Research on Intelligent Millimeter Wave Beam Selection Technology for Propagation Environment Awareness

This paper focuses on mmWave communication, and based on artificial intelligence technology, studies the intelligent mmWave beam selection technology of propagation environment perception, and solves the problems of insufficient efficiency and insufficient robustness of the current beam selection technology. Firstly, the beam selection problem in the outdoor block scene is modeled, the channel model is constructed, and then the data sets in multiple scenarios are obtained by simulation. In addition, a reinforcement learning model is established according to the problem, and a deep learning network is designed to learn the implicit relationship between the optimal beam pair index in the Massive MIMO mmWave environment and the environment at that moment. Finally, the Top-K precision and mean Reference Signal Received Power (RSRP) metrics are deployed to assess the accuracy of the experimental approaches. The experimental data show that both the Top-K accuracy and the average RSRP using deep learning are close to exhaustive search, which achieves the original intention of reducing computational overhead and improving efficiency.