数据驱动的保证收敛速率最优输出调节  

Data-driven Optimal Output Regulation With Assured Convergence Rate

作　　者：姜艺 范家璐 柴天佑 JIANG Yi;FAN Jia-Lu;CHAI Tian-You(State Key Laboratory of Synthetical Automation for Process Industries,Northeastern University,Shenyang 110819)

机构地区：[1]东北大学流程工业综合自动化国家重点实验室,沈阳110819

出　　处：《自动化学报》2022年第4期980-991,共12页Acta Automatica Sinica

基　　金：国家自然科学基金(61991404,61991403,61991400,61533015);中央高校基本科研专项资金(N180804001);2020年度辽宁省科技重大专项计划(2020JH1/10100008)资助。

摘　　要：针对具有外部系统扰动的线性离散时间系统的输出调节问题,提出了可保证收敛速率的数据驱动最优输出调节方法,包括状态可在线测量系统的基于状态反馈的算法,与状态不可在线测量系统的基于输出反馈的算法.首先,该问题被分解为输出调节方程求解问题与反馈控制律设计问题,基于输出调节方程的解,通过引入收敛速率参数,建立了可保证收敛速率的最优控制问题,通过求解该问题得到具有保证收敛速率的输出调节器.之后,利用强化学习的方法,设计基于值迭代的数据驱动状态反馈控制器,学习得到基于状态反馈的最优输出调节器.对于状态无法在线测量的被控对象,利用历史输入输出数据对状态进行重构,并以此为基础设计基于值迭代的数据驱动输出反馈控制器.仿真结果验证了所提方法的有效性.This paper investigates the output regulation problem for linear discrete-time systems with disturbances caused by exosystem and proposes data-driven optimal output regulation approaches with assured convergence rate,including the state feedback based algorithm for the system whose state can be measured online,and the output feedback based algorithm for the system whose state cannot be measured online.Firstly,this problem is decomposed into an output regulation equation solving problem and a feedback control law design problem.Based on the solutions of the output regulation equation,by introducing the convergence rate parameter,an optimal control problem with assured convergence rate is formulated and an assured convergence rate output regulator can be obtained by solving this problem.Then,by using the reinforcement learning approach,this paper designs a value iteration based data-driven state feedback controller which can learn the state feedback based optimal output regulator.For the systems whose states cannot be measured online,the state is reconstructed by using historical input and output data,and a data-driven output feedback controller based on value iteration is designed.Simulation results show the effectiveness of the proposed approaches.

关 键 词：保证收敛速率 最优输出调节 强化学习 值迭代 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]