Optimal Parallel Tracking Control of Unknown Time-Varying Nonlinear Systems Based on Adaptive Dynamic Programming

Abstract

This article investigates an optimal parallel tracking control problem for a class of nonaffine time-varying nonlinear systems (NATVNSs) under the unknown-model adaptive dynamic programming (UMADP) framework. First, a parallel control strategy is developed to address the tracking problem, which directly decouples the nonaffine characteristics by constructing an affine augmented system (AAS) and an augmented performance index (API). Second, to cope with the lack of an accurate system model, integral reinforcement learning (IRL) is extended to the constructed augmented system with completely unknown dynamics, thereby eliminating the reliance on model reconstruction. Third, an online learning strategy within the UMADP framework is proposed to achieve real-time optimal tracking control without assuming bounded input dynamics. Furthermore, rigorous theoretical analysis is carried out to prove that all closed-loop signals are uniformly ultimately bounded (UUB). Finally, the simulation results demonstrate that our proposed theoretical framework ensures effective attitude tracking of a morphing vehicle (MV), even under sweep-angle variations and unknown system dynamics.