Load-adaptive control for quadruped mobile manipulation platforms

Abstract Currently, quadruped robots are widely used in diverse scenarios due to their high mobility, creating a demand for more advanced interaction capabilities. This study proposes a whole-body planning and control framework that integrates adaptive control into a hierarchical model predictive control (MPC) and whole-body control (WBC) structure, enhancing the environmental adaptability and interaction performance of quadruped mobile manipulators. Key innovations include: a recursive least squares and feedforward compensation strategy for accurate end-effector force estimation; relaxed barrier functions embedded in the MPC to combine dynamic obstacle avoidance with adaptive control; and a WBC-based priority hierarchy to enforce critical constraints. Validated in Gazebo simulation and on the B1-Z1 platform, the method allows the robot to handle unknown loads up to 3 kg and maintain tracking errors under 2 cm despite 35 N external disturbances. It also demonstrates strong adaptability in non-uniform object transportation, providing a reliable solution for unstructured environments.