Disturbance-Aware Model Predictive Control
Real time robust MPC for underactuated robots subject to unmatched disturbances.
Real time robust MPC for underactuated robots subject to unmatched disturbances.
Overview
Disturbance-Aware MPC improves the robustness of model predictive control for underactuated robotic systems. A nonlinear disturbance observer estimates unknown disturbances online, and the controller explicitly incorporates those estimates into its future state predictions.
The approach works with standard optimal control solvers and is suitable for real time control. Its effectiveness was demonstrated on an underactuated quadrotor and compared with L1 adaptive MPC in simulation.
Video
Publication
J. Kim and M. J. Kim, “Disturbance-Aware Model Predictive Control of Underactuated Robotics Systems,” IEEE/RSJ IROS, 2024. IEEE