中文摘要：

针对多智能体系统在动态演化过程中容易出现的＂局部聚集＂现象,融合复杂网络中的拓扑结构优化理论与多智能体系统协调蜂拥控制研究,提出了一种基于邻域交互结构优化的多智能体快速蜂拥控制算法.该算法首先从宏观上分析多智能体的局部聚集现象,利用社团划分算法将局部相对密集的多个智能体聚类成一个社团,整个多智能体系统可以划分成多个相对稀疏的社团,并为每个社团选择度最大的个体作为信息智能体,该个体可以获知虚拟领导者信息;随后从多智能体系统中不同社团相邻个体间的局部交互结构入手,取消社团间相邻个体的交互作用,设计仅依赖于社团内部邻居个体交互作用的蜂拥控制律;理论分析表明,只要每个社团存在一个信息智能体,在虚拟领导者的引导作用下,整个多智能体系统就可以实现收敛的蜂拥控制行为;仿真实验也证实了对多智能体系统进行邻域交互结构优化可以有效提高整个系统的收敛速度.

英文摘要：

A fast flocking algorithm for multi-agent systems is presented to improve the speed of consensus of multiagent systems based on local interactive topology optimization by considering the phenomenon of ＂local cohesion＂ in the dynamic process of flocking control, and combining the theory of topology optimization with the research of multi-agent flocking control. Firstly, the phenomenon of ＂local cohesion＂ is analyzed macroscopically, and fast Newman algorithm is used to form multiple communities for multi-agent system where agents connect familiarly in the same community and connect sparsely in the different community. Then the agent which possesses the maximum degree is defined as the informed agent in every community which can obtain information from the virtual leader. Furthermore, to cut off the joint among agents in different communities, the flocking control law is proposed to only consider local interactions between neighbor agents in the same community. Theoretic analysis shows that the multi-agent system can achieve the goal of flocking control when every community has at least one informed agent. And the simulation results show that the speed of flocking control can be improved by optimizing the local interactive topology.