Optimized command filtered backstepping asymptotic tracking control for a class of nonlinear systems with unknown dynamic functions

Abstract This article addresses the challenge of achieving optimal asymptotic tracking control for nonlinear systems characterized by unidentified dynamic functions. The reinforcement learning approach is employed to design an optimized control strategy, using neural network approximation within an identifier‐critic‐actor framework to derive the reinforcement learning update law. The identifier component models the unknown system dynamics, the critic assesses the effectiveness of the control policy, and the actor generates the control actions, collectively ensuring robust and adaptive control. By integrating command filters into the backstepping design, the need for repeated differentiations of virtual control signals is eliminated, enhancing the efficiency of the control process. The proposed approach guarantees the asymptotic convergence of tracking errors to zero, offering a robust solution for precise control, ensuring precise and stable tracking performance even under disturbance conditions. The performance of this approach is demonstrated through two comprehensive simulation examples.