摘要
研究一阶主从多智能体网络的汇聚问题,基于多智能体网络群体的社会制度,提出3种汇聚策略,即民主、独裁和混合策略。假设在策略中,只有领导者知道全局任务,即汇聚到一个已知的目标点,跟随者只需要与领导者群体保持联系,从而能够到达目标点。研究发现,主从网络多智体个体在属性上可以分为目标智能体,领导者智能体,非孤立跟随者从智能体和孤立跟随者从智能体。研究结果表明民主策略容易实现所有智能体到达目标点,但移动速度较慢;在独裁策略作用下,智能体网络可以很快实现汇聚,但个别跟随者从智能体由于初始条件下感应不到环境中的领导者智能体而永远停留在原来的位置;引进混合策略后,即可以保证所有的智能体到达目标点而且又可以保持较快的到达速度。通过对模型和拓扑结构的分析,得出3种策略各自的特点,并以4个领导者和5个跟随者组成的主从智能体网络为例进行仿真。
In order to touch upon the rendezvous problems of leader-follower multi-agent systems, three different kinds of strategies, namely the democracy strategy, the autarchy strategy as well as the mixed strategy were suggested based on the social system of the swarm of the multi-agents system, whereas the last strategy was composed of the democracy one and the autarchy one. It was assumed that only the leader agents were capable of knowing the details of the global task, which was to come to the same targeted object, for the follower agents, if they were kept con- nected with the corresponding leader ones, they could also realize the rendezvous task in indirect ways. It was found that among this arena, agents were finally divided into four different catego ries, that they were the agent where they represented the targeted object, the leader agents where they played the leading roles and the following agents where they had some connections with the leader ones and the follower agents where they were failed in connection with the leader ones, i. e. the isolated following agents. Results demonstrated that, the democracy strategy could fulfill the rendezvous task in an easiest way in expenses with a long convergent time and the autocracy democracy could come to the targeted object quickly with the cost of the risk of failing in completing the rendezvous task of the isolated follower agents,, while the mixed strategy cared about both of the convergent speed and the utmost task, would take autocracy strategy when the connections were available for the follower agents to the leader ones, and the democracy rules had to be used when the connection to the leader agents was not possible. Both the models and the to- pological structure were provided to elucidate the specifications of the three strategies and they were explained even further with simulation work under the leader follower agent network which consisted of 4 leader agents and 5 follower agents.
出处
《复杂系统与复杂性科学》
EI
CSCD
北大核心
2012年第1期88-94,共7页
Complex Systems and Complexity Science
基金
国家自然科学基金(60850004)
河南省教育厅科技创新人才支持计划项目(2009HASTIT021)
关键词
主从多智能体网络
汇聚问题
民主策略
独裁策略
混合策略
leader follower multi-agents network
rendezvous problems
the democracy strategy
the autarchy strategy
the mixed strategy
