Distributed Genetic Algorithm: Learning by Direct Exchange of Chromosomes

摘要

Genetic algorithms is a technique widely used to evolve controllers of agents or robots in dynamic environments. In this paper we describe a modification to a single-robot-based evolution of a controller - a distributed parallel genetic algorithm where the pool of chromosomes is dispersed over a multi-robot society. Robots share their experience in solving the task by direct exchange of individually evolved successful strategies coded by chromosomes.