Reinforcement learning‐based adaptive predefined‐time optimal tracking control for strict‐feedback nonlinear systems

Abstract

SummaryIn this paper, an adaptive predefined‐time optimal tracking control method is presented for a class of strict‐feedback nonlinear systems. Different from the existing optimal control results, these reinforcement learning methods only ensure the system is semiglobally uniformly ultimately bounded or finite‐time stable, which cannot achieve the precise control of the convergence time and accuracy. The designed predefined‐time optimal control strategy can guarantee that the convergence time and error accuracy can be predefined by users, while optimizing the performance index function by constructing the identifier‐actor‐critic neural networks. Meanwhile, an improved piecewise continuous function is devised to solve the problem of containing the sign function in the controller under the predefined‐time stability framework, which can both decrease the chattering phenomenon and avoid the possible singularity problem. Moreover, it can be demonstrated that all signals within the closed‐loop systems are predefined‐time stable. Finally, some simulation results confirm the effectiveness of the proposed control strategy.