Research On Multi-objective Resource Optimization Algorithm For Shared Transportation

Nowadays,with the rapid development of urbanization,the number of urban population is increasing,the number of travel tools is also increasing,and problems such as traffic congestion and environmental pollution are becoming more and more serious,which poses new challenges to the development of urban transportation.In the era of sharing economy,the development of shared transportation provides a new idea for solving urban traffic problems.The improvement of the utilization rate of traffic resources can alleviate many problems caused by excessive traffic pressure.Optimizing the allocation of shared transportation resources can further improve the utilization rate of shared transportation resources and reduce the investment in transportation resources,which can not only save costs,but also reduce the degree of urban road traffic congestion and reduce pollution emissions.Shared transportation has many modes of transportation,and most of its resource optimal allocation problems belong to multi-objective optimization problems.Thesis mainly chooses the fixed crowd carpooling problem and the shared bicycle dynamic rebalancing problem,establishes multi-objective problem models,and studies heuristic algorithm to solving multi-objective optimization problems.For the problem of carpooling with fixed groups of people,thesis takes the people working in the same industrial park as the research object,and establishes a multi-objective carpooling problem model.The goal of this problem includes minimizing the total mileage of vehicles,the total mileage of employees,and the extra time consumed by employees.A serial code is designed to represent the carpooling scheme.A multiple leaders particle swarm optimization algorithm MPSO-VNS combined with variable neighborhood search is proposed.The leaders is selected from the optimal solution set of particle motion according to the direction distance index which proposed in thesis.Experiments show that MPSO-VNS can obtain better quality non-dominated solution sets compared with the other six algorithms NSGA-II,MOEA/D,PSO,Ma PSO,VNS and Two-Level VNS.Dynamic rebalancing of shared bicycles can solve the problem of imbalance between supply and demand of bicycles in different locations in daily life,which is more meaningful than static rebalancing.Thesis considers a multi-objective dynamic rebalancing problem that minimizes the total distance traveled by rebalancing truck and the total unmet demand.A mathematical model is established to formulate this problem.A sequence flow multi-objective evolutionary algorithm SFMEA-VNS with variable neighborhood search is proposed.A sequence flow encoding and non-dominated solution sorting method with similarity comparison are proposed.The algorithm designs and applies six local search neighborhoods.Experiments show that the non-dominated solution set of SFMEA-VNS has better quality than NSGA-II,MOEA/D,SPEA2,ILS and LSMOVRPTW.Finally,an application platform is designed and implemented for the above algorithms.Users can input their own application examples in the platform and choose the optimization algorithm to solve their own problems.The requirements of the system are analyzed,the structure of the system and the detailed design of each functional module are introduced,and the operation effect of the system is shown.