Research Of Stochastic Vehicle Routing Problem Based On Hybrid Quantum-inspired Evolutionary Algorithm

Vehicle routing problem (VRP) is an important point of logistics and distribution ends, and the rapid development of modern logistics industry provides a wide application space for its research. With the rapid growth of China's economy, increasing intensification of market competition, and accelerating industrial upgrading, there are more and more enterprises regard improving logistics efficiency and reducing logistics costs as an important way to exalt their core competitiveness. Recently, most studies about VRP are based on the case of deterministic information. But in the ever-changing market, some distribution-related information often presents some statistical law, which is the point of stochastic vehicle routing problem (SVRP). Therefore, the research about SVRP has important academic significance and practical value.To study vehicle routing problem with stochastic demand (VRPSD) , a model with soft time window and the multi-objective problem based on fuzzy due-time were established respectively in this paper, and the hybrid quantum evolutionary algorithms were designed to solve the above models. Overall, the main work and innovation of this paper is as follows:(1) A mathematical model about vehicle routing problem with stochastic demand and time window (VRPSDTW) was established in this paper. In order to solve the above problem, a quantum-inspired evolutionary algorithm based on immune operator (IQEA) was proposed. Characteristics for VRP, proposed a sort of quantum integer-based encoding, simplified the calculation by referencing a"virtual customer". The probabilistic optimal solution selecting and the dynamic rotation angle improved the algorithm's search ability. In the iteration, the good quantum fragments were preserved by immune operating, and local optimum and regress was avoided. The results of simulation on multiple instances show that IQEA can get 90% of the optimal solution, and the greater vaccination probability, the more accurate results. Besides, IQEA qualified faster convergence ability.(2) Based on VRPSDTW, a fuzzy due-time based multi-objective problem model was raised. To solve the problem, a stage optimization method based on QEA and PSO was proposed. In the initial stages, using the search space characteristics of QEA to generate a non-dominated solutions set with considerable size, which is can be convert into continuous encoding by a transforming program. Then, obtain the final Pareto optimal solution by PSO. To maintain diversity and uniform distribution of Pareto solution set, an adaptive grid algorithm was proposed. In the process of evolution, the grid density can be adaptive adjusted. The results of experiments show that the proposed method can not only get a sufficient number of Pareto solutions, and comparison with NSGA2 results indicate that the method has absolute advantages in distribution, diversity, and convergence of Pareto solution set.(3)On the basis of the above work, a stochastic vehicle scheduling verification platform was designed, which tested the models and algorithms proposed in this paper. Combined with GIS, the results can be shown up through Web page.