Admittance Control for Robot Polishing Force Tracking Based on Reinforcement Learning.

To achieve a stable interaction between the robot and the environment, stable force control of the robot is required. In this paper, a position-based impedance adaptive controller is proposed. The proposed method tracks the expected contact force based on estimating the parameters of the environment. Within this framework, we analyze environmental parameter estimation, introduce an adaptive algorithm based on reinforcement learning to adjust control parameters, and verify the stability of the system based on the Routh criterion and Lyapunov equation. The collaborative robot for workpieces with different surfaces is used for polishing. The controller is constructed to adjust the reference trajectory and use reinforcement learning training to adaptively adjust the control parameters to reduce the force error. Polishing experiments were carried out on the UR16e robot, and constant force control was carried out on beveled and curved workpieces. The proposed method improved the tracking accuracy of the robot polishing task.