LAUSR

Real-time matching of millimeter wave (mmWave) narrow beams is a great challenge in dynamic vehicle to infrastructure (V2I) environments due to the mobility of vehicles. In this paper, a novel beam search strategy based on spectrum-environment awareness is proposed. Combining the technique of label iterative optimization with three-dimensional (3D) grid encoding, the strategy treats the optimal beam pair indexes (BPIs) as labels and encodes the environments as features. Three-dimensional grid encoding is a symmetry-based environmental coding technology. A new Convolutional Neural Network (CNN) model is also constructed, which is trained by the features. The situational beam search of actual vehicles is performed under the trained CNN model. As a result, real-time mmWave narrow beam matching can be achieved. Simulation results demonstrate that the proposed strategy can effectively reduce the beam search overhead and improve the efficiency while guaranteeing the matching accuracy.