MAS系统的问题求解能力分析

本文用状态空间搜索模型分析了多Agent系统(MAS)的问题求解能力，认为MAS系统中Agent之间知识的组合应用和对问题搜索方向的交互和决策是影响MAS系统问题求解能力的主要原因，在状态空间搜索模型下可以将Agent间知识的组合应用表达为不同Agent的搜索路径的组合，而Agent对搜索方向的判断是基于启发式信息做出的，从而为形式化分析MAS系统的性能建立了通用的模型．本文以A*算法为例探讨了可采纳算法下多Agent合作求解效果与Agent的知识和启发信息之间的关系，指出只有在一定条件下MAS系统才会获得更好的解题能力．本文还对非可采纳算法下MAS系统性能分析方法提出了初步看法．

Analysis of Problem Solving Capability of MAS

In this paper, a method for analyzing problem solving capability of Multi-Agent System (MAS) is proposed, this method is based on state space searching model. We point out that synthesis of knowledge and interaction of searching direction among agents are main reasons that could influent problem solving capability of MAS. In state space searching model, synthesis of knowledge could be represented as combination of path and interaction of searching direction could be represented as interaction of heuristic information among agents. By this means a general model for formalized analysis of capability of MAS could be established. In this paper we discussed how the knowledge and heuristic information of agents influence the problem solving capability of MAS based on A* algorithm, and find that only in specific conditions could MAS perform better. In this paper we also proposed some opinions about how to analyze problem solving capability of MAS in which agent adopt unacceptable searching algorithms.