Deep Learning-Based Beam Pair Prediction With Finite Beam Quality Information.

With the development of 5G/B5G wireless communication technology, millimeter wave is the key to achieve ultra-high data transmission rate, and beam management is an effective means to reduce the path loss and expand the coverage of high-frequency communication. This paper proposes deep learning (DL)-based spatial and temporal beam prediction schemes, which utilize finite beam quality information. Compared with non-AI methods, the proposed DL-based method can significantly improve beam selection accuracy and/or reduce training overhead, and is suitable for deployment in lightweight terminal equipment with low energy consumption and small storage.