Joint Fragment Dissemination and Edge Fusion for Fast Target Detection in UAV-Assisted Urban IoT

Abstract

Urban edge-camera and IoT sensing networks increasingly support time-critical monitoring, where rapid target detection is essential in emergency scenarios. In these environments, Unmanned Aerial Vehicles (UAVs) serve as mobile data ferries to disseminate high-value recognition payloads to distributed ground nodes. However, large payloads are severely constrained by short air-to-ground (A2G) contact windows and packet-level impairments such as path loss, shadowing, smallscale fading, and stochastic packet errors (PER). This paper addresses the problem of minimizing mission-level time-to-detection under realistic packet-level PHY constraints. We propose a coordinated framework that partitions recognition payloads into compact fragments. A UAV continuously broadcasts these fragments, rather than the full payload, along a coverage-aware trajectory, while ground nodes cooperatively exchange and fuse missing evidence through localized, confidencetriggered collaboration. Unlike monolithic UAV-only delivery and infrastructure-bound cooperative inference, our design explicitly couples fragment broadcasting with edge cooperation under packet-level feasibility. Simulation results show that the proposed framework significantly reduces time-to-detection compared to monolithic transmission and UAV-only baselines, particularly under stringent A2G conditions with short contact windows and high packet error rates. The results highlight the value of jointly optimizing fine-grained dissemination, edge cooperation, and PHY-aware communication in UAV-assisted urban sensing systems.