自动化计算总的越来越趋向于网络化、智能化、代理化、人性化发展,多Agent系统正是在这种环境之下发展起来的一种先进的计算模式.通常情况下,多Agent系统中的Agent在逻辑上或者物理上是分离的,Agent间交互与协作的智能行为及其有效性是多Agent系统研究的重点.对多Agent协作机制的研究主要是在动态环境下,多个独立自主的Agent通过相互的交互与协作等智能行为去完成对任务的求解.在多Agent系统中,传统的基于管理Agent协作机制属于集中式协调方式,采用这种协作模型的多Agent系统导致其管理Agent成为整个系统的瓶颈,从而影响系统完成任务的效率与稳定性,也增加对系统额外的维护费用.文章通过在传统基于管理Agent协作机制中加入了“熟人”、状态值元素,平衡了系统Agent的负载,降低了对管理Agent 的维护成本,整体上提高了系统解决问题的效率.%Themajor development tend of automation computing is more and more networked, intelligent, agent, humane; and working with this situation and development requirements, the multi-Agent System appeared and is developed as a new computing model. Usually, the agents in multi-agents system are logically or physically separated, so the intelligent behavior and effectiveness of agents' interaction and collaboration is one of the hot spots in research area The study of multi-agents collaborative mechanism is to design a model in a dynamic environment and arrange multiple independent agents to solve problems through their interaction and collaboration. For traditional management agent based multi-agents system, its collaborative mechanism is a centralized model which means all collaborations and interactions are going through the management agent, and the management agent will be bottleneck of whole system, it also affects the system performance, effectiveness and stability; furthermore it increases the whole system maintenance cost. Here it bases on the study of the traditional management agent based multi-agents systems, by adding the "acquaintance" and "status value" elements to the traditional model, to illustrate a new improved management agent based multi-agents system design. Through this new model, it can reduce the whole system maintenance cost and overally enhance the efficiency of whole system in solving problems.
展开▼
