Research And Design Of Embedded Mobile Agent Negotiaition Platform

With the development of network manufacturing, Multi-Agent technology embedded networked equipment have been successfully applied, which makes network manufacturing more intelligent. In Multi-Agent System, Agents negotiate with each other to complish complicated tasks. The task dispatch in intelligent manufacturing system can be classified to three levels. They are enterprise level, jobshop level and equipment level. Currently, equipment level Agent negotiation is the hardest part to implement.To run mobile Agent in embedded equipment, a embedded mobile Agent platform is essential. But no embedded mobile Agent platform can be directly used. We developed such a platform to test negotiation. To make our Agent widely executable, our schema is a popular ARM and Linux embedded system architecture, which is now widely applied in manufacturing equipment and smart terminal. Java is a cross platform computer language, so this paper discussed J2ME architecture and KVM technology. When porting KVM to ARM9, some related functions and the mechanism of KVM were discussed in detail, including memory allocation garbage collection and class load mechanism. Then Java program can run any platform with KVM. Based on Java and lightweight Agent software open source, JADE-LEAP platform was built. The characteristic is compatible to other common platform. Besides, embedded mobile Agent negotiation was analysised. Negotiation repeatedly leads to the huge quantity of data in network. While, the Agent with mobility can be flexible to collect bids, which reduces data transfer. This paper studied 3 tasks dispatch tactics. One of them comparatively with less data transfer was choosed as the tactic to migrate embedded mobile Agent.This paper discussed mobile Agent theory and its application, KVM porting, Agent oriented design and test. Besides, a scene designed for equipment Agent negotiation was design with performance on constructed platform. On JADE-LEAP Agent communicate with others in FIPA ACL, concede and reach agreement.