Tensor product model transformation based integral sliding mode control with reinforcement learning strategy

This paper presents a new chattering elimination method, and an optimal adaptive integral sliding mode controller design based on reinforcement learning for translational oscillations by a rotational actuator (TORA) system is demonstrated. At first, we introduce the tensor product model transformation based adaptive integral sliding mode controller. Next, we utilize an adaptive boundary layer width saturation function to get better performance. Reinforcement learning algorithm is employed to find the instantaneous optimal value for the boundary layer width of saturation function appeared in the adaptive integral sliding mode controller. The proposed tensor product model transformation based adaptive integral sliding mode controller with reinforcement learning strategy is verified by TORA system whereas the agent is rewarded for lower chattering, and punished for higher chattering. Simulation results show that chattering can be reduced effectively by incorporating reinforcement learning strategy into the adaptive integral sliding mode controller.