Reinforcement Learning Based Truck-and-Drone Coordinated Delivery

Abstract

Coronavirus disease 2019 has brought a great challenge to the supply of daily necessities and medical items for home-quarantined people. Considering the unmanned operation, agility and use of clean energy of drones, we propose a novel truck-and-drone coordinated delivery system that does not require direct human contact during the delivery process. All final deliveries are completed by the drone while the truck acts as a movable charging station and a carrier, so that the contagion risks are reduced. Moreover, we spilt the whole delivery problem into the customer clustering problem and the routing problems of both the truck and the drone. An encoder-decoder framework combined with reinforcement learning is created to solve the routing problems without handcrafted designed heuristics. We design the different problem contexts specific to the truck routing problem and the drone routing problem. The experimental results show the proposed approach has good generality and can consequently be applied to problems of different scales with high time efficiency.