分布式反馈优化研究现状与发展

Distributed feedback optimization:Status and prospects

对于多自主体系统,分布式反馈优化研究各个自主体如何利用自身状态及局部梯度信息的实时反馈值以及自主体之间的实时信息交换使整个多自主体系统的输出趋近于某个总体优化指标的最优值点.在特定情况下,分布式反馈优化可以简化为分布式优化或集中式反馈优化,基于此介绍分布式反馈优化的研究现状与发展.利用文献调查法和归纳推理法,首先介绍分布式反馈优化的研究背景与意义以及相关领域发展现状;然后分析分布式反馈优化的基本难点,从线性系统和非线性系统两方面综述分布式反馈优化的研究现状;最后展望未来发展方向.

For multi-agent systems,the problem of distributed feedback optimization is about forcing the output of the multi-agent system to converge to the optimizer of certain global cost function by means of a distributed controller that depends on the local state of each agent,the real-time feedback value of the local gradient information,and the real-time information exchange with other agents.In some special cases,the distributed feedback optimization problem can be reduced to a distributed optimization problem or a centralized feedback optimization problem.This paper reviews the current state of the art of distributed feedback optimization and discusses its outlook.Through literature survey and inductive reasoning,this paper first introduces the research background and significance of distributed feedback optimization and the development status of related fields,then analyzes the fundamental difficulties of distributed feedback optimization,and then summarizes the research status of distributed feedback optimization from the aspects of linear systems and nonlinear systems.Finally,this paper gives the future development direction.