基于预测机制的MAS协商通信模型  

Prediction Rules Based Communication Model of MAS

作　　者：王晓伶[1] 慕德俊[1] 袁源[1] 刘哲元[1] 

机构地区：[1]西北工业大学控制与网络研究所,西安710072

出　　处：《计算机科学》2008年第12期65-68,共4页Computer Science

基　　金：航空科学基金(01F53031);教育部博士点基金(20020699026)资助

摘　　要：网络代价直接影响MAS通信性能,特别是在服务器过载情况下影响更为明显。为减少网络通信代价、减轻服务器负载,引入预测机制,提出一种根据网络代价、服务器负载的预测结果,动态选择agent远程通信或MA迁移方式的协商通信模型MRCM,模型中引入黑板结构,实现预测结果在MAS系统中的共享,增强预测成功率、准确率。实验表明该方法对网络代价、服务器负载预测近似准确,能有效地降低通信开销。Network cost directly affects the communication performance of MAS, especially when the server is over loading. In order to reduce the communication cost and the server load, the prediction model was introduced. It selects dynamically the remote communication of agent or consulting communication model MRCM based on the transferring type of MA, according to the prediction of network cost and server load. The blackboard model was proposed to realize the sharing of the prediction results in the MAS system,and to improve the success rate and the accuracy of the prediction. The simulation shows that the method predicts the network cost and server load approximately accurately and effi- ciently reduces the communication cost.

关 键 词：多智能体系统 移动智能体 网络拥塞 网络代价 预测模型 黑板模型 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] TU243[自动化与计算机技术—计算机科学与技术]