| Millimeter wave is one of the key technologies of 5G,which can effectively solve the problem of insufficient frequency spectrum resources.However,due to its physical properties,the signal will suffer from severe path loss and atmospheric absorption during propagation,and the range of signal transmission is thus limited.To overcome these drawbacks,beamforming technology based on the large antenna array has become an important technology in mmWave systems.Beamforming can concentrate the energy of the signal in a certain direction to resist losses encountered during propagation,and the directional beam generated by the beamforming method can also reduce interference.For the mmWave system,obtaining reliable and efficient system performance is crucial.This paper investigates two aspects:maintaining the stability of the beam link and obtaining channel information.Regarding the aspect of maintaining the stability of the beam link,this paper focuses on optimizing the beam handover process.By using reinforcement learning methods to obtain an adaptive beam handover threshold and comparing it with the fixed threshold method,this paper models the beam handover scenario in the mmWave system with reinforcement learning and matches different states in the Q table with different environments after training with the Q-Learning algorithm.The simulation results show that the adaptive beam handover algorithm based on reinforcement learning can reduce the frequency of beam handover while ensuring the performance of service beams.For the channel information acquisition,this paper mainly considers reconstructing the complete channel using beam information obtained through periodic beam sweeping in the mm Wave scenario of FDD.Due to the sparsity and correlation of mm Wave channels in space,this paper uses deep learning methods to reduce the feedback overhead of beam information by exploiting these features,and improves the performance of channel estimation under limited feedback conditions.In general,this paper combines different machine learning algorithms to design the corresponding algorithm in the corresponding scenario of millimeter wave,proposes innovative algorithms in beam handover and limited feedback channel estimation,and optimizes the performance of the system in the corresponding scenario. |
