Request prompt response in same-day delivery problem with drone resupply

The present research considers the request for prompt response in the same-day delivery (SDD) problem with drone resupply. SDD has gained popularity since customer satisfaction is highly valued in the logistics industry. Customers, even those residing far from distribution centres, anticipate the opportunity to get prompt responses. The decision-maker (DM) promptly decides whether to accept the order or not as customers’ requests stochastically occur throughout the day. We assume that the truck promises to deliver packages within the promised time while a drone performs multiple trips from the warehouse to replenish the truck at any required time. A novel Deep Q-Learning (DQL) approach is proposed to maximise the acceptance rate of requests and simultaneously ensure prompt responses. Two nested agents are utilised to (a) determine order acceptance and (b) dynamically adjust drone deployment based upon its availability. Comprehensive testing and analysis demonstrate the superior effectiveness of our approach compared to benchmarks. Findings suggest that: (1) Drone resupply enables SDD to remote customers within designated distances; (2) Deep Q-learning optimises drone’s waiting times to dynamically adjust payload capacity; and (3) it is the increase in resupply frequency rather than payload size per resupply that more effectively improve the whole delivery volume.