Research On Communication Mechanism Of Mobile Agent

Mobile agent technology, which combines Artificial Intelligence (AI) with distributed computing, has been a hotspot in research of distributed computing and AI. Essentially, Mobile agent is an object that looks for cooperators in the Internet on behalf of a user, and cooperates with other agents in order to implement the task of the user. Therefore, cooperation is the essential demand of mobile agent. Cooperation is a very challenging issue of mobile agent key technology. At present, the cooperation of mobile agents mainly focuses on the research of communication mechanism. There are two kinds of representative working: one is KQML research based on knowledge exchanging, the other is communication mechanism research on message transfer. The work in this thesis belongs to the latter. Communication mechanism of mobile agent is designed mostly from the perspective of supporting moble computing. In traditional distributed computing model, the position of object is fixed and not changed in the whole life circle since it is created. So the message sender is easy to get the location of a target mobile agent for communication. But for a mobile agent, its position maybe changed frequently, the following problems need to be considered under mobile agent computing mode: Firstly, how to give a position independence naming method for mobile agent in distributed environment. Secondly, how to implement transparent locating of a mobile agent. Thirdly, how to handle messages that are send to mobile agent when it is being in the procedure of migration. Finally, how to resolve the problem of communicating invalidation. Many researchers have put forward some kind of solutions of agent location tracking, message storage and transfer for problems above, such as Broadcasting method, Link Tracing method, Home Proxy method Central Registration method, and etc. Although these methods have their advantages respectively, there are some disadvantages in tracing efficiency, expansibility and communication reliability. In order to resolve the problems above, this thesis proposes a high-performance, reliable mobile agent communication model which combines name resolution mechanism with message transfer mechanism. We divide he mobile agent environment to domains. Being constructed by Domain Name Resolver and Communicator, the model is called D-C communication model. To achieve the model, we propose a naming method - Universal Agent Identity (UAID) to name all the agents in the system. Based on this, we implement the transparent locating for mobile agents. We use message storage and transfer mechanism to handle delivering messages. Additionally, the presentation of timeout and priority percept gives a good way to resolve communicating invalidation. Finally, we analyze the performance of D-C communication model, and compare it with other related model. Generally, this is a flexible-structured, high-performance and high-reliability model for mobile agent communication. These advantages make a good foundation for subsequence research.