal robot while navigating under a social norm. The proposed planner leverages a conditional variational autoencoder architecture and learns from human crowd datasets to produce a socially acceptable path plan. Robot-specific locomotion safety is formally enforced by incorporating signal temporal logic specifications during the learning process. We demonstrate the integration of the social path planner with a model predictive controller and a lowlevel passivity controller to enable comprehensive full-body joint control of Digit in a dynamic simulation.
Workshop
Kelin Yu, Yunhai Han, Matthew Zhu, Ye Zhao
NeurIPS Workshop on Touch Processing: A New Sensing Modality for AI, 2023
@workshop{yu_touch_2023,
title = {Touch Insight Acquisition: Human’s Insertion Strategies Learned by Multi-Modal Tactile Feedback},
author = {Kelin Yu and Yunhai Han and Matthew Zhu and Ye Zhao},
url = {https://www.touchprocessing.org/2023/camera_ready/camera_ready_9.pdf},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
booktitle = {NeurIPS Workshop on Touch Processing: A New Sensing Modality for AI},
abstract = {In the evolving landscape of robotics and automation, the application of touch processing is crucial, particularly in learning to execute intricate tasks like insertion. However, existing works focusing on tactile methods for insertion tasks predominantly rely on sensor data and do not utilize the rich insights provided by human tactile feedback. For utilizing human sensations, methodologies related to learning from humans predominantly leverage visual feedback, often overlooking the invaluable tactile insights that humans inherently employ to finish complex manipulations. Addressing this gap, we introduce "MimicTouch", a novel framework that mimics a human’s tactile-guided control strategy. In this framework, we initially collect multi-modal tactile datasets from human demonstrators, incorporating human tactile-guided control strategies for task completion. The subsequent step involves instructing robots through imitation learning using multi-modal sensor data and retargeted human motions. To further mitigate the embodiment gap between humans and robots, we employ online residual reinforcement learning on the physical robot. Through comprehensive experiments, we validate the safety of MimicTouch in transferring a latent policy learned through imitation learning from human to robot. This ongoing work will pave the way for a broader spectrum of tactile-guided robotic applications.},
note = {Best Paper Award},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
In the evolving landscape of robotics and automation, the application of touch processing is crucial, particularly in learning to execute intricate tasks like insertion. However, existing works focusing on tactile methods for insertion tasks predominantly rely on sensor data and do not utilize the rich insights provided by human tactile feedback. For utilizing human sensations, methodologies related to learning from humans predominantly leverage visual feedback, often overlooking the invaluable tactile insights that humans inherently employ to finish complex manipulations. Addressing this gap, we introduce "MimicTouch", a novel framework that mimics a human’s tactile-guided control strategy. In this framework, we initially collect multi-modal tactile datasets from human demonstrators, incorporating human tactile-guided control strategies for task completion. The subsequent step involves instructing robots through imitation learning using multi-modal sensor data and retargeted human motions. To further mitigate the embodiment gap between humans and robots, we employ online residual reinforcement learning on the physical robot. Through comprehensive experiments, we validate the safety of MimicTouch in transferring a latent policy learned through imitation learning from human to robot. This ongoing work will pave the way for a broader spectrum of tactile-guided robotic applications.
Bridge Mixed-Integer Convex Program and Linear Temporal Logic: Reactive Gait Synthesis and Footstep Planning for Agile Perceptive Locomotion
Workshop
Ziyi Zhou, Eohan George, Ye Zhao
IROS Workshop on Formal Methods Techniques in Robotics Systems: Design and Control, 2023
@workshop{zhou_bridge_2023,
title = {Bridge Mixed-Integer Convex Program and Linear Temporal Logic: Reactive Gait Synthesis and Footstep Planning for Agile Perceptive Locomotion},
author = {Ziyi Zhou and Eohan George and Ye Zhao},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
booktitle = {IROS Workshop on Formal Methods Techniques in Robotics Systems: Design and Control},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Interval MDPs for Optimal Robotic Motion Planning with Temporal Logic Constraints
Workshop
Jesse Jiang, Samuel Coogan, Ye Zhao
IROS Workshop on Formal Methods Techniques in Robotics Systems: Design and Control, 2023
@workshop{jiang_interval_2023,
title = {Interval MDPs for Optimal Robotic Motion Planning with Temporal Logic Constraints},
author = {Jesse Jiang and Samuel Coogan and Ye Zhao},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
booktitle = {IROS Workshop on Formal Methods Techniques in Robotics Systems: Design and Control},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Workshop
Zhaoyuan Gu, Nathan Boyd, Ye Zhao
ICRA Workshop on Legged Robots, 2022
@workshop{gu_reactive_2022,
title = {Reactive Locomotion Decision-Making and Robust Motion Planning for Real-Time Perturbation Recovery},
author = {Zhaoyuan Gu and Nathan Boyd and Ye Zhao},
url = {http://arxiv.org/abs/2110.03037
https://www.youtube.com/watch?v=3ucH3Rflvwo},
doi = {10.48550/arXiv.2110.03037},
year = {2022},
date = {2022-03-01},
urldate = {2022-03-01},
booktitle = {ICRA Workshop on Legged Robots},
publisher = {arXiv},
abstract = {In this paper, we examine the problem of push recovery for bipedal robot locomotion and present a reactive decision-making and robust planning framework for locomotion resilient to external perturbations. Rejecting perturbations is an essential capability of bipedal robots and has been widely studied in the locomotion literature. However, adversarial disturbances and aggressive turning can lead to negative lateral step width (i.e., crossed-leg scenarios) with unstable motions and self-collision risks. These motion planning problems are computationally difficult and have not been explored under a hierarchically integrated task and motion planning method. We explore a planning and decision-making framework that closely ties linear-temporal-logic-based reactive synthesis with trajectory optimization incorporating the robot’s full-body dynamics, kinematics, and leg collision avoidance constraints. Between the high-level discrete symbolic decision-making and the lowlevel continuous motion planning, behavior trees serve as a reactive interface to handle perturbations occurring at any time of the locomotion process. Our experimental results show the efficacy of our method in generating resilient recovery behaviors in response to diverse perturbations from any direction with bounded magnitudes.},
note = {arXiv:2110.03037 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
In this paper, we examine the problem of push recovery for bipedal robot locomotion and present a reactive decision-making and robust planning framework for locomotion resilient to external perturbations. Rejecting perturbations is an essential capability of bipedal robots and has been widely studied in the locomotion literature. However, adversarial disturbances and aggressive turning can lead to negative lateral step width (i.e., crossed-leg scenarios) with unstable motions and self-collision risks. These motion planning problems are computationally difficult and have not been explored under a hierarchically integrated task and motion planning method. We explore a planning and decision-making framework that closely ties linear-temporal-logic-based reactive synthesis with trajectory optimization incorporating the robot’s full-body dynamics, kinematics, and leg collision avoidance constraints. Between the high-level discrete symbolic decision-making and the lowlevel continuous motion planning, behavior trees serve as a reactive interface to handle perturbations occurring at any time of the locomotion process. Our experimental results show the efficacy of our method in generating resilient recovery behaviors in response to diverse perturbations from any direction with bounded magnitudes.
Human Instruction Following: Graph Neural Network Guided Object Navigation
Workshop
Hongyi Chen, Letian Wang, Yuhang Yao, Ye Zhao, Patricio A. Vela
CVPR Workshop in Embodied AI, 2022
@workshop{chen_human_2022,
title = {Human Instruction Following: Graph Neural Network Guided Object Navigation},
author = {Hongyi Chen and Letian Wang and Yuhang Yao and Ye Zhao and Patricio A. Vela},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
booktitle = {CVPR Workshop in Embodied AI},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Leveraging Quadrupedal Robots in Heterogeneous Multi-Robot Teaming with Run-Time Disturbances
Workshop
Ziyi Zhou, Ye Zhao
IROS workshop in Decision-making in Multi-agent Systems, 2022
@workshop{zhou_leveraging_2022,
title = {Leveraging Quadrupedal Robots in Heterogeneous Multi-Robot Teaming with Run-Time Disturbances},
author = {Ziyi Zhou and Ye Zhao},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
booktitle = {IROS workshop in Decision-making in Multi-agent Systems},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Towards Data-Driven Contact Model Estimation using Inverse Optimization
Workshop
Luke Drnach, Ye Zhao
Proceedings of Dynamic Walking, 2021
@workshop{drnach_towards_2021,
title = {Towards Data-Driven Contact Model Estimation using Inverse Optimization},
author = {Luke Drnach and Ye Zhao},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Navigation in Dynamic Workspaces: Integrated Task and Motion Planning for Bipedal Locomotion
Workshop
Abdulaziz Shamsah, Jonas Warnke, Zhaoyuan Gu, Ye Zhao
Proceedings of Dynamic Walking, 2021
@workshop{shamsah_navigation_2021,
title = {Navigation in Dynamic Workspaces: Integrated Task and Motion Planning for Bipedal Locomotion},
author = {Abdulaziz Shamsah and Jonas Warnke and Zhaoyuan Gu and Ye Zhao},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Momentum-Aware Planning Synthesis for Dynamic Legged Locomotion
Workshop
Ziyi Zhou, Bruce Wingo, Nathan Boyd, Seth Hutchinson, Ye Zhao
Proceedings of Dynamic Walking, 2021
@workshop{zhou_momentum-aware_2021,
title = {Momentum-Aware Planning Synthesis for Dynamic Legged Locomotion},
author = {Ziyi Zhou and Bruce Wingo and Nathan Boyd and Seth Hutchinson and Ye Zhao},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Can Chance-Constrained Contact Uncertainty Quantification Improve Feasibility of Robust Trajectory Optimization?
Workshop
John Z. Zhang, Luke Drnach, Ye Zhao
Proceedings of Dynamic Walking, 2021
@workshop{zhang_can_2021,
title = {Can Chance-Constrained Contact Uncertainty Quantification Improve Feasibility of Robust Trajectory Optimization?},
author = {John Z. Zhang and Luke Drnach and Ye Zhao},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Robust Task and Motion Planning for Agile Locomotion In The Now
Workshop
Ye Zhao
Proceedings of Dynamic Walking, 2020
@workshop{zhao_robust_2020,
title = {Robust Task and Motion Planning for Agile Locomotion In The Now},
author = {Ye Zhao},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Robust Trajectory Optimization for Safe Locomotion over Uncertain Terrain
Workshop
Luke Drnach, Ye Zhao
RSS Workshop on Robust Autonomy, 2020
@workshop{drnach_robust_2020,
title = {Robust Trajectory Optimization for Safe Locomotion over Uncertain Terrain},
author = {Luke Drnach and Ye Zhao},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {RSS Workshop on Robust Autonomy},
abstract = {Safe and autonomous locomotion for legged robots in real-world environments requires generating motion strategies that are robust to uncertainty in the terrain. Current trajectory optimization methods rely on specifying the geometry and friction properties of the terrain; however, errors in the terrain model can lead to failure through slipping and falling. Here we develop a trajectory optimization approach that explicitly incorporates parametric uncertainty in the terrain model. We demonstrate that our method produces a spectrum of robust trajectories: the method produces robust trajectories when uncertainty is large and the nominal optimal trajectories when uncertainty is small. Our study represents a step towards generating safe locomotion behaviors which are robust against uncertainty in the terrain.},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Safe and autonomous locomotion for legged robots in real-world environments requires generating motion strategies that are robust to uncertainty in the terrain. Current trajectory optimization methods rely on specifying the geometry and friction properties of the terrain; however, errors in the terrain model can lead to failure through slipping and falling. Here we develop a trajectory optimization approach that explicitly incorporates parametric uncertainty in the terrain model. We demonstrate that our method produces a spectrum of robust trajectories: the method produces robust trajectories when uncertainty is large and the nominal optimal trajectories when uncertainty is small. Our study represents a step towards generating safe locomotion behaviors which are robust against uncertainty in the terrain.
Robust Locomotion Navigation in Partially Observable Environments with Safety Guarantees
Workshop
Jonas Warnke, Abdulaziz Shamsah, Yingke Li, Samuel Coogan, Ye Zhao
RSS Workshop on Robust Autonomy, 2020
@workshop{warnke_robust_2020,
title = {Robust Locomotion Navigation in Partially Observable Environments with Safety Guarantees},
author = {Jonas Warnke and Abdulaziz Shamsah and Yingke Li and Samuel Coogan and Ye Zhao},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {RSS Workshop on Robust Autonomy},
abstract = {This study is working towards an integrated task and motion planning method for dynamic locomotion in partially observable environments with safety guarantees. This planning framework is composed of a symbolic task planner and a reduced-order-model-based motion planner, which are connected by a mid-level keyframe decision-maker as seen in Fig. 2. The mid-level keyframe decision maker generates a keyframe plan via reachability analysis and proposes a robust keyframe policy, which is used to generate low-level phase-space trajectories. The high-level task planner employs a linear temporal logic approach for a reactive game synthesis between the robot and its environment while incorporating the robust keyframe transition policies into the formal task specification design. A belief abstraction method in the task planner enables belief estimation of dynamic obstacle locations and guarantees safe locomotion with collision avoidance.},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
This study is working towards an integrated task and motion planning method for dynamic locomotion in partially observable environments with safety guarantees. This planning framework is composed of a symbolic task planner and a reduced-order-model-based motion planner, which are connected by a mid-level keyframe decision-maker as seen in Fig. 2. The mid-level keyframe decision maker generates a keyframe plan via reachability analysis and proposes a robust keyframe policy, which is used to generate low-level phase-space trajectories. The high-level task planner employs a linear temporal logic approach for a reactive game synthesis between the robot and its environment while incorporating the robust keyframe transition policies into the formal task specification design. A belief abstraction method in the task planner enables belief estimation of dynamic obstacle locations and guarantees safe locomotion with collision avoidance.
Ye Zhao, Jonathan S. Matthis, Sean L. Barton, Mary Hayhoe, Luis Sentis
2017 IEEE Workshop on Advanced Robotics and its Social Impacts (ARSO), 2017
@workshop{zhao_towards_2017,
title = {Towards understanding visually guided locomotion over complex and rough terrain: A phase-space planning method},
author = {Ye Zhao and Jonathan S. Matthis and Sean L. Barton and Mary Hayhoe and Luis Sentis},
url = {http://ieeexplore.ieee.org/document/8025198/},
doi = {10.1109/ARSO.2017.8025198},
isbn = {978-1-5090-0475-1},
year = {2017},
date = {2017-03-01},
urldate = {2017-03-01},
booktitle = {2017 IEEE Workshop on Advanced Robotics and its Social Impacts (ARSO)},
pages = {1–3},
publisher = {IEEE},
address = {Austin, TX, USA},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Towards Formal Planner Synthesis of Unified Legged and Armed Dynamic Locomotion in Constrained Environments
Workshop
Ye Zhao, Ufuk Topcu, Luis Sentis
Proceedings of Dynamic Walking, 2016
@workshop{zhao_towards_2016,
title = {Towards Formal Planner Synthesis of Unified Legged and Armed Dynamic Locomotion in Constrained Environments},
author = {Ye Zhao and Ufuk Topcu and Luis Sentis},
year = {2016},
date = {2016-01-01},
urldate = {2016-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Robust Optimal Control and High-Level Planner Synthesis for Locomotion over Various Terrain Topologies
Workshop
Ye Zhao, Luis Sentis
ICRA Workshop on Legged Robot Falling: Fall Detection, Damage Prevention, and Recovery Actions, 2016
@workshop{ye_zhao_robust_2016,
title = {Robust Optimal Control and High-Level Planner Synthesis for Locomotion over Various Terrain Topologies},
author = {Ye Zhao and Luis Sentis},
year = {2016},
date = {2016-01-01},
urldate = {2016-01-01},
booktitle = {ICRA Workshop on Legged Robot Falling: Fall Detection, Damage Prevention, and Recovery Actions},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Exploring Visually Guided Locomotion over Rough Terrain: A Phase Space Planning Method
Workshop
Ye Zhao, Jonathan Matthis, Sean L. Barton, Mary Hayhoe, Luis Sentis
Proceedings of Dynamic Walking, 2016
@workshop{ye_zhao_exploring_2016,
title = {Exploring Visually Guided Locomotion over Rough Terrain: A Phase Space Planning Method},
author = {Ye Zhao and Jonathan Matthis and Sean L. Barton and Mary Hayhoe and Luis Sentis},
year = {2016},
date = {2016-01-01},
urldate = {2016-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Towards Formal Planner Synthesis for Unified Legged and Armed Locomotion in Constrained Environments
Workshop
Ye Zhao, Ufuk Topcu, Luis Sentis
Proceedings of Dynamic Walking, 2016
@workshop{zhao_towards_2016-1,
title = {Towards Formal Planner Synthesis for Unified Legged and Armed Locomotion in Constrained Environments},
author = {Ye Zhao and Ufuk Topcu and Luis Sentis},
year = {2016},
date = {2016-01-01},
urldate = {2016-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Reactive Task Planner Synthesis of Dynamic Multi-Contact Locomotion in Constrained Environments
Workshop
Ye Zhao, Ufuk Topcu, Luis Sentis
RSS 2016 Workshop on Task and Motion Planning, 2016
@workshop{ye_zhao_reactive_2016,
title = {Reactive Task Planner Synthesis of Dynamic Multi-Contact Locomotion in Constrained Environments},
author = {Ye Zhao and Ufuk Topcu and Luis Sentis},
year = {2016},
date = {2016-01-01},
urldate = {2016-01-01},
booktitle = {RSS 2016 Workshop on Task and Motion Planning},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Whole-Body Operational Space Control: Undirectional Walking of a Point-Foot Series Elastic Biped
Workshop
Donghyun Kim, Ye Zhao, Gray Thomas, Luis Sentis
Proceedings of Dynamic Walking, 2015
@workshop{donghyun_kim_whole-body_2015,
title = {Whole-Body Operational Space Control: Undirectional Walking of a Point-Foot Series Elastic Biped},
author = {Donghyun Kim and Ye Zhao and Gray Thomas and Luis Sentis},
year = {2015},
date = {2015-01-01},
urldate = {2015-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}
Towards Agility in Compliant Point-Foot Bipeds
Workshop
Donghyun Kim, Ye Zhao, Gray Thomas, Luis Sentis
Proceedings of Dynamic Walking, 2014
@workshop{kim_towards_2014,
title = {Towards Agility in Compliant Point-Foot Bipeds},
author = {Donghyun Kim and Ye Zhao and Gray Thomas and Luis Sentis},
year = {2014},
date = {2014-01-01},
urldate = {2014-01-01},
booktitle = {Proceedings of Dynamic Walking},
keywords = {},
pubstate = {published},
tppubtype = {workshop}
}