The millimeter wave(mm Wave)is a potential solution for high data rate communication due to its availability of large bandwidth.However,it is challenging to perform beam tracking in vehicular mm Wave communication sys...The millimeter wave(mm Wave)is a potential solution for high data rate communication due to its availability of large bandwidth.However,it is challenging to perform beam tracking in vehicular mm Wave communication systems due to high mobility and narrow beams.In this paper,an adaptive beam tracking algorithm is proposed to improve the network throughput performance while reducing the training signal overhead.In particular,based on the mobility prediction at base station(BS),a novel frame structure with dynamic bundled timeslot is designed.Moreover,an actor-critic reinforcement learning based algorithm is proposed to obtain the joint optimization of both beam width and the number of bundled timeslots,which makes the beam tracking adapt to the changing environment.Simulation results demonstrate that,compared with the traditional full scan and Kalman filter based beam tracking algorithms,our proposed algorithm can improve the time-averaged throughput by 11.34%and 24.86%respectively.With the newly designed frame structure,it also outperforms beam tracking with conventional frame structure,especially in scenarios with large range of vehicle speeds.展开更多
基金supported by the National Key R&D Program of China(2020YFB1807204)Beijing Natural Science Foundation(L212003)。
文摘The millimeter wave(mm Wave)is a potential solution for high data rate communication due to its availability of large bandwidth.However,it is challenging to perform beam tracking in vehicular mm Wave communication systems due to high mobility and narrow beams.In this paper,an adaptive beam tracking algorithm is proposed to improve the network throughput performance while reducing the training signal overhead.In particular,based on the mobility prediction at base station(BS),a novel frame structure with dynamic bundled timeslot is designed.Moreover,an actor-critic reinforcement learning based algorithm is proposed to obtain the joint optimization of both beam width and the number of bundled timeslots,which makes the beam tracking adapt to the changing environment.Simulation results demonstrate that,compared with the traditional full scan and Kalman filter based beam tracking algorithms,our proposed algorithm can improve the time-averaged throughput by 11.34%and 24.86%respectively.With the newly designed frame structure,it also outperforms beam tracking with conventional frame structure,especially in scenarios with large range of vehicle speeds.
