The Comparative Of Different VRP Algorithms In Logistics

Logistics, as the third profit source of modern enterprise, plays a more and more important role in the economic activities of enterprise, in which vehicle routine problem (VRP) is the key of vehicle scheduling during logistics distribution process. Rational vehicle dispatching method will not only enhance customers’ satisfaction, but also increase the operation rate of distribution vehicles and warehouses. Most important of all, it will decrease the company’s economic cost during logistics distribution process and expand profit margin of enterprise, in order to strengthen the competitiveness of enterprise.Many scholars have made intensive studies of vehicle routing problem (VRP). They got best practices about vehicle routing problems through different algorithms. On the basis of existing research, this paper makes detailed introduction and comparison upon common vehicle dispatching methods.Firstly, the paper introduces the basic conception of vehicle optimum scheduling, classifies vehicle optimum scheduling according to different restricted conditions and expounds to divide vehicle scheduling algorithm into accurate algorithm and heuristic algorithm.Secondly, the paper explains the mathematical model, algorithm thought of Table Dispatching Method and Graphic Dispatching Method in accurate algorithm of vehicle scheduling..However, these accurate algorithms can be only used in vehicle routing problems with small scales. With the expansion of problem scale, the algorithm complexity increases exponentially. That’s why accurate algorithm is not applicable for real scheduling work of enterprise.Then,the paper introduces the algorithm thought and algorithm flow of Ant Colony Optimization (ACO), Genetic Algorithm (GA), Simulated Annealing (SA) and Particle Swarm Optimization (PSO) in heuristic algorithm. Ant Colony Optimization is a method to simulate the biological behavior of ants’ routing to get optimal solution of problems. The algorithm owns positive feedback mechanism and comparatively strong robustness. Genetic Algorithm refers the model "survival of the fittest of chromosome" in the process of biological genetic. The algorithm owns good global searching capacity. Simulated Annealing takes example by internal-energy balance theory that solid has in the process of annealing. Particle Swarm Optimization simulates the situation that during the process of foraging, some birds learn from other excellent birds positively to reach the result of acting in concert. The algorithm makes the best information of individuals and groups among swarms reflected effectively.Thirdly,the paper uses Matlab to execute programs for every algorithm and adopts each algorithm to solve a virtual logistics distribution problem.Finally,according to a postal distribution and demand case in China Post Group of city S,this paper uses Matlab to execute this case by every algorithm,then compares the advantages and disadvantages of heuristic algorithm comprehensively according to the conclusions of program operation. The papers puts forward that there are still come problems in each algorithm, further consideration is required for relation between algorithm parameter and algorithm result, perfecting parameter value and strengthening algorithm stability. Meanwhile, the paper also puts forward that each heuristic algorithm may interact with other algorithms combining with their own characteristics, to make up for their own lack and improve algorithm efficiency, which can solve vehicle routing problems better.The paper is characterized to have relatively comprehensive statement upon common algorithms in vehicle scheduling, and use Matlab to present the realization situation that each algorithm has in detailed cases. Different from qualitative comparison for every algorithm in previous articles, the paper adopts a logistics distribution problem in China Post Group of city S to carry out quantitative comparison for every algorithm. Finally, the paper summarizes that Ant Colony Optimization has simple thought and strong stability, but costs most time in operation, and is also sensitive about parameter; Genetic Algorithm works fast, but is easily involved in local optimum, and the stability of optimal solution is not high while the variance is large; Simulated Annealing has strong global searching capacity, but easily converged to inferior solution with low stability of optimal solution.