离散非线性系统的事件驱动最优控制  

Event-triggered optimal control for discrete-time nonlinear systems

作　　者：张欣 薄迎春 ZHANG Xin;BO Yingchun(College of Information and Control Engineering,China University of Petroleum,Qingdao 266580,China)

机构地区：[1]中国石油大学信息与控制工程学院,山东青岛266580

出　　处：《沈阳师范大学学报（自然科学版）》2018年第4期318-323,共6页Journal of Shenyang Normal University:Natural Science Edition

基　　金：国家自然科学基金资助项目(61573378);山东省自然科学基金资助项目(BS2015DX009)

摘　　要：为了降低数据传输次数和计算量,针对离散非线性系统的最优控制问题,提出了一种基于单网络值迭代算法的事件驱动最优控制方案。首先,设计了一种新型事件驱动阈值,当事件驱动误差大于该阈值时事件触发;然后,仅利用一个神经网络来构建评价网,直接计算获得系统状态和控制策略,省略了典型自适应动态规划中模型网和执行网的构建,从而减少了神经网络权值的训练量,通过在评价网和控制策略之间不断地迭代,获得事件驱动近似最优控制策略;接着,基于Lyapunov稳定性理论分别证明了闭环系统的稳定性和评价网络权值的一致最终有界性;最后,将该方法应用于一个离散非线性系统上进行仿真,实验结果验证了所提出的事件驱动最优控制方案的有效性。In order to reduce the data transmission frequency and control input calculation,an event-triggered optimal control scheme for discrete-time nonlinear systems is investigated by adopting the single network value iteration algorithm.Firstly,a new event-triggered threshold is defined,which is triggered when the event-triggered error is greater than the threshold.Then,only one neural network（NN）is used to construct the critic network.The system state and control policy are obtained by calculation.The model network and action network in the typical adaptive dynamic programming architecture are omitted to reduce the NN weight training.Through the iteration between the critic NN and the control policy,the approximate optimal control policy is obtained.Based on Lyapunov stability theory,the stability of closed loop system and the uniform ultimate boundedness of the critic NN weight estimation error are demonstrated,respectively.Finally,the control scheme is applied to a discrete-time nonlinear system.The results show that the proposed scheme is effective.

关 键 词：离散非线性系统 事件驱动控制 值迭代算法 最优控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] O221[自动化与计算机技术—控制科学与工程]