协同进化多代理系统

协同进化与进化稳定遗传算法相结合，并应用于多代理系统形成协同进化多代理系统。协同进化多代理系统将每个代理看作一个群体，所有代理按顺序进化，以达到最高适应度。本文采用进化稳定遗传策略，采用稳定参数控制下的突变因子，使算法具有较高的全局寻优能力和较快的收敛速度。同时采用基于适应度大小顺序的选择算子代替常用的轮盘赌选择法。避免了群体中各个染色体适应值相差不大而导致优秀个体得不到足够的后代。

Co-evolutionary Multi-agent System

Co-evolutionary multi-agent systems employ co-evolutionary combining with evolutionarily stable genetic algorithms to solve the multi-agent learning problem. Each agent of CEMAS evolves sequentially to maximize the full utility of the system. By using evolutionarily stable strategy we improve the ability of mutation processs. With the mutation operator, the diversity of the population is improved, and the full utility convergence is also accelerated. With the selection operator based on the sequence of fitness, we avoid the best can't evolve in the next generation because the little difference of the fitness of the chromosomes.