A Research on Logistics Distribution Path Planning for Using Unmanned Aerial Vehicle(UAV)

• Main Authors: ZHAO,LI-YIN, 趙立愔
• Other Authors: Fang,XIA-HUA
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/7rfuz8

碩士 === 世新大學 === 資訊管理學研究所(含碩專班) === 105 === In 2013, Amazon announced that it would use UAV to do home delivery services in the future. From that time many countries of logistics industry in the world began to research the area of UAV on logistics service. However, the Federal Aviation Administration's newly-proposed drone guidelines have strict restrictions of using UAV for delivery services in the United States. The UAV flight environment belongs to 3D environment which is different from the traditional path planning for 2D environment. Therefore, in this research explores the UAV using in delivery logistics, hoping this study would provide decision makers a better choice about the arrangement of the UAV’s path planning after the relevant rules of International Air Traffic Regulation are allowed and the UAV capability has progressed. Practically, the logistics center usually has more than one conveyance to do delivery services in real life. With the change of order groups and the individual destination in this research uses K-means for clustering order groups at first, and Multiple Traveling Salesman Problem (MTSP) is regarded as solution of this study. As MTSP is a complex combination optimization problem, it is very difficult to find out optimal solution in polynomial time, and its calculation times will be with problem complex degrees into index growth. Therefore, in this study will use Genetic Algorithm (GA) to find out the path planning of UAV. GA’s characteristics of parallel multiple-point search can upgrade execution efficiency. The system is designed into four parts, order clustering, clustering TSP, genetic algorithm, and customer service level analysis. First, order clustering would use K-means for clustering order groups. And then the clustering result would execute TSP calculation for obtaining the first result of initial population. And other results of initial population are based on the first result randomly generated. Second, genetic algorithm design would start chromosome encoding used two-part chromosome technique, and customer service level is used define fitness function. The method of selection and reproduction adopts binary tournament selection method. Two-part chromosome crossover is used in Crossover and swap mutation is adopted in Mutation. The termination condition is set by the number of iterations. Finally, we would carry out customer service level analysis, if the result of path planning for UAV is within the customer service level, operation is ended. Otherwise, we will increase a UAV for delivery services, and one group will be added and clustering order groups again. This study has been proved to obtain an effective result. The decision maker could make decision easily by this model in the future.