Mining Theory And System Implementation Based On The Components Of Multi-agent Running Platform

With the rapid development of computer network and communication technology, people have more and more available devices connected to Internet. In terms of the different devices and running environment, people may require some different versions of the same function to meet their individual needs. In order to fulfill personal services on demand, we put forward the concept of Component Mining. The basic idea of Component Mining is making use of several task-specific software agents, analyzing user's requests for computing, searching and retrieving candidates from online component repositories according to this request, composing and reassembling them to form programs that perform the expected computing.Component Mining is involved with many techniques, such as component, database, mobile and intelligent agents, and XML etc. In order to run Component Mining smoothly, we have to develop a good and strong environment to support Component Mining. Because Component Mining is an extremely intricate process, we define the role of agents, the interface of agents and the lifecycle of agents etc, and propose agent interaction protocols. Component Mining is distributed computing system. We introduce agents to facilitate the whole process of Component Mining. However agents need to communicate and cooperate with each other to do Component Mining. So we define the message format between agents and the way of routing a message to a target. During Component Mining, there are a series of complex interactions among mobile agents and intelligent agents. Thus some detailed interaction protocols are defined in each step of the process.Intelligent agents need Multi Agent Running Platform (MARP) to support their running. MARP provides five functional modules including agent communication, agent management, agent migration, persistence,security, agent directory. Because of many advantages in Java, we choose it to build our MARP. Agent migration and agent communication are based on Remote Method Invocation. We design agent migration protocol and agent persistence protocol. In addition, we propose a communication mechanism of transparency location and reliable message delivery.At last, we introduce our prototype of Component Mining. The prototype helps people delve into the Component Mining. We explain the structure of prototype and provide an example.