Intelligent Migration Model Of Mobile Agent

Compared to current object-oriented software component for distributed computing, Software Agent has bigger granularity and more intelligence, which becomes the research hotspot of current software technology, i.e. software technology changes from procedure-oriented to object-oriented, component-oriented and to agent-oriented. Mobile Agent is a program that can moves from one host to another host in the heterogeneous networks and can intercourse with other agent or resources. Firstly, this paper analyses concepts, characteristics and logic development relation of agent, software agent and mobile agent. This paper pays more attention to key components of intelligent model of mobile agent and migration plan of mobile agent, training and learning of mobile agent, choice of host, migration strategy of mobile agent based on machine learning and correlation analysis. Referring to MASIF and the exploitation of mobile agent platforms at present , this paper hammers at interlinking these platforms to realize true remotion independent of them, which premises for manufacturing a mobile agent platform.One of important characteristics of agent is intelligence. How to realize intelligence of agent attracted many researchers. After analyzing criterion of agent system, Agent intelligent architecture is proposed. Two approaches potential destination hosts based on accumulating knowledge is proposed. 1) Through correlation analysis between programming time and influential performance parameter, this performance parameter is considered as the weight mapping to the expected time; 2) Considered the influence of each performance parameter, the input number of ANN is reduced by declining the performance parameter that have little effect on the programming time. We solved the agent execution host choice between hosts itself and many uncertain influencing factors, moreover agent study accumulation knowledge method is given. In our simulation experiment in Aglet, stochastic selecting method takes more time than ANN method to prove our suppose correct. By learning Agent can obtain a more superior result and get the basis for reasonable choice host.This paper is organized as follow. The first chapter analyzes the existing problem of Agent and presents the meaning and content, which this paper confined to; The general structure, mobile agent platform and key technologies and researches of specifications are given in the second chapter; The paper attaches more to intelligence model of mobile agent, migration plan, training and learning of mobile agent, migration strategy based on correlation, migration strategy based on correlation and neural network in the third chapter4; In the fourth chapter we give simulation experiment based on above migration strategy and analyze the experiment result; The last chapter sums up the researches of this paper and the unsettled problems.