Distributed genetic algorithm with multiple populations using Multi-agent

Abstract

This paper designs a distributed genetic algorithm in order to reduce the execution time and obtain more near optimal using master-slave multi-agent model for the TSP. Distributed genetic algorithms with multiple populations are difficult to configure because they are controlled by many parameters that affect their efficiency and accuracy. Among other things, one must decide the number and the size of the populations (demes), the rate of migration, the frequency of migrations, and the destination of the migrants. In this paper, I develop two dynamic migration window methods, increasing dynamic window and random dynamic window, that control the size and the frequency of migrations. In addition to this, I design new genetic migration policy that selects the destination of the migrants among the slave agents. © 2002 Springer Berlin Heidelberg.