Chance-Constrained Convex MPC for Quadrupedal Robots

In this video, we showcase our Chance-Constrained Model Predictive Control (CCMPC) framework, developed to enhance stability and control for quadrupedal robots navigating uncertain terrains and variable payloads. Our CCMPC approach models terrain and payload variations as uncertainties, optimizing ground reaction forces in real-time to keep the robot stable across challenging surfaces.
Through simulations and real-world experiments, our method demonstrates superior stability and adaptability compared to traditional Linear MPC. Watch as the Unitree Go1 robot maneuvers confidently across terrains like gravel, mud, grass, and stairs-even with added payloads exceeding 50% of its weight.
For more details and access to code, visit our project page: cc-mpc.github.io/.