Beam Tracking Based on Multi-Point Radar Sensing and DNN Prediction

Millimeter-Wave (mmWave) communication based on massive MIMO and beamforming has become a hotspot owing to its advantage on capacity. However, a rapidly varying mm Wave channel in high mobility scenario will lead to frequent beam switching and high signaling overhead. Integrated Sensing and Communication (ISAC) gives a new idea to the above challenges by using radar sensing to assist beam management. Nevertheless, most of the sensing -assisted beam tracking are limited to the scenario where Line-of-Sight (LOS) paths exist, and fail when a blockage occurs. To address this issue, a beam tracking scheme based on multi-point radar sensing and Deep Neural Network (DNN) is proposed in this paper, where the sensing information from multiple radars are aggregated and a DNN is utilized to predict the beam direction by learning the channel nonlinear correlation in a specific environment. The simulation results indicate that the proposed scheme operates effectively in blockage scenario, and surpasses the baselines in the field of spectral efficiency.