3-D LiDAR and GPS Aided Beam Tracking in Millimeter Wave Vehicular Communications

Vehicle-to-vehicle (V2V) millimeter wave communication scenarios have attracted more and more attention. One important problem in V2V mmWave communication is continuously tracking the narrow beams with the movement of vehicles. To our best knowledge, our work is the first method fusing measured 3-D LiDAR and GPS data to aid beam tracking in V2V scenario. We develop a multi-modal complementary scheme aided beam tracking (MCBT) method, including data preprocess, index decision part (IDP), direction decision part (DDP) and direction limitation mechanism (DLM). To evaluate the method, we use the DeepSense 6G real-world V2V dataset. The results show that Top-5 and Top-20 accuracy can reach about 76% and 92% among the overall 256 beams. We also prove by experiment that under the same deep neural network after the extracted feature, multi-modal data can improve the prediction accuracy about 8% in Top-5 accuracy.