A Particle Swarm Optimization Solution Approach for the Multi-Compartment Vehicle Routing Problem

• Main Authors: Lin, Jhih-Syuan, 林致瑄
• Other Authors: Han, Anthony Fu-Wha
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/81294848574936044774

碩士 === 國立交通大學 === 運輸與物流管理學系 === 103 === Multi-compartment vehicle routing problem (MC-VRP) is an extension of the classical Vehicle Routing Problem with multiple products which must be stored in the given compartment in the vehicle. The main difference between MC-VRP and VRP is the compartment capacity limit for various products and customer’s different products could be delivery separately namely set of requested products can be delivered in several times. In this thesis, we applied the particle swarm optimization (PSO) algorithm for solving MC-VRP. We modified a solution representation and the corresponding decoding method proposed by Ai and Kachitvichyanukul [1] then created a new and streamlined solution representation method to generate the initial solution for both non-split and split problem. In addition, we added the Inter-Route Exchange、2-Opt*、2-Opt and Or-Opt local improvement procedures into the PSO framework. We also improved the local search methods for the split problem and designed the 1-0* methods in response to problem features which various products could be distribute respectively. To make a more efficient implementation of PSO, we used 30 instead of 50 particles. Our proposed algorithms tested on a set of 40 benchmark problems described by Fallahi et al.[2]. By our algorithm we found out the solution gap in four patterns of problems are 1.10%, -0.26%, -1.09% and -0.82% to the best known solution, respectively. Therefore our algorithm is appropriate in solving the MC-VRP.