Impedance control of slag removal robot based on Q-Learning

Due to the time-varying contact force of molten metal, it is a challenging work to realize the accurate slag pushing of industrial robot. In order to solve this problem, this paper proposes an impedance control method based on reinforcement learning. Firstly, the impedance model of the industrial robot system is established, and the Maxwell stress model which describes the contact between the end-effector of the manipulator and the molten metal surface is established. Then, on the basis of the idea of Q-learning and system linearization, an online correction strategy of impedance parameters based on reinforcement learning is designed. At the same time, the stability of the proposed method is analyzed theoretically. Finally, the proposed intelligent impedance control strategy is successfully applied to the slag pushing simulation of Rokae robot. And compared with the existing control methods, the superiority of the control method is proved to a certain extent.