MIMO optimization with deep learning: thematic analysis of 5G and beyond applications
Telecommunication Systems, cilt.89, sa.3, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 89 Sayı: 3
- Basım Tarihi: 2026
- Doi Numarası: 10.1007/s11235-026-01508-x
- Dergi Adı: Telecommunication Systems
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, ABI/INFORM, Aerospace Database, Applied Science & Technology Source, Compendex, INSPEC, Business Source Ultimate (EBSCO), Engineering Source (EBSCO), Technology Collection (ProQuest)
- Anahtar Kelimeler: AI-enabled beamforming, Channel estimation and CSI feedback, Deep learning-based wireless communication, MIMO system optimization
- Marmara Üniversitesi Adresli: Evet
Özet
Fifth generation (5G) and beyond wireless communication systems involve multi-dimensional and complex optimization problems due to requirements such as high data rates, low latency, reliable connectivity, and dense device access. Deep learning (DL)-based methods offer significant advantages in physical layer design due to their ability to model nonlinear channel behavior, perform high-dimensional representation learning, and generalize to multiple tasks in solving these problems. This study systematically reviews the current literature focusing on DL-based physical layer solutions in 5G/6G MIMO systems. Articles are grouped under thematic headings, comprehensively analyzing trends in beamforming, channel estimation and CSI feedback, resource allocation, RIS/STAR-RIS optimization, generative models, digital twin-based systems, cell-free MIMO, and ISAC applications. In addition, the citation distribution of the reviewed studies and the Q-levels of the journals in which they are published are also evaluated, revealing influential studies and prominent research trends in the literature using objective criteria. The findings demonstrate that DL methods provide significant performance gains at the physical layer, but significant challenges remain in critical areas such as data efficiency, generalizability, computational cost, and scalability to real-time applications. This study comprehensively summarizes the current state of AI-enabled physical layer design for 5G and beyond and provides a systematic roadmap for future research. In this context, this study aims to systematically identify existing achievements, current limitations, and future research opportunities in this rapidly evolving field.