UAV-Based Pickup and Delivery with Different Customers Classes

The use of small UAVs for delivery of goods is advantageous when there is a lack of roads, or the delivery time is critical. We consider a scenario where a company offers a delivery service to small shops or private customers. The system is composed of customers, goods, and UAVs. Customers request the delivery of goods from A to B. UAVs fly to A-locations to load goods and deliver them to B-locations. We consider two classes of customers that differ in the willingness to pay an higher price to get a better service, and on their impatience: customers abandon the system after a certain time without receiving any service. The system is modeled as a semi-Markov decision process, which is embedded in each UAV. Our analysis considers two performance metrics: the average price per delivery and the fraction of served customers. These metrics cannot necessarily be jointly optimized. Consequently, we propose a distributed policy for the customers' scheduling to balance the two metrics. In this poster we outline the system model, introduce a policy for the customer scheduling, and provide preliminary results on the performance of the policy at different arrival intensities.