基于ZGS的大规模多智能体系统的分布式优化算法

A ZGS based on distributed optimization algorithm for large scale multi-agent system

分布式优化算法已广泛用于解决大规模多智能体系统优化问题,其中异步分布式方法由于其应用于多智能体系统时的灵活性和自主性而受欢迎。在本文中,我们针对多智能体系统一致性问题提出了一种基于的Zero-Gradient-Sum(ZGS)算法的异步分布式优化算法Accelerated-ZeroGradient-Sum(AZGS),其通过提高智能体之间的信息交互模式来加速ZGS算法的收敛速度。同时改进其信息交互方式使其在实际通信过程中节省通信量。在多智能体系统网络连通的条件下,证明了所提出的算法相比于原算法更快实现渐近收敛。最后,我们通过一个数值示例验证所提出的算法的有效性。

Distributed optimization algorithms have been widely-used in solving large-scale multi-agent optimization problems,among which asynchronous distributed methods are of particular interest due to their flexibility and autonomy when applied to multi-agent systems.In this paper,we propose an asynchronous distributed optimization algorithm Accelerated-Zero-Gradient-Sum(AZGS)based on the recently developed Zero-Gradient-Sum(ZGS)algorithm,which accelerates the convergence speed of the ZGS algorithm by improving the efficiency of the agent interactions.Mean While,we improve its information interaction so that it saves actual amount of information during communication.Under mild connectivity condition of the network,asymptotic convergence of the proposed algorithm is proved.Finally,we verify the effectiveness of the proposed algorithm via a numerical example.