基于局部优化的多智能体网络牵制蜂拥算法

Local optimization-based flocking control algorithm for multi-agent networks

针对现有的多智能体网络牵制蜂拥算法中网络在演化过程中可能出现'分裂'的问题,以代数连通度作为衡量网络连通的指标,提出一种基于局部优化的牵制蜂拥控制算法,以保持多智能体网络在演化过程中的全局连通性,从而保证所有智能体跟随虚拟领导者运动.该算法将局部优化策略引入牵制蜂拥控制器中,考虑网络中仅有部分智能体(称为信息智能体)具有虚拟领导者的信息,使信息智能体不断朝着其在当前时刻和下一时刻所有邻居的并集所组成网络的代数连通度最大的方向运动,以通过优化局部连通性达到提高整个网络连通性的目的.通过理论分析和仿真实验并与现有牵制蜂拥控制算法相比,进一步证明了算法可有效保持多智能体网络在演化过程中的全局连通性,且所有智能体均聚集在虚拟领导者周围更小的范围内,更能体现信息智能体在牵制蜂拥控制网络中的相对重要性.

In existing flocking control algorithm of multi-agent networks via pinning,networks may appear＂splitting＂in process of evolution.Therefore,a flocking control algorithm of multi-agent network via local optimization and pinning was proposed.This algorithm joins the local optimization algorithm in flocking control algorithm of multi-agent network via pinning,considering that there were only part of the agents（called information agents）have the information of the virtual leader,and they continuously move toward the direction that has the largest algebraic connectivity of network consisting of all the neighborhood of them in this moment and next moment.Thus,the connectivity of the whole network through local optimization was improved.Finally,better performance was obtained by the comparison with the previous methods.Theory analysis and numerical experiment show that the proposed algorithm can effectively keep global connectivity of multi-agent network during evolution.All the agents gather around the virtual leader in smaller area which can reflect relative importance of the information agent in multi-agent network.