The objective of this research project is to design closed-loop perception, navigation, task and motion planners for dynamic-legged robots to autonomously navigate complex indoor environments while accomplishing specific tasks such as office disinfection. In the long-term run, a multi-agent cooperation scenario involving both quadrupedal and bipedal robots will be designed in both simulation and the mock indoor facility to validate the feasibility for real-life application scenarios. A team of legged robots will be able to autonomously navigate through various types of indoor terrain to disinfect appropriate objects while keeping humans in the loop through front-end user interfaces. [More details upcoming!]  

Unitree A1 quadruped with Slamtec 2D lidar sensor (left) & Generated 2D static global map from GTMI lounge area (right)
Navigation demo to a target location while demonstrating global path planning & local collision avoidance capabilities

———————————————————————————————————————-

A conceptual illustration of a three robots (Unitree A1, UBTECH DR, UBTECH Wassi) performing tasks in a simulated hospital setting. An illustration of team automaton is shown in the top left.
Topological map of a hospital environment consisting of locations that are explicitly defined in each robot’s transition system.
Operating state diagram for legged robot capable of both delivery and training task.

Representative publications:

Ziyi Zhou, Dong Jae Lee, Yuki Yoshinaga, Dejun Guo, and Ye Zhao. Reactive Task Allocation and Planning of A Heterogeneous Multi-Robot System, Submitted, 2021.