Spatio-Temporal SSB Prediction Using Deep Neural Network in 5G Advanced Networks

Effective Beam Management (BM) in 5G advanced networks is essential for ensuring reliable connectivity, especially in environments characterized by high mobility and varying signal conditions. Traditional methods depend on frequent measurements of Synchronization Signal Block (SSB) quality which increases power consumption and negatively impact battery life. In this paper, we propose a novel spatiotemporal prediction model that leverages Deep Learning (DL) techniques to forecast future SSB Received Signal Received Power (RSRP) in both space and time domain. By predicting the best SSB index in advance, our model eliminates the need of continuous SSB measurements. This approach significantly reduces power consumption, thereby enhancing device battery life. The model combines Convolutional and Long Short-Term Memory (LSTM) networks, which capture temporal dependencies, with Fully Connected (FC) network for spatial feature extraction, resulting in a robust prediction mechanism. We also provided the mechanism to choose between the DL and legacy mechanism to have balance between global performance and power consumption.