Node-Availability-Aware Cooperative RF Identification for Low-Altitude UAVs with Distributed Multi-Receiver Sensing

This study addresses cooperative radio-frequency (RF) identification of low-altitude unmanned aerial vehicles (UAVs) under dynamic receiver availability, where single-receiver observations are susceptible to spatial variability and distributed sensing nodes may become unavailable. First, an event-level cooperative sensing architecture comprising three distributed edge RF sensing nodes and a central fusion service is developed. A unified event identifier associates observations acquired at different spatial locations, while a binary availability mask represents receiver outages, communication timeouts, and acquisition failures. Second, a node-availability-aware cross-receiver fusion framework is proposed. A shared MobileNet-1D encoder, receiver-position embeddings, and availability-masked cross-receiver multi-head attention are employed to learn complementary representations from spatial observations. During training, Random Node Dropout exposes the model to variable available-receiver sets, enabling event-level inference with three, two, or one receiver inputs. In the reported historical three-fold closed-set evaluation comprising 720 acquisition events, the Random-Node-Dropout-trained configuration achieved mean event-level identification accuracies of 98.19%, 95.14%, and 84.63% with all three receivers, any two receivers, and any single receiver, respectively. The prototype also implements an operational pipeline for multi-node acquisition, event aggregation, fusion inference, and online alerting. In a separately evaluated low-bandwidth edge-feature upload mode, the three-receiver communication payload per round decreased from 91.553 MiB to 22.5 KiB, while end-to-end latency decreased from 2226.27 to 282.21 ms. These results characterize identification performance under different receiver-input configurations and demonstrate the engineering feasibility of the distributed sensing prototype under the reported evaluation conditions.