自适应分布式聚合博弈广义纳什均衡算法
Distributed Adaptive Generalized Nash Equilibrium Algorithm for Aggregative Games
时侠圣 1任璐 1孙长银1
作者信息
- 1. 安徽大学自主无人系统技术教育部工程研究中心 合肥 230601;安徽大学安徽省无人系统与智能技术工程研究中心 合肥 230601;安徽大学人工智能学院 合肥 230601
- 折叠
摘要
随着信息物理系统技术的发展,面向多智能体系统的分布式协同优化问题得到广泛研究.主要研究面向多智能体系统的受约束分布式聚合博弈问题,其中局部智能体成本函数受到全局聚合项约束和全局等式耦合约束.首先,面向一阶积分型多智能体系统设计一种基于估计梯度下降的纳什均衡求解算法.其中,利用多智能体系统平均一致性方法设计一种自适应估计策略,以实现全局聚合项约束分布式估计,并据此计算出梯度函数估计值.其次,利用状态反馈策略和输出反馈策略将上述算法推广至状态信息可测和状态信息不可测一般线性异构多智能体系统.最后,利用拉萨尔不变性原理证实上述算法收敛性,并提供多组案例仿真用以验证算法有效性.
Abstract
With the development of cyber-physical system technology,the distributed cooperative optimization problem for multi-agent systems has been widely studied.This study focuses on the distributed constrained aggreg-ative game for multi-agent systems,where the local cost function is subject to the global aggregative and global equality constraints.Firstly,a Nash equilibrium seeking algorithm based on estimation gradient descent is designed for the first-order integrator-based multi-agent systems.To this end,an adaptive estimation scheme is designed us-ing the average consensus method of multi-agent systems to realize the distributed estimation of global aggregative function.Based on this,the estimation gradient function is calculated.Secondly,the above algorithm is extended to the state-accessible and state-inaccessible general heterogeneous linear multi-agent systems using the state and out-put feedback control scheme,respectively.Finally,the convergence proof is provided using the LaSalle's invariance principle and several simulation examples are provided for illustrating the effectiveness of our proposed algorithms.
关键词
聚合博弈/自适应/比例积分/梯度跟踪/一般线性多智能体系统Key words
Aggregative game/adaptive/proportional-integral/gradient tracking/general linear multi-agent system引用本文复制引用
基金项目
国家自然科学基金创新研究群体科学基金(61921004)
国家自然科学基金重点项目(62236002)
国家自然科学基金重点项目(62136008)
国家自然科学基金(62303009)
出版年
2024
国家科技期刊平台
NSTL
万方数据