Optimization Of Logistics Deliver Region Partition For Large-scale VRP

With the rapid economic development of the modern city, the development of commodity trading is fast. And the rapid logistics demonstrate the power of logistics.As an important part of the logistics system, distribution closely related to people’s lives is an important factor to reflect the level of urban logistics services. With the expansion of the city, the commodities trading have become increasingly frequent, and the urban distribution customers point scale has become larger. It is critical challenges for the logistics enterprises in the distribution of large-scale customer point to plan vehicle distribution lines, reduce distribution vehicle travel distance, reduce cost and maximize efficiency of the enterprises. Vehicle Routing Problem (VRP) abstracted from real life and closely related to it has been a hot research problem of academia in multiple disciplines. Algorithms research of small-scale vehicle routing problem with has been relatively ripening with the exact solution available in a range of50points. However, due to the scale and diversity of practical problems, the problem is complex.Based on the study of the subarea clustering algorithm, the paper presents the three-stage algorithm "devising, scheduling then improving" according to the characteristics of large-scale VRP. Firstly, the use of to the large-scale customer point is devised by the improved k-means algorithm, so the large-scale data set is divided into a number of small-scale datasets, and the influence of the partitions number on the results is tested by comparison. Secondly, schedule vehicle in each region by the parallel-saving algorithm on the basis of the partition of the first stage. Finally, use neighborhood search to improve results in order to improve the correlation between the regions.We make a large number of simulation tests in order to verify the effectiveness of the algorithm. Test examples include not only the uniform distribution density of customer counting cases, but also combined with the true distribution of data in real life, uneven density distribution was constructed customers counting cases, the number of different partitions of different sizes customer point simulation results after partition the test results than the partition results deviation of1%to3%, verifying the feasibility of the algorithm. In addition, the algorithm solved the large-scale VRP of3189customers in short time, reflecting a high solution quality and efficiency of the algorithm.Finally, the paper made prospects for regional studies of large-scale VRP prospect, combined with the lack and presented a follow-up study direction.