Adaptive Dynamic Programming-based Event-triggered Fault Tolerant Control For Nonlinear Systems

The dynamic programming in traditional control theory often encounters the difficulty of“Curse of Dimensionality” when solving optimal control problems.The difficulty greatly limits its application in parctice.By combining the idea of optimal control theory,artificial neural networks and reinforcement learning,adaptive dynamic programming(ADP)is an intelligent control method with self-learning and optimization abilities for dealing with approximate optimal control problems effectively.It is widely used in addressing the optimal control problem of complex systems with high nonlinearities and uncertainties.In the practical industrial environment,many controlled systems have higher dimensions and more complex sturctures.Then,the failures occurring on actuators,sensors and other components may lead to damage or even disasters.The acutator failure is one of the most common and serious failures in the controlled systems.Therefore,it is significant and meaningful to study the fault tolerant control for the continuous-time nonlinear systems subject to actuator failures.On the other hand,as one hotspot of modern control theorey research,eventtriggered control method is implemented by sampling system states and updating controller aperiodically according to the specific indicators such as system states,which has great advantages in reducing energy consumption and saving resources.This paper mainly studies the ADP-based event-triggered fault-tolerant control method for nonlinear systems.The main works and contributions of this thesis include the following aspects.(1)Considering nonlinear interconnected systems with unkonwn mismatched interconnections,this paper addresses the decentralized fualt-tolerant control problem subject to actuator bias failures,and proposes an ADP-based event-triggered decentralized optimal faulttolerant control method.First of all,for the dynamic models without failures,a series of eventtriggered decentralized optimal control strategies are proposed for isolated subsystems based on the idea of decentralized control,and critic neural networks are used to approximate the improved value functions of isolated subsystems.The neural network-based decentralized observer is established to approximate actuator failures and mismatched interconnections.The event-triggered decentralized optimal controller and the neural network weights in the observer are updated aperiodically determined by the designed triggerring condition.Then,the eventtriggered decentralized optimal fault-tolerant contro scheme is proposed by combining the event-triggered decentralized optimal controller and the decentralized observer.By using the Lyapunov stability analysis,it proves that all signals of the closed-loop system are ultimately uniformly bounded(UUB).Simulation results of two examples verify the effectiveness of the proposed event-triggered decentralized optimal fault-tolerant control scheme.(2)To handling the fault-tolerant control problem for nonlinear interconnected systems with unknown actuator saturation,an event-triggered adaptive critic fault-tolerant control method is proposed.The control scheme includes event-triggered decentralized optimal control strategy and neural network-based saturation compensator.For the nonlinear interconnected system without unknown input saturation,by combining the event-triggered control mechnism,the improved value function is approximated by the critic neural network,and then the Hamiltonian-Jacobi-Bellman(HJB)equation is solved iteratively to obtain the decentralized optimal control strategy.However,the input saturation nonlinearity is unknown in practice,it is impossible to directly solve the optimal control problem by constructing a value function with the information of saturation nonlinearity.A feedforward neural network is bulid to approximate the unknown saturation nonlinearity,and the output of neural network is employed to compensate the decentralized optimal control strategy.Theoretical analysis proves the stability of the closed-loop system,and the effectiveness of the proposed method is verified by the simulation results.Finally,the conclusion and the future research are presented at the end of this thesis.