摘要
对车辆的动态线路进行实时准确的优化调整,有利于运输成本与运输效率的控制。针对当前线路优化算法存在的动态性能不佳,以及大规模数据处理性能方面的缺陷,提出了关于大数据的车辆动态线路遗传优化方法。方法充分考虑了配送过程中各种因素导致的开销,以及线路与时间对成本的影响,设计惩罚因子,同时建立了车辆的动态线路调度模型。并根据最小运输成本确定遗传算法的评价函数,同时设计了自整定优化的算子交叉策略,从而改善遗传算法的收敛性与寻优性。最后将算法部署于分布式集群上,利用的在线处理能力,将一个算法任务拆分成多个并行任务分别执行,改善算法的处理效率。通过仿真,验证了所提方法能够更加准确的对车辆动态线路做出优化调整,在数据规模增加时,仍然能够保证良好的动态线路优化精确度与实时性,并且显著降低车辆的运输成本。
Real-time and accurate optimization and adjustment of the dynamic line of vehicles is conducive to the control of transportation cost and efficiency.In view of the poor dynamic performance of current line optimization algorithms and the defects of large-scale data processing performance,a genetic optimization method for vehicle dynamic line with large data is proposed.Taking full account of the cost caused by various factors in the distribution process,as well as the impact of route and time on the cost,we designed penalty factors based on the method.At the same time,the dynamic route scheduling model of vehicles was established.The evaluation function of genetic algorithm was determined according to the minimum transportation cost,and the operator crossover strategy of self-tuning optimization was designed.So the convergence and optimization of genetic algorithm can be improved.Finally,the algorithm was deployed on Spark distributed cluster.Using Spark’s online processing capability,an algorithm task was divided into several parallel tasks and executed separately to improve the efficiency of the algorithm.Through simulation experiments,it is verified that the proposed method can optimize and adjust the vehicle dynamic line more accurately.With the increase of data scale,it can still ensure good accuracy and real-time performance of dynamic route optimization,and significantly reduce the transportation cost of vehicles.
作者
许智宏
王怡峥
王利琴
董永峰
XU Zhi-hong;WANG Yi-zheng;WANG Li-qin;DONG Yong-feng(School of Artificial Intelligence,Hebei University of Technology,Tianjin 300401,China)
出处
《计算机仿真》
北大核心
2020年第6期122-125,231,共5页
Computer Simulation
关键词
动态线路优化
惩罚因子
优化遗传算法
自整定算子交叉
分布式集群
Dynamic line optimization
Punishment factor
Optimized genetic algorithm
Self-tuning operator crossing
Distributed cluster