Learning-Based Optimal Impedance Control for Space Manipulator Contact Tasks

This paper focuses on the contact control problem during operation tasks executed by space manipulator systems. First, a discounted value function is established to describe the interaction performance between the end-effector and the contact surface of object. Then, taking into account the effect of contact surface position, a model-free integral reinforcement learning (IRL) method with state feedback is applied to solve the optimal impedance parameters. Later, a state reconstruction technique with immunity to control noise is developed, whereby a novel model-free IRL algorithm is proposed which obviates the reliance on the velocity measurement and contact dynamics knowledge. Numerical simulations reveal the effectiveness of the proposed algorithm in space contact tasks.