讨论了MAS环境下实现Agent交互协作的关键性问题。在分析Agent协同设计目标的基础上，将遗传算法应用于子任务的调度，提高了Agent协同工作的效率。应用Agent的通信语言建立了一个基于知识查询与操纵语言的Agent通信模型，通过引入Agent位置追踪和Agent消息缓存机制解决了Agent移动过程中遇到的位置追踪和通信失效问题。关 键 词 多主体系统; 移动代理; 知识查询与操纵语言; 通信模型; 协作Abstract In MAS environment, some key techniques are needed to realize interaction and cooperation of mobile Agents. In this paper, genetic algorithm is introduced to assign Agent to a suitable node, improved the coordination efficiency of Agents. An Agent communication language—knowledge query and manipulation language is analyzed and an Agent communication model is established based on knowledge query and manipulation language. Also, an effective way to solve Agent location tracing and invalidated communicating during Agent roaming among network nodes is provided.Key words multi agent system; mobile agent; knowledge query and manipulation language; communication model; cooperation