Research And Application Of Intelligent Component Coordination Based On MAS

The technology of Multi-Agent is a very important and active field of computer science recently. Academe and industry have all payed much attention to the field of Multi-Agent. But the technology has not been mature,and the field of application is not wide.One of the reason is that the communication among the Agents of MAS is very complex ,coordination among the Agents is difficult.So it is very hard to control it.With the increasing demand of intelligentized software ,the require of improving the MAS is imperious. The paper is on the background of 863 project:flatform of WEB intelligent application system,and is a important component of the project .The project is based on the technology of middleware.The intelligent assemblies of the project are followings:coordination assembly,KIF inferengine assembly,relating rule arithmetic assembly,interengin based on rule,bayisan inferenginee assembly,DS inferenginee assembly and etc.This paper completed coordination assembly,the function of the coordination assembly is to coordinate all kinds of inferengine in order to coordinating infer. And synthesize results of all kinds of inferengine.The results are more reasonable.The paper resolved coordination of intelligent assemblies utilizing task shared and result shared of Multi-Agent theory. The project extracts and encapsules speciality of intelligent field,and develops international intelligent IDE that is intelligent,integrated and network.The plattform composes of middlewares,soft-component,develop tools and FC. Brief introduction of Agent and MAS.So far,the history of computer science has five important and related developing trend:general fit,contact,intelligent ,Agent and humanism.All of above lead the appearance of MAS field.Agent is a computer system.It can delegate user or its ower.MAS is a system composed of Agents.Commonly,these Agents exchange information via basic equipment of network.It is difficult to construct a large and uniform intelligent system with multi-form of knowledge and multi-mechanism of resolving question.But It is effective way that Agents coordinate to resolve question without fail.The specialty of MAS is:coordination,independence,distribution and isomerous. The architecture of MAS is :centralization,distribution and composite. If you discuss how to coordinate among Agent,you must consider how to design Agents in order to coordinate effectively."Coordinate"is used to describe interactional system that must interact in order to accomplish intending tasks.The paper introduce tow kinds of Agent coordination way that is task shared and result shared.During the Agents coordination ,it is hard to avoid conflict.Under the instance, it is required to conflict resolve in order to coordinate each Agent. According to spatially coupled and temporary coupled, the model made off four classes: direct,blackboard-based,meeting-oriented and linda-like. On the theory foundation of Agent coordination in MAS ,considering the specialty of the project we apply Agent coordination theory in MAS to intelligent assembly coordination.The implement in the project is that result shared is to handle every infer result of inferengine, conflict resolve,give user a reference result after synthesis.Whether the coordination result is grouped or not,there are two arithmetic.If the result need group,conflict resolve arithmetic is utilized.If the coordination result need not group,result synthesis arithmetic is utilized.Task shared is that the middle results of every inferengine are as subresult to be shared,in order to other inferengine can use them,and use the middle result to infer.So we need a global space to store the middle result.According to the different storage form of middle result,there are two coordination arithmetic.If applying distributed blackboard to store middle result,it is distributed lackboard-based coordination arithmetic.If applying tuple space to store middle result,it is linda-like arithmetic.