Optimization Sliding Mode Controller Using Reinforcement Learning

This article considers the optimization using reinforcement learning to determine the sliding mode slope and sliding mode gain of a sliding mode level controller. The switching of the sliding mode level controller depends on the switching surface, which is determined by the error value and the rate of change of the system error value. The objective function of the reinforcement learning method is the value of the rising edge time and the integral time absolute error value. The reward for motivating reinforcement learning is learning with a sliding mode controller. The result is that the sliding mode controller, optimized through reinforcement learning, can be controlled to a specified value with minimal rise time and error. In summary, a reinforcement learning system outperforms the nonlinear system, resulting in more accurate trajectory tracking and fast response.