Using Genetic Algorithm to Solve the Shortest Driving Time Problem and Improvement of Population Initialization

• Main Authors: Hao-Tian Zuo, 左浩天
• Other Authors: Chu-Hsing Lin
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/19502145854766465880

碩士 === 東海大學 === 資訊工程學系 === 98 === In this thesis we investigate the influences on the genetic algorithm of a shortest driving time problem due to factors such as nodes on a map, the population size, the mutation rate, the crossover rate, and the converging rate. When the nodes on the map increase, more execution time is needed and larger difference between the approximate solution and the exact solution appears on running genetic algorithms. Some chromosome initialization methods may lead to initialization failure for maps of large amount of nodes, and their algorithms may be in danger of converging prematurely. Moreover, inappropriate genetic operator may extend the execution time of the genetic algorithm, and it is difficult to improve execution efficiency. The characteristics of the factors investigated in the thesis provide us insight to improve the genetic algorithm for the shortest driving time problem. Besides, from the viewpoint of the population initialization, the restart and back-and-start methods affect precision of approximate solutions and the execution time. In the thesis, we also propose an automatic mechanism for initialization: threshold values are used to automatically switch between the restart and back-and-start methods during the initialization process to have faster and superior approximate solutions.