并行多种群自适应遗传算法在COW集群上的实现

为改善标准遗传算法的求解效率,提出一种基于6模糊控制器的并行多种群自适应遗传算法。利用MPI(Messagepassinginterface)技术建立了一个COW(Clusterofworkstation)集群,将算法在该硬件平台上进行了实现。3机COW集群的仿真实验结果在演示算法设计可行性的同时,表明该算法的求解效率明显优于用于对照的单种群算法,具有在解决组合优化问题上广泛应用的可能。本文还对影响并行算法的参数进行了探讨。

Parallel Multi-Deme Adaptive Genetic Algorithm on Cluster of Workstation

To improve the solving efficiency of standard genetic algorithms,a novel parallel multi-deme adaptive genetic algorithm is proposed based on six fuzzy logic controllers(6FLC-MDPFGA).A PC cluster of workstation(COW) using the message passing interface(MPI) technology is built.Furthermore,the 6FLC-MDPFGA is realized on the hardware platform.When the possibility of the algorithm design is illustrated,results from initial experiments on the 3 PCs COW platform indicate that the algorithm efficiency is improved.Meanwhile,experiments show that the new multi-deme algorithm can provide more stable results.The algorithm is run on the 3 PCs COW platform,and easily implemented on a large-scale PCs COW platform using MPI.The 6FLC-MDPFGA can be applied to a wide range of combinatorial optimization.Finally,how to select parameters is discussed.