Nonlinear Control of Partially Unknown Systems Based on ReinforcementLearning and Polynomial Representation

摘要

In this study we consider the problem of designing a controller for partially unknown plant, instead of unknown plants. Partially unknown plants are assumed to be consisting of a class of polynomial systems and systems whose mathematical model is unknown. For polynomial systems we give a stabilizing controller and the domain of attraction(DOA) by means of Lyapunov function. In the unknown space a reinforcement learning controller is synthesized by SARSA(λ)method. Reward of the reinforcement learning is given by the Lyapunov function which is derived in the synthesis of stabilizing controller of the polynomial system. Since the reinforcement learning is carried out by the Lyapunov function, the reinforcement learning controller can be connected to the stabilizing controller of the polynomial plant on the DOA. To demonstrate the usefulness and validity of the proposed method, we apply the proposed control method to the Furuta pendulum.