Learning based Beam Tracking in 5G NR

Abstract

We propose a beamforming framework model which contains multiple mobile users in a 5G NR setting. The base station (BS) synchronizes with the user equipment (UE) using Synchronization Signal Bursts (SSB) that are beamformed in different directions. After establishing connection with the UE and determining the UE's location, it is important to keep track of the UE's position coordinates with respect to the BS to transmit data with an appropriate beamforming angle. The data directed at an angle aligning with the UE's location shall increase the signal to noise ratio (SNR) and reduce the EVM of the received waveform. Our goal here is to predict and track the UE's location using an efficient deep learning algorithm. We shall use the Deep SORT algorithm for this purpose. The predicted information regarding the UE location shall be helpful in directing data towards its location without any feedback mechanism involving channel state information reference signal (CSI-RS) report feedback. Ergo, the proposed model shall help to reduce the feedback and thus improves the overall network efficiency. Simulation results consisting of error vector magnitude (EVM) and constellation plots demonstrating the best transmission parameters and beamforming angle among all the SSB blocks transmitted are presented in this paper.