多智能体系统中基于演化博弈的群体状态控制

Control of states based on evolutionary games in multi-agent systems

中文关键词  博弈论  随机规则网络  复制动力学

英文关键词  game theory  random regular network  replicator dynamics

中文摘要

      在对多智能体系统的研究中, 如何通过施加最少的控制来使某种策略在群体中占优是一个未解的难题. 本 文借助演化博弈理论, 通过设置一定比例节点为指定策略作为控制手段, 分别研究了在无结构群体和随机规则网络 群体中的策略演化情况. 在随机规则网络中, 本文进一步研究了在控制手段下, 一种新策略是如何演化并成功占据 整个网络的. 结果表明在无结构的情况下, 强制策略对群体的影响受限于博弈的类型; 而在随机规则网络中, 在任 何的博弈类型下, 只要给定足够多的强制策略就可以使其突破成功. 在理论分析的基础上, 本文进行了计算机仿真 验证, 仿真结果与理论结果一致. 本文的结果揭示了如何对群体施加影响, 进而对群体中的个体状态进行控制.

英文摘要

      When the evolution of strategies in the network is studied through game theory, it is still an open question as to how to make a strategy to take over the whole network by exerting minimum control. In this paper, we investigate the evolution of strategies in unstructured networks and random regular networks by forcing a certain proportion of nodes’ strategies as a control. Furthermore, in the random regular network, we have studied how a new strategy invades and succeeds in dominating the whole network under control. The results show that the effect of the forcing strategy in the unstructured group depends strongly on the type of game; however, in the random regular network and under any game type, the intrusion can be successful as long as a sufficient number of forcing strategies are given, which is validated by our simulation results.