Federated Learning-Based Joint Radar-Communication mmWave Beamtracking with Imperfect CSI for V2X Communications

Robust beamforming is used to ensure communication quality, particularly for vehicle-to-everything (V2X) communications. Moreover, millimeter waves (mmWaves) are used in a new era of communication, providing a high data rate and high spectrum efficiency. However, mmWave links are prone to short-range communication because of serious free-space path loss and blockage by obstacles. In addition, owing to constraints on bandwidth and privacy considerations, it is computationally difficult to transmit the entire dataset between vehicle users and road side units (RSUs). As a result, FL-BT, a federated learning (FL)-based joint radar and communication mmWave beamtracking with imperfect channel state information (CSI) for V2X communications, is proposed. FL-BT is a federated module at vehicle user (local clients) and an RSU (main server) consisting of support vector regression (SVR) along with an unscented Kalman filter (UKF) that predicts the distribution of future channel behavior based on a sequence of input signals available. The simulation results demonstrate that the proposed FL-BT approach offers significant improvements in V2X communication performance by incorporating the UKF and SVR in a state-space model for precise future channel-value prediction and a federated learning framework for handling bandwidth and privacy considerations. This leads to a more robust and accurate beamtracking performance compared to traditional beamtracking methods.