An Agent-based Co-decision Support System For Crop Production Management

Agent and Multi-Agent System theory are one of the hotspot of DAI (Distributed artificial intelligence) , whose basic idea is to decompose one complex issue into a number of singles which one agent can solve, then these agents collaborate to solve the issue. For the deficiencies of traditional agricultural information system in solving complex problems in distributed network environment, and the excellent capability as intelligent, interactive, collaborative, this paper built a crop co-decision management platform based on agent technology, for the purpose of sharing information and software teamwork and group decision support, and realized the system in java environment with the help of tool-MAGE.In this paper we discussed the conception, characteristic, architecture and other related technology of MAS, elaborated on the Agent-Oriented Software Engineering and Agent UML modeling method and FIPA MAS architecture, based on witch the architecture modeling framework of MAS for crop management was given. Based on analysis of crop management system and MAS, the function and structure of each role agent is discussed. Combining with specific features of agricultural DDS, we introduced several parameters for mental state such as trust degree, cooperation frequency and activeness degree into the collaboration of contract net. And we introduced the strategy of filtering the "acquaintances". The normal model of announcing and bidding based on extended contract net model makes it more rationalization for the task's the commission and contracting. This project not only perfects and enriches the theory of MAS cooperative mechanism, reduces communication expenses and calculation amount, improved systematic efficiency, but also offers an effective method of realizing technology for the development of practical MAS.