We Will Find You: An Edge-Based Multi-UAV Multi-Recipient Identification Method in Smart Delivery Services

Unmanned aerial vehicle (UAV) is increasingly becoming a promising solution for last-mile delivery in smart logistics, and multi-UAV scenarios have become increasingly common. In multi-UAV delivery services, the ability to accurately and efficiently identifying multiple target recipients is a critical issue. Face recognition has been widely used in UAV delivery services but since the UAVs need to passively wait for the target recipient to arrive at the designated location, the efficiency of identification is undesirable. Furthermore, as one UAV is intended only for one recipient, there is no collaboration between multiple UAVs at the same location which could otherwise reduce the computational and communication time for some common tasks. To address the aforementioned issues, in this paper, we propose an active multi-UAV multi-recipient identification method (named MM4ID) in edge-based smart delivery services, based on person re-identification (ReID) technology. Specifically, multiple UAVs scan the recipient’s activity area and transmit their video streams to a nearby edge server. In the meantime, a multiple object tracking (MOT) algorithm is performed at the edge server to track the recipients and obtain one image for each recipient from the video streams to reduce duplicated data. We perform person re-identification algorithm on the obtained recipient images to determine the location of the recipient, and then we use the face recognition algorithm to confirm the identity of the recipient. Finally, successful identification results including the target recipients and their locations are sent to all UAVs via broadcast so that each UAV can complete their final delivery task. Experimental results based on a real-world edge-based smart UAV delivery service system successfully demonstrate the effectiveness of our proposed method and yield better performance compared with other representative solutions.