Route Generation for Drone Delivery Service Using an Evolutionary Multi-Objective Optimization Method

Abstract

The purpose of our research is to formulate a drone delivery problem (DDP) as a constrained multi-objective optimization problem and evaluate the cost-reduction effect of a drone delivery service using the provisional-ideal-point (PIP) method proposed in this paper. The original PIP method is a genetic algorithm-based (GA-based) optimization method that can efficiently generate a preferred solution for a decision-maker. However, there are two problems occur when this method is applied to the DDP. The first problem is that there exist some cases wherein the evaluation function becomes infinite in the search process, making it impossible to sort the generated solutions. The second problem is that a long time is needed for the solution search to converge. Accordingly, the process had to be aborted at the halfway point. We present an improved PIP method to overcome these two problems. The proposed method is a solution search comprising a GA combined with tabu search. It converts the DDP into a single-objective optimization problem of a delivery cost using conversion factors. This paper presents several things understood regarding the cost-reduction effect on drone delivery services using our newly proposed method.