Genetic Algorithms for a Scheduling Problem of Parcel Delivery and Pickup by Drones

Recent studies have been conducted actively on scheduling in a parcel delivery system using a truck and drones to deliver parcels more effectively. In logistics, it is necessary to provide not only delivery service but also pickup service, and drones can be used for the pickup service. This paper extends our previous delivery system to a new one handling both delivery and pickup and then formulates a scheduling problem for the new system. A challenge in this problem is to consider how to appropriately select one of three options: whether a drone only delivers a parcel, only picks up, or both delivers and picks up. This paper proposes two methods based on a genetic algorithm (GA) introducing procedures to select an appropriate option for solving the problem. The first method uses a maximum weight matching procedure before executing a GA, and the procedure is applied to a bipartite graph of customers whose parcels are delivered and customers whose parcels are picked up. The second uses a greedy pairing procedure to reduce drone flight time in evaluating each individual of a GA. Experimental results show that the two proposed methods can successfully find good feasible solutions. The second method can often find better solutions but consumes longer computation time than the first.