Reinforcement Learning for Non-Affine Nonlinear Non-Minimum Phase System Tracking Under Additive-State-Decomposition-Based Control Framework

Abstract

This paper proposes a reinforcement-learning additive-state-decomposition-based tracking controller for a class of non-affine nonlinear non-minimum phase systems. Because the tracking performance is not satisfied with the model-based additive-state-decomposition tracking control with an approximate ideal internal model, two reinforcement learning schemes are introduced to improve the performance under the proposed additive-state-decomposition-based control framework. One is used to generate control commands, and the other is used to generate tracking reference commands. Finally, numerical simulations show the effectiveness of the proposed controller.