Adaptive data driven controller for nonlinear systems

A novel adaptive data driven control strategy is proposed for general discrete nonlinear systems. The parametric estimation is designed to be adaptive, which greatly improves the control ability. During the control process, the Simultaneous Perturbation Stochastic Approximation (SPSA) method is used to do the estimation of the control parameters, and the controller is fixed as a neural network here. In this paper, the proposed control strategy is applied to solve nonlinear tracking problems for discrete-time nonlinear systems, as well as nonlinear near-optimal control problems. The traditional model-free control strategy is introduced for comparison, and the feasibility and effectiveness of the proposed adaptive data driven control strategy is well demonstrated through simulation comparison results.