Nearly optimal stabilization of unknown continuous-time nonlinear systems: A new parallel control approach

This paper develops a novel online nearly optimal control (ONOC) method for unknown continuous-time (CT) nonaffine nonlinear systems without recovering unknown systems. First, a dynamic control law is proposed for CT nonaffine nonlinear systems using parallel control. To achieve the proposed dynamic control law, an affine augmented system (AAS) is constructed according to the original system, and an augmented performance index (API) is constructed on the basis of the original performance index (OPI). Then, the stability relationship between the original system and the AAS is provided, and it is proven that, by selecting a suitable parameter in the API, optimal control of the AAS with the API is equivalent to near-optimal control of the original system with the OPI. Subsequently, based on the proposed dynamic control law, we extend integral reinforcement learning (IRL) to completely unknown CT nonaffine systems, and it is further proved that closed-loop signals are uniformly ultimately bounded (UUB) without the assumption that the input dynamics are bounded. Furthermore, the OPI can be set to an arbitrary positive-definite form, and the UUB bound for the state vector can be predetermined. Lastly, simulations are offered to exhibit the correctness of the developed ONOC method. Source code of this paper is available at: https://github.com/lujingweihh/Adaptive-dynamic-programming-algorithms/tree/main/model_free_integral_reinforcement_learning.