A DYNAMIC PLANNING METHOD OF MOBILE AGENT PATH BASED ON WINDOW STRATEGY

The mobile agent is a user program that is able to migrate continuously among sites in a network and use site services to complete task. In a social cooperative system, the site can be regarded as a proxy of the social member and all sites form a work environment for mobile agent. The sequence of sites on which the mobile agent completed tasks is called mobile agent path. The path planning has been one of the key issues in the mobile agent research. This paper discusses a mobile agent path planning model based on Markov decision process, and proposes a dynamic mobile agent path planning method based on the referral network with window strategy, of which the referral network is used to characterize the partial work environment of mobile agent, the window strategy is used to find the optimal workplace on the referral network, similar to a 'thinking for multi-step before taking one step' behavior mode in the chess game. The simulation results and an example have shown that, by utilizing the referral network with window strategy, the mobile agent is able to dynamically construct an optimal path for executing sequence of tasks.