Abstract: In order to solve the dynamic multi-vehicle selective and repeatable pick-up and delivery problem with time windows (DMVSRPDPTW), a dynamic vehicle routing model based on dynamic requirements is established. For the real-time arrival requirements, the dynamic problem is divided into several static problems. Each time a static problem is solved, the initial solution is constructed by the insertion algorithm, and the improved variable neighborhood search algorithm is designed to improve the initial solution. In order to accurately calculate the time when the vehicle arrives at each request, the traffic time of any two points is read from the Baidu map. Multiple instances are constructed based on the historical data from the China Tower Co., Ltd illustrate: the proposed algorithm could get better results in one minute, and it has an average improvement of 46.47% compared with the existing experience of the company, and an average improvement of 30.38% compared with the insertion algorithm. The algorithm has been applied by the company in actual base station operation and maintenance to reduce the maintenance costs efficiently.

Key words: selective and repeatable pickup and delivery, dynamic requirements, insert algorithm, variable neighborhood search, base station operation and maintenance