A Study of the Effectiveness of Various Combined Control Schemes Based on MPC and WBC in Humanoid Control
摘要
The development of motion control frameworks for legged robots primarily focuses on Model Predictive Control (MPC) and Whole-Body Control (WBC) strategies, both of which have proven effectiveness in achieving dynamic locomotion for bipedal and quadrupedal robots. MPC offers predictive capabilities by optimizing control inputs over a future horizon, while WBC enables real-time task prioritization and joint torque generation. In this work, we design a WBC motion control scheme based on nullspace projection and compare it with an existing linear MPC-based controller and combined MPC-WBC approach. These schemes are evaluated in simulation with respect to maximum walking speed, trajectory tracking accuracy, and disturbance rejection capability. The WBC scheme is shown to achieve faster movement (up to 0.9 m/s) and superior robustness against external perturbations compared to the MPC-only controller (up to 0.3m/s). Moreover, it significantly reduces computational complexity relative to the combined MPC-WBC framework, making it more suitable for real-time or resource-limited applications.