AI and ML Strategies for Robust Beam Management in 5G Systems

Le résumé fourni par la source

The millimetre wave (mmWave) frequency band (sub-6 GHz), spanning from 30 to 300 GHz, is the most important among the wide range of frequency bands used in fifth-generation (5G) wireless communication systems. Although extremely broad bandwidth and spectrum resources can be utilised in the mmWave band, mmWave signals suffer from a severe path loss. This is mitigated by employing massive multiple-input multiple-output (MIMO) with beamforming. However, each beam (spatial beam in two-dimensional space and temporal beam in the time domain) has its unique quality, leading to the necessity of measurements from the users or their side information for reports for connecting the users to the best beams. Unfortunately, these measurements and reports incur high time, power, teaching-learning overheads. In this paper, we examine the feasibility of using artificial intelligence (AI) and machine learning (ML) to improve the beam management by simply predicting the best beams in this connection to millimetre wave (mmWave) technology. Results show that the AI/ML models greatly outperform the traditional method in several key performance metrics including accuracy, latency and throughput. In addition to their high classification and prediction accuracy, AI/ML models also have a good generalization capability across different geographical environments and a solid robustness and reliability against channel estimation and quantization errors. The findings are expected to promote the use of AI/ ML models in future wireless networks and help achieve more efficient and reliable beam management in 5G networks.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.