Distributed UAV Detection and Classification in Cell-Free ISAC With Real RCS Signatures
This paper proposes a distributed sensing framework for Cell-Free (CF) Integrated Sensing and Communications (ISAC) aimed at efficient target detection and classification. To mitigate fronthaul overhead, we implement a decentralized architecture where Access Points (APs) perform local detection and Radar Cross-Section (RCS) estimation, forwarding only refined metrics to a Central Processing Unit (CPU). We adapt the Partially Informed Maximum A Posteriori Ratio Test (MAPRT) detector to a monostatic scenario, incorporating an AP selection strategy tailored for the realistic 3GPP Urban Macro (UMa) channel model. Beyond binary detection, the framework integrates a multi-stage classification layer based on Gradient Boosting Machines and Recursive Bayesian Classification (RBC). By leveraging empirical mmWave RCS measurements of commercial drones, our approach enables high-accuracy Unmanned Aerial Vehicle (UAV) recognition. Simulation results demonstrate that the proposed framework achieves robust detection and classification performance even in sparse AP configurations. Notably, we show that exploring temporal diversity through RBC allows a reduced number of APs to match the accuracy of denser deployments over time, offering a scalable and communicationefficient solution for 6G environmental awareness.