基于集成协同PSO算法的车辆路径优化仿真  

Ensemble Collaborative PSO Algorithm for Vehicle Routing Problem Optimization and Simulation

作　　者：施彦[1] 韩力群[1] 陈秀新[1] 

机构地区：[1]北京工商大学计算机与信息工程学院,北京100048

出　　处：《计算机仿真》2012年第6期339-342,350,共5页Computer Simulation

基　　金：北京市自然基金项目:连锁零售企业精益物流供应链智能协同决策管理模式研究(9102005);北京市教委科技发展计划项目:连锁零售企业供应链风险预警与控制系统的研究(KM201210011005)

摘　　要：为提高物流配送效率,减小配送车辆的运输成本,提出采用改进的集成协同粒子群优化(PSO)算法来对路径进行优化。根据车辆路径问题的特点,采用极坐标对路径上的节点编码,并用权重表示其先后顺序,将其转化为连续PSO算法解决该问题。并且针对标准PSO算法存在的早熟问题,通过划分子种群来提高粒子的多样性,并利用集成学习,将粒子的每个维度视为个体学习者进行结合,提高搜索精度,构建了集成协同PSO算法。理论分析和实验表明,所采用的编码方式结合改进的集成协同PSO算法可以有效解决车辆路径问题。To improve logistics efficiency and reduce transportation costs, an improved ensemble collaborative particle swarm optimization （PSO） algorithm was proposed to optimize the path of distribution vehicles. According to the Vehicle Routing Problem＇s characteristics, the rout nodes were encoded in polar coordinate and the route order was represented by weights. Based on this, a continuous style PSO algorithm was applied to the VRP problem. Fur- thermore, in order to solve the prematurity problem in standard PSO, the whole swarm was divided into many sub - populations to increase diversity and particles＇ dimensions as individual learners were combined through ensemble learning to increase search accuracy. Theory analysis and experiment results show that the improved PSO algorithm combined with the above encode method is effective to solve vehicle routing problem.

关 键 词：集成学习 粒子群优化算法 车辆路径问题 极坐标 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] F252[自动化与计算机技术—控制科学与工程]