Abstract
Millimeter wave (mmWave) bands formulate the standalone (SA) operation mode in the new radio (NR) access technology of 5G systems. These bands rely on beamforming architectures to aggregate antenna array gains that compensate for dynamic channel fluctuations and propagation impairments. However, beamforming results in directional transmission and reception, thus resulting in beam management challenges, foremost initial access, handover, and beam blockage recovery. Here, beam establishment and maintenance must feature ultra-low latencies in the control and data planes to meet network specifications and standardization. Presently, existing schemes rely on arrays redundancy, multi-connectivity, such as dual-beam and carrier aggregation, and out-of-band information. These schemes still suffer from prolonged recovery times and aggregated power consumption levels. Along these lines, this work proposes a fast beam restoration scheme based on deep learning in SA mmWave networks. Once the primary beam is blocked, it predicts alternative beam directions in the next time frame without any reliance on out-of-band information. The scheme adopts long short-term memory (LSTM) due to the robust memory structure, which uses past best beam observations. The scheme achieves near-instantaneous recovery times, i.e., maintaining communications sessions without resetting beam scanning procedures.
Funder
Deanship of Scientific Research, King Faisal University
Subject
Fluid Flow and Transfer Processes,Computer Science Applications,Process Chemistry and Technology,General Engineering,Instrumentation,General Materials Science
Reference37 articles.
1. Minimum Requirements Related to Technical Performance for IMT-2020 Radio Interface,2017
2. Channel Estimation and Hybrid Precoding for Millimeter Wave Cellular Systems
Cited by
7 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献