Reserch On Radar-assisted Multiple Base Station Cooperative Millimeter Wave Beam Tracking

The millimeter-wave band has very rich spectrum resources to support higher-speed data transmission,thus becoming a research hotspot in the communication field in recent years.Beamforming based on massive MIMO is able to significantly make up for high mmwave path loss and is a critical technology in 5G millimeter-wave communications.However,in high-speed mobile scenarios,channel changes are accelerated and frequent beam switching is necessary to ensure good communication quality,thus greatly increasing signaling overhead and resource consumption.How to reduce beam switching overhead and improve beam switching speed and efficiency in highly dynamic scenarios has become a critical problem.Integrated Sensing And Communication(ISAC),as one of the key technologies of 6G,proposes to use sensing information to assist communication,i.e.,the future location and beam direction of the terminal can be predicted with the aid of radar sensing,which provides a new idea to solve the above problems.Therefore,this thesis will focus on the new technology of millimeter wave beam tracking assisted by radar echo under the highly dynamic scenario of ISAC.Firstly,to address the challenges of high beam tracking overhead and single base station beam tracking misalignment in high dynamic scenario,this thesis proposes a radar-assisted multi-base station cooperative millimeter wave beam tracking algorithm,which uses radar echoes received from multiple base stations and Extended Kalman Filter(EKF)algorithm to cooperatively predict the optimal beamforming angle.Considering more complex road conditions,the analysis scenario is divided into vehicle linear and curvilinear mobilities.The algorithm is performed in two steps.In the mobility state prediction and calibration step,the Road Side Unit(RSU)predicts the mobility state by the mobility state evolution model,and calibrates the prediction result by the radar echo and EKF.In the multi-base station cooperation step,by deploying multiple RSUs,the Edge Server(ES)uses the distributed sensing information uploaded by each RSU for joint estimation,thus improving the accuracy of vehicle location prediction.Since the radar echoes do not require additional pilot and feedback,the beam switching overhead is effectively decreased;meanwhile,the multi-base station cooperation takes advantage of distributed sensing to significantly improve the angle prediction accuracy.Simulation results show that the multi-base station cooperation approach improves the spectral efficiency by 34%and 20%in both vehicle linear and curvilinear mobility scenarios,respectively,with a lower beam tracking overhead compared to the single base station beam tracking.Secondly,to improve the performance of multi-base station cooperative beam tracking when the line of sight(LOS)path is blocked,this thesis further proposes a multi-base station cooperative beam tracking algorithm based on radar sensing and Deep Neural Networks(DNN)prediction,which improves the decreased accuracy of multi-base station cooperation due to radar echoes being blocked by obstacles.Considering the beamforming methods in practical applications,two prediction schemes based on fixed codebook and adaptive codebook are proposed in this thesis.The algorithm is divided into three steps.In the data set generation step,the features and labels of each time slot are determined by beam traversal search.In the model training step,ES trains the DNN network with the radar echo power of each RSU and information such as Angle of Departure(AOD)or beam pair number at the current moment as features,so that the model learns the nonlinear correlation between channels at adjacent moments.In the beam prediction step,the DNN predicts the optimal base station index and its corresponding optimal beam pair direction for the next moment based on the sensed features of the current moment.The simulation results show that the spectral efficiency of the proposed DNN prediction method can reach 86%and 91%of the upper bound under both adaptive and fixed codebooks,respectively.Compared with the EKF and radar echo-based multi-base station cooperation schemes,the spectral efficiency is improved by 5.4 times and 5.1 times,respectively.