| 引用本文: | 薛文艳,黄捷.基于滚动分布式鲁棒微分博弈的多智能体编队避碰控制[J].控制理论与应用,2026,43(4):805~813.[点击复制] |
| XUE Wen-yan,HUANG Jie.Multi-agent formation collision avoidance control with receding distributed robust differential game[J].Control Theory & Applications,2026,43(4):805~813.[点击复制] |
|
| 基于滚动分布式鲁棒微分博弈的多智能体编队避碰控制 |
| Multi-agent formation collision avoidance control with receding distributed robust differential game |
| 摘要点击 203 全文点击 23 投稿时间：2024-03-22 修订日期：2025-06-11 |
| 查看全文 查看/发表评论 下载PDF阅读器 HTML |
| DOI编号 10.7641/CTA.2024.40167 |
| 2026,43(4):805-813 |
| 中文关键词 多智能体 微分博弈 近似纳什均衡 反馈鲁棒纳什策略 稳定性 |
| 英文关键词 multi-agent differential game approximate Nash equilibrium feedback robust Nash strategy stability |
| 基金项目 国家自然科学基金项目(92367109), 航空科学基金项目(20230001144001), 科研启动费项目(360302022401)资助. |
|
| 中文摘要 |
| 针对未知障碍物环境下受有限通信能力以及干扰约束的多智能体编队控制存在鲁棒性差、安全性不佳的
问题, 本文提出了一种新颖的滚动分布式鲁棒微分博弈方法. 首先, 与现有忽略外部干扰影响的流行分布式微分博
弈方法比较, 将外部干扰看作虚拟的恶意参与者, 每个智能体基于局部交互信息优化最坏情况下的个体性能指标,
以实现对外部干扰的鲁棒性. 其次, 基于分布式终端状态估计器构造了一种近似开环鲁棒纳什均衡解, 以摆脱传统
方法依赖全局状态信息求解的困境, 并从数学角度分析了近似纳什均衡的收敛性. 为保证编队的安全性, 在推导的
近似纳什均衡解中引入避碰惩罚项, 并证明了系统的稳定性. 最后, 为了提升系统的安全性能, 基于滚动优化控制
合成近似反馈鲁棒纳什策略, 通过对未知状态信息的预测, 减少了智能体在运动过程中死锁现象的发生. 仿真结果
验证了所提方法的有效性. |
| 英文摘要 |
| A novel receding distributed robust differential game method is proposed for the formation control problem
of multi-agent systems under limited communication capabilities and disturbance constraints in unknown obstacle environments. Firstly, compared with existing popular distributed differential game methods that ignore external disturbance
effects, external disturbance is regarded as virtual malicious participants, and each agent optimizes the worst-case individual performance indicators based on local interaction information to achieve robustness against external disturbance.
Secondly, an approximate open-loop robust Nash equilibrium solution is constructed based on a distributed terminal state
estimator to overcome the dilemma of relying on global state information in traditional methods, and theoretically ensure
that the convergence of the approximate Nash equilibrium. Then, to ensure the safety of the formation, a collision avoidance
penalty term is introduced into the derived approximate Nash equilibrium, and the stability of the system is proved. Finally,
based on the receding horizon control, an approximate feedback robust Nash strategy is further synthesized to reduce the
occurrence of deadlock by predicting unknown state information. The simulation results verify the effectiveness of the
proposed method. |
|
|
|
|
|