Research On Fast Beam Search Method Based On Deep Learning In Millimeter Wave Vehicular Networking

Millimeter wave is considered as one of the most promising technologies to meet the future demand for ultra-high rate communications due to its abundant spectrum resources.With its shorter wavelength,millimeter wave can integrate more antennas in the same area compared to lower frequency bands to form large-scale antenna arrays and improve the signal transceiver gain by forming narrow beams in specific directions to overcome the high path and penetration loss of millimeter wave.The introduction of millimeter wave massive antenna array technology in vehicular network can greatly improve the data transmission rate and meet the ultra-high data transmission demand of networked autonomous driving.However,with massive antenna arrays,due to the narrow beam and high vehicle mobility,high frequency beam search is required to ensure that the beam always covers the vehicle,resulting in higher frequency guidance,time overhead and communication delay.In order to reduce the overhead and communication delay of beam search in millimeter-wave vehicular network,this paper investigates fast beam search methods based on deep learning,which are including two parts depending on the phase the beam search is in.The first part addresses the beam search in the initial access process,explores two properties of the beam search space in millimeter-wave vehicular network,i.e.,incoming vehicle direction invariance and beam symmetry,proposes a method to compress the beam search space,and proposes a fast initial access method based on recurrent neural networks and particle swarm optimization algorithms based on the correlation of the compressed beam space with vehicle traffic and time.The simulation results show that the proposed method can effectively reduce the number of beam searches and discovery delays in the initial access process.The second part explores the restrictive nature of the road on vehicle movement in the millimeter-wave vehicular network and proposes a fast beam tracking method based on LSTM network considering the correlation between the millimeter-wave channel and the location of the transceiver.Simulation results show that the proposed method can reduce the number of beam searches in the beam tracking process and improve the spectral efficiency while ensuring a lower outage rate.