基于零和博弈的多智能体网络鲁棒包容控制
作者:
作者单位:

北京信息科技大学

作者简介:

通讯作者:

中图分类号:

TP273

基金项目:

北京信息科技大学学科群建设


Robust containment control of multi-agent networks based on zero-sum game
Author:
Affiliation:

Beijing Information Science and Technology University

Fund Project:

Supported by the Subject Group Construction Project of Beijing Information Science and Technology University

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对受扰非线性多智能体网络, 本文研究分布式鲁棒包容控制方法. 采用微分博弈理论将有界$\mathcal {L}_{2}$增益包容控制问题描述成多玩家零和博弈问题. 对于每个跟随者,至少存在一个领航者与其存在有向路径通信时, 基于局部邻居信息定义每个跟随者的性能指标, 从而得出包容误差$\mathcal {L}_{2}$有界且零和博弈解存在的结论. 在系统动态完全未知的情况下, 采用积分强化学习算法和执行—评价—干扰网络, 在线得到近似最优策略. 仿真结果表明了本文所提方案的有效性和正确性.

    Abstract:

    The distributed robust containment control methods are investigated in the paper for disturbed nonlinear multi-agent networks. Applying differential game theory, the bounded $\mathcal {L}_{2}$ gain containment control problem is described as multi-player zero-sum game one. Suppose there exists at least one leader that has a directed path from it to each follower, its performance index is defined based on the information of local neighbors. Furthermore, it is proved that the containment errors are $\mathcal {L}_{2}$ bounded and there exists Nash equilibrium solution. With the completely unknown system dynamics, the integral reinforcement learning method and critic-actor-disturbance neural networks are used to solve the approximate optimal strategy online. Simulation results verify the effectiveness and validity of the proposed scheme.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2019-09-25
  • 最后修改日期:2020-07-29
  • 录用日期:2020-05-12
  • 在线发布日期:
  • 出版日期: