Radar-Aided Beam Selection in MIMO Communication Systems: A Federated Transfer Learning Approach

By leveraging massive available data and hidden communication patterns, deep learning (DL) has enabled diverse applications in wireless network operations. In this paper, we consider radar-aided beam prediction in multi-input multi-output (MIMO) communication systems with federated transfer learning (FTL) to preserve users' location privacy. Specifically, we propose a novel structure, i.e., radar-aided federated transfer beam prediction (RaFT-BP), to achieve few samples-enabled distributed beam selection in internet of vehicles (IoV) scenarios. Simulation results show that the proposed RaFT-BP can achieve the 93.78% top-5 accuracy with 600 samples in the distributed node, enabling 11.9% to 33.2% beam selection accuracy improvement compared with baseline schemes.