Event-triggered Robust Adaptive Dynamic Programming Control For Nonlinear Continuous System

For nonlinear continuous systems based on the network,compared with the traditional time-triggered periodic sampling mechanism,in order to reduce the amount of calculation and communication transmission and make reasonable use of limited network bandwidth resources,this thesis proposes to apply the event-triggered control method to the network-based nonlinear continuous system.Non-periodic sampling and transmission enable an on-demand control mechanism.At the same time,disturbances in the network-based nonlinear continuous system are unavoidable,and it is a major difficulty to control the direction of how to solve the robust control optimization problem of the nonlinear continuous system.In the field of optimization control,adaptive dynamic programming(ADP)is a very effective optimization method.It not only overcomes the dimension-disaster problem in solving dynamic programming problems,but also can obtain the optimal controller based on the approximate properties of the neural.network.In the context of the network control system,for the problem of robust control of nonlinear continuous systems,this thesis studies the adaptive dynamic programming based on event-triggered.The research work done in this thesis is as follows:(1)Based on adaptive dynamic programming,the robust control problems of nonlinear continuous systems with matched and unmatched uncertainties are analyzed theoretically.At the same time,the adaptive dynamic programming method is also used to solve the H? optimal control problem of nonlinear continuous systems.In order to facilitate the understanding of the H?,problem,the related zero sum game is introduced,and the tracking control problem is briefly analyzed.This thesis also describes the event-triggered feedback in the control system,which lays a theoretical foundation for the following chapters.(2)For a class of robust optimal control problems with uncertain nonlinear continuous systems,this thesis presents a new algorithm based on event-triggered and adaptive critic-control learning method.The main focus is to combine the eventtriggered method with the adaptive identification-critic-control scheme to solve the robust optimal control problem for continuous systems with uncertain nonlinearities.Firstly,the unknown dynamics of nonlinear robust systems are reconstructed by neural network identification.Then the robust optimal control is implemented using the newly defined value function and optimal control law using the structure of the critic-control network;a new type of event-triggered condition is designed.The superior control law only updates when the set triggered condition is violated,thus,the computing and communication resources are reduced,and the uniformly ultimately boundedness of the closed-loop system is proved by using the Lyapunov method.At the same time,in order to avoid the accumulation of events(ie,the Zeno behavior).The minimum sampling time interval is given.Finally,the performance of the proposed eventtriggered robust control scheme is verified by simulation.(3)For the H? optimal tracking control problem of nonlinear continuous system,this paper proposes a new algorithm based on event-triggered.The novelty lies in combining the event-triggered method and adaptive dynamic programming to solve the H? tracking control optimization problem.First,in order to realize H? optimal tracking control,based on the augmented system and the reference system,the nonlinear H? optimal tracking control problem is transformed into a two-person zero-difference strategy using the event sampling state.By establishing a critic network,a time-triggered value function for the augmented system and a worst-case disturbance law are obtained.The H? optimal tracking controller is obtained by adjusting the weight of the control network and selecting the appropriate optimal discount factor.A novel adaptive event-triggered condition is designed so that the control network is updated only when the trigger condition is violated,resulting in a non-periodic adjustment relative tothe periodical form of the traditional time-triggered.In order to prove the stability of closed-loop system,an impulse power system framework was used and a robust term was added.At the same time,in order to avoid the Zeno behavior,the lower bound of the minimum sampling interval is determined,and the relationship between the sampling interval and the neural network estimation weight is given.Finally,three examples are given to verify the effectiveness of the algorithm.