Research On Multi-objective Swarm Intelligence Algorithms For Door-to-door Railway Freight Transportation Routing Design

With the implementation of the railway's freight organization reform and the unswerving advancement of the railway's corporation system reform,it promotes the transportation management model of the national railway enterprise from production to operation,which makes the railway freight fully market-oriented.It promotes the railway transportation enterprises to improve the quality and efficiency of the transportation to promote supply-side structural reform of the railway transportation in the down-to-earth manner.It will comprehensively improve the supply quality and efficiency of the railway transportation and promote the complementary advantages and integration of various modes of transportation,which will further reduce the logistics cost of the whole society.The railway transportation enterprises expand their freight transportation from station-to-station transportation to door-to-door transportation,which makes the routing design more complicated.In the new complex practical application environment,the use of swarm intelligence algorithms that has been widely used in various practical application,has theoretical significance and practical application value.The swarm intelligence algorithms can obtain optimization results in a reasonable time,which provides routing design decision support for the railway transportation enterprises to reduce the transportation cost and the transportation time of the system.In this dissertation,the swarm intelligence algorithms are reviewed,and its adaptability in the routing design of the door-to-door railway freight transportation is analyzed.Based on the analysis of the whole process of the door-to-door railway freight transportation,and taking the door-to-door railway traffic of wagon loading goods as the specific research object,the railway freight transportation routing design(RFTRD)system optimization model is constructed.Two improved discrete multi-objective swarm intelligence algorithms and two improved continuous multi-objective swarm intelligence algorithms are proposed to solve the RFTRD model,respectively.Considering resources can be expanded,and assuming the location of the originating stations and the destination stations are unknown or to be determined,the resource-expandable railway freight transportation routing design(RERFTRD)system optimization model is constructed.The RERFTRD model is also solved by the two proposed continuous multi-objective swarm intelligence algorithms.The performance of the four proposed algorithms is compared and analyzed quantitatively and qualitatively.The optimization results obtained by the four proposed algorithms are input into the Simio Simulation Software for simulation analysis.This dissertation can be divided into the following sections.In chapter 2,swarm intelligence algorithms and its application and improvement are reviewed.From the perspective of single-objective optimization and multi-objective optimization,respectively,the adaptability of the swarm intelligence algorithms to the door-to-door railway freight transportation routing design problem is analyzed in boths relationship between optimization algorithm and practical problem and the practical verification of the swarm intelligence algorithms in actual application.The characteristics,techniques,and performance metrics of multi-objective optimization are introduced.The improved two-set coverage(ISC)is introduced.Based on the fast non-dominated sorting approach,the population entropy for multi-objective optimization is proposed to measure the diversity of population during the optimization process of multi-objective optimization algorithms.In chapter 3,the process of the door-to-door railway freight transportation is analyzed in detail from the perspective of complex system and multi-objective modeling.According to the basic requirement of the activity-based costing,the process of the door-to-door railway freight transportation is divided into five operations:sending,operating,transit,arrival and service at both ends.According to the literature and actual requirement,the key impact factors for the door-to-door railway freight transportation routing design are set to the transportation cost and the transportation time.According to the basic requirement of the activity-based costing,the workload indices and their workload calculation formula for each operation are determined,and the time calculation formula for each operation is also given.From the perspective of railway transportation enterprises,and taking the door-to-door railway traffic of wagon loading goods as the specific research object,the railway freight transportation routing design(RFTRD)system optimization model is constructed,which assigns the originating station and the destination station for the goods of multiple consignors,and minimizes the transportation cost and the transportation time of the system.In chapter 4,base on three random multi-neighborhood structures(random block-insertion,random block-swap and 2-opt)for the sequencing and three random multi-neighborhood structures(random replacement,random generation and random critical replacement or generation)for the assignment,two improved discrete multi-objective swarm intelligence algorithms,the random multi-neighborhood based multi-objective shuffled frog-leaping algorithm with path relinking(RMN-MOSFLA-PR)and the random multi-neighborhood based multi-objective intelligent water drops algorithm with path relinking(RMN-MOIWD-PR),are proposed by combining the shuffled frog-leaping algorithm and the intelligent water drops algorithm with the path relinking,respectively.Base on the discrete coding for the flexible job shop scheduling problem(FJSSP),the RMN-MOSFLA-PR and the RMN-MOIWD-PR are applied to solve the FJSSP benchmark instances to verify their performance.Base on the discrete coding for the railway freight transportation routing design,the RFTRD model is solved by the RMN-MOSFLA-PR and the RMN-MOIWD-PR.In chapter 5,base on the multi-phase particle swarm optimization(MPPSO)and the quantum-behaved particle swarm optimization(QPSO),respectively,two improved continuous multi-objective swarm intelligence algorithms,the improved multi-objective multi-phase particle swarm optimization(IMOMPPSO)and the improved multi-objective quantum-behaved particle swarm optimization(IMOQPSO),are proposed by optimizing the parameter settings,improving the update formula of the individual position and introducing a mutation operation.Base on the continuous coding for the flexible job shop scheduling problem,the IMOMPPSO and the IMOQPSO are also applied to solve the FJSSP benchmark instances to verify their performance in solving combinatorial optimization problem.Base on the continuous coding for the railway freight transportation routing design,the RFTRD model is also solved by the IMOMPPSO and the IMOQPSO.The performance of the four proposed multi-objective swarm intelligence algorithms for the RFTRD model is compared and analyzed quantitatively and qualitatively.The optimal solution selected from the best Pareto front obtained from the four proposed multi-objective swarm intelligence algorithms is input into the Simio Simulation Software for simulation analysis,to verify the performance of the four proposed multi-objective swarm intelligence algorithms.In chapter 6,considering that resources can be expanded,assuming that the location of the originating stations and the destination stations are unknown or to be determined,the resource-expandable railway freight transportation routing design(RERFTRD)system optimization model is constructed to explore the impact of the change in the transportation distance of the railway and road in the door-to-door railway traffic of wagon loading goods.Base on the coding for the resource-expandable railway freight transportation routing design,the RERFTRD model is solved by the IMOMPPSO and the IMOQPSO that verify their performance in solving continuous optimization problem by ZDT function instances.The performance of the four proposed multi-objective swarm intelligence algorithms for the RFTRD model and the two proposed continuous multi-objective swarm intelligence algorithms for the RERFTRD model is compared and analyzed quantitatively and qualitatively.The optimal solution selected from the best Pareto front obtained from the two proposed continuous multi-objective swarm intelligence algorithms is input into the Simio Simulation Software for simulation analysis,to verify the performance of the two proposed continuous multi-objective swarm intelligence algorithms.In this dissertation,from the perspective of complex system,the door-to-door railway freight transportation routing design is modeled and simulated,and solved by the four proposed multi-objective swarm intelligence algorithms,which effectively enriches the research of door-to-door railway freight transportation routing design and swarm intelligence algorithms.Furthermore,the research of this dissertation can provide decision support for the routing design and decision reference for the location of the new station and the layout optimization of the stations for the railway transportation enterprises.