多智能体系统的性能优化
Performance Optimization of Multi-agent Systems
探讨多智能体系统的性能优化及相关问题. 多智能体系统性能优化问题是指给定性能评价指标, 设计分布式协议或者在某类分布式协议下优化通信拓扑的边权重或设计通信拓扑图, 使系统以最优的性能完成既定任务. 按性能指标的评价对象, 可将多智能体系统性能优化问题分为基于系统整体性能的优化和基于个体性能的优化. 文章首先针对系统整体性能优化问题, 分别介绍了多智能体系统的快速一致性问题和综合最优控制问题; 并基于线性二次型最优控制理论, 得到领航者------跟随者多智能体系统达到一致的最优拓扑是星拓扑. 其次, 对个体性能优化问题, 介绍了利用博弈论研究这一问题的相关成果; 并基于零和博弈, 得到存在两个竞争性领航者的多智能体系统最优拓扑的判别条件. 最后, 对这一领域的未来发展趋势做出了一些展望.
In this paper, we investigate some problems related to performance optimization for multi-agent systems. For a multi-agent system, performance optimization is optimizing the edge weigh of the interaction topology or designing a protocol or an interaction topology so as to minimize (or maximize) a given performance criterion. There are two kinds of performance criterions: One is to evaluate the entire system, the other is to evaluate an individual agent. Firstly, for the performance optimization problem based on the entire system, we present an overview on the fast consensus problem and the synthesized optimal control problem for multi-agent system, respectively. Then, based on Linear Quadratic Regulator theory, we obtain that the optimal topology for leader-following consensus problem of multi-agent systems is a star topology. Secondly, in the case of performance for an individual agent, we introduce some related results based on game theory. By using zero-sum game, we get the criterion on the optimal topology of a multi-agent system with two competitive leaders. Finally, some future research directions for this field are presented at the end of the paper.
多智能体系统 / 最优控制 / 有限时间一致性 / 博弈 {{custom_keyword}} /
/
〈 | 〉 |