2020 Volume 85 Issue 775 Pages 1877-1886
In recent years, the number of packages to be delivered has been rapidly increasing according to the pandemic usage of online shopping, while hireling delivery persons is becoming rather difficult. Given this background, some delivery companies employed a Team Delivery System (TDS), in which a Sales Driver (SD) and a Field Cast (FC) cooperate with each other to deliver packages efficiently. SD drives a truck to carry all packages and handover some packages to FC at parking spots. SD and FC deliver packages separately until the time to meet at the next parking spot. In this research, we proposed a method to find the best locations of parking spots and the best travel route for delivery so that the time to finish delivery of all the packages is minimized.
First, we formulated a Delivery Traveling Problem (DTP) as a mathematical model, which can be solved by minimizing an evaluation index including the total delivery time. This problem is regarded as a variation of multiple Traveling Salesman Problem, which is a one of NP-hard problem. In order to solve this problem efficiently within finite time, we proposed two new methods by combining the well-known methods that have been used for multiple Traveling Salesman Problems.
Next, we conducted the field surveys on the actual delivery service activities by Yamato Transport Co., Ltd, which is the largest delivery service company. Based on the surveys, we estimated basic statistics, such as the number of packages to deliver, the service time at delivery destinations, and the movement speed of delivery persons, and constructed a simulation model to describe the movement of SD and FC.
Next, we validated the effectiveness and usefulness of our proposed method by comparing with the existing methods of heuristics through the numerical simulation on the assumed virtual area. More specifically, employing the basic statistics obtained by surveys to the parameters of simulation, we compared each methods and examined the most efficient method. As the result, we showed the best method is the following procedure. (1) cluster delivery destinations into some zones using the fuzzy c-means method, (2) determine the locations of optimal parking spots for package handover using the genetic algorithm, (3) cluster the delivery destinations for each delivery person using the perspective ratio.
Next, we executed the delivery simulation in the survey areas and analyzed the movement of delivery persons. Comparing the actual-delivery route and the optimal-delivery route calculated by the proposed method, the latter's total delivery time was shorter. Hence, it was shown that more efficient delivery can be achieved by optimizing the locations of parking spots and travel routes.
Finally, we evaluated the performance of delivery person combinations by calculating the social cost for delivery activities. The social cost is an economical evaluation index obtained by accumulating the labor costs and the environmental costs in addition to the travel distance and the total delivery time. It was shown that the social cost varied depending on the areas, and the evaluation of the actual delivery person combination was excellent. In other words, it was found that the optimal delivery was practiced with comprehensive consideration of time, economy and safety, based on the long experiences in the actual delivery operations.
