链式多种群多智能体进化算法

Chainlike multi-population multi-agent evolutionary algorithm

将多种群的进化方式和链式结构的动态邻域引入到多智能体进化算法中,提出了一种链式多种群多智能体进化算法.算法设置了多种群交互的演化结构.各种群中的智能体通过与其动态邻域智能体的竞争、合作及自学习操作来增加自身的能量;动态邻域的链式结构提高了算法的效率、降低了计算复杂度;多个种群之间的信息定期以一定的方式进行交互,增强了种群的多样性,减小了算法陷入局部最优的机率.理论分析和多个测试函数的仿真结果均表明:链式多种群多智能体进化算法在求解高维优化问题上具有很好的性能.

We propose a novel chainlike multi-population multi-agent evolutionary algorithm which combines the dy- namic neighborhood environment chainlike structure with the evolutionary framework of multi-population. This algorithm provides the evolution structure for multi-populations interaction. Agents in the population increase their own energy by competition, cooperation and self-study with its dynamic neighborhood agents. The chainlike structure improves the effi- ciency of algorithms and reduces the computational complexity. The interaction of information among various populations in a regular period of time improves the diversity of the population and decreases the possibility of sticking at local op- tima. Theoretical analysis and simulation of multiple test functions show that the new algorithm is very good for handling high-dimension optimization problems.