Hierarchical Optimization Based Loco-Manipulation Control for Quadruped Robot Carrying Heavy Payload

In recent years, there has been significant interest in the legged robotics community in enhancing the manipulation capabilities of quadruped robots. However, unlike fixed-base robots, quadruped robots are inherently floating-base systems, and manipulating heavy objects introduces substantial disturbances that affect their motion. This paper proposes a hierarchical optimization-based control framework for quadruped robot loco-manipulation that accounts for whole-body coordination. First, an online manipulation planner computes the manipulation force and the reference trajectory for the manipulated object. Next, the robot's trajectory is optimized for pose using kinematic constraints. The resulting reference trajectory is then executed by a Model Predictive Controller (MPC) that incorporates the required manipulation force into its prediction model. Finally, a Whole-Body Controller (WBC) computes the joint torques based on the MPC's output. The effectiveness of the proposed method is validated in simulations, where a designed quadruped robot, equipped with a 1-DoF manipulator arm, successfully lifts and carries payloads of up to 6 kg.