Improved Particle Swarm Optimization and Its Application to Travelling Salesman Problem

• Main Authors: Tun-han Yao, 姚敦瀚
• Other Authors: Chieh-li Chen
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/80151291086895203548

碩士 === 國立成功大學 === 航空太空工程學系碩博士班 === 96 === This dissertation proposed a improved version of Particle Swarm Optimization (PSO) to solve Traveling Salesman Problem (TSP). This evolutionary algorithm includes two phases. First phase includes Fuzzy C-Means clustering, a rule-based route permutation, a randomly swap strategy and a cluster merge procedure. This approach generates an initial route without cross link problem, which make TSB problem can be solved more efficiently by the proposed PSO algorithm. The use of sub-cluster to solve TSP is also a novel procedure in the literature, which reduces the complexity and obtains better performance for problems with a large number of cities. The second phase is the use of the improved PSO procedure, a Genetic Algorithm hybrid strategy, which solves the TSP with better performance. In this study, the space transformation is applied for route tour in which space vector transfers into tour through mapping function. This method could decrease the complexity and raise the evolutionary efficiency. Fixed runtime was applied to compare the efficiency of novel algorithm with other algorithms. This research also focuses on the comparison of clustering efficiency, and the result reveals that the clustering method reaches a better result for the cases with a large number of cities.