正交Multi-agent遗传算法及其性能分析

Orthogonal multi-agent genetic algorithm and its performance analysis

将Multi-agent系统、遗传算法与正交试验设计方法相结合,提出一种新的遗传算法——正交Multi-agent遗传算法.其主要思想是:利用正交设计的方法产生初始化种群;用正交交叉算子代替传统的算术交叉算子;利用agent间的竞争作用与每个agent所具有的知识和自学习能力进行启发式搜索,以提高进化的速度.仿真试验和性能分析表明,正交Multi-agent遗传算法不但具有很强的全局优化能力和较快的收敛速度,而且具有很强的鲁棒性.

An algorithm, orthogonal multi-agent genetic algorithm (OMAGA), is proposed, which is based on multi-agent systems, genetic algorithm and orthogonal experimental design method. The initial population of points is generated by orthogonal experimental design. An orthogonal crossover operator instead of arithmetic crossover operator is adopted and heuristic search is implemented by using the competition between agents and the knowledge and self-learning ability of each agent to accelerate the convergent speed. The simulation results and performance analysis show that OMAGA has a stronger ability of global optimization and is more robust than genetic algorithm.