Modeling And Multi-objective Optimization Algorithms For Vehicle Routing Problem In Logistics Distribution

Logistics, which is regard as "the third profit source" of enterprises, has received extensive attention of people. Distribution is an important link of logistics system, accounting for more than 50% of the logistics cost. Vehicle routing problem is the core issue of logistics distribution link, it becomes hot issue in the field of logistics distribution, operational research and combinatorial optimization problem since it is proposed. Reasonable planning distribution route not only can effectively reduce logistics cost, increase the profit of the enterprises, but also can improve the quality of service, increase customer satisfaction degree. Therefore, the research about vehicle routing problem has important practical value and scientific significanceVehicle routing problem is a typical multi-objective optimization problem, which often has multiple goals that have conditions each other, but the traditional methods to solve the multi-objective optimization problem are difficult to converge to the Pareto optimal solution set, therefore the model and algorithm for vehicle routing problem are studied from the perspective of multi-objective in this paper. This paper mainly studied two kinds of vehicle routing problem: vehicle routing problem with time windows and open vehicle routing problem with fuzzy time windows. The multi-objective mathematical model is established, and different multi-objective evolutionary algorithm is designed for solving two models. This main research content of this paper is as follows:(1) The vehicle routing problem with time windows. This paper sets up a multi-objective mathematical model to minimize travel distance and minimize number, and puts forward a multi-objective hybrid differential evolution algorithm to solve this problem. Natural number coding mechanism is adopted, and the methods to produce new individual are redefined in the algorithm. The advantages and disadvantages of individuals are evaluated by introducing the concept of Pareto dominance, and non-dominated set is constructed by using the Arena's principle. In the process of evolution, differential evolution algorithm is regard as the main body of solving problem, global exploration ability and local exploitation ability of the population are effectively balanced in the solution space by introducing mechanism of double populations and variable neighborhood search strategy, and the searching efficiency of the algorithm is improved. The experimental results show that multi-objective hybrid differential evolution algorithm is effective to solve the vehicle routing problem with time windows.(2) The open vehicle routing problem with fuzzy time windows. This paper sets up a multi-objective mathematical model to minimize the total cost and maximize customer satisfaction degree by considering the vehicle capacity constraint and vehicle maximum mileage constraint, and proposes a multi-objective hybrid genetic algorithm to solve this model. Algorithm uses natural number coding mechanism, and adopts mixed method to construct the initial population and suitable genetic operators for this problem. In the process of evolution, the local search algorithm that performs simulated annealing mechanism based on meta-Lamarckian learning strategy is implemented, and duplication individuals of population are removed, to improve local exploitation ability of the algorithm. Arena's principle is adopted to construct non-dominated solution set, to reduce time complexity of algorithm.The harmonic average distance is adopted to evaluate the crowded degree of individual, to improve the distribution of non-dominated set. Experimental results show that the algorithm can effectively solving the open vehicle routing problem with fuzzy time windows, and the results also verify the algorithm's advantages in the number of Pareto solutions, the distribution of solutions and the convergence of algorithm compared with NSGA-? algorithm.