Decentralized Multi-Biped Control
Developed a decentralized control system for multi-biped robot carriers, enabling payload transport over rough terrain using bipedal robots.
Designed and implemented a reinforcement learning-based controller that adapts to varying numbers and configurations of rigidly attached robots without retraining.
Conducted extensive simulation experiments and demonstrated successful real-world tests with Cassie robots.
Achieved a scalable solution for multi-biped transport systems, marking a first in the field for handling challenging terrains with legged robots.
Above image shows the decMBC controller structure
The project webpage contains the arXiv paper link and the video for the project.