Abstract
This paper investigates the fault tolerant control problem for a class of continuous-time nonlinear systems with completely unknown dynamics via the data-based adaptive dynamic programming method. The proposed controller can be divided into two parts: (1) optimal control policy of the fault-free systems and (2) fault compensator. Firstly, a model-based policy iteration algorithm is introduced to obtain the optimal control law. Subsequently, a fault compensator is derived to get rid of the impact of the actuator fault. The stability analysis of the model-based control scheme is presented by using Lyapunov theory. However, for the complex practical systems, system models are generally unavailable, and thus the model-based approaches may be invalid. To overcome this difficulty, we provide a data-driven reinforcement learning method and an identification approach to design the two parts of the proposed controller, respectively, without any knowledge of the system models. Neural networks are employed to implement these two data-based methods. Finally, two simulation examples are shown in details to demonstrate the effectiveness of our proposed scheme.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have