FDD Massive MIMO Channel State Information Compression and Reporting Technology Based on AI/ML—Take Reducing Feedback Overhead as the Main Line

Yi Jiang; Xianmeng Wang

doi:10.54254/2755-2721/2026.37637

Abstract

FDD Massive MIMO is an important enabling technology for 5G-Advanced and 6G, but its downlink channel state information (CSI) feedback overhead increases significantly with the increase in the number of antenna array elements and has become the main obstacle affecting system performance. The traditional schemes based on codebooks and compressed sensing rely on preset constraints and a step-by-step processing approach, and their compression performance has approached the optimal compression ratio. In recent years, AI/ML methods represented by deep learning have provided a new idea of "Data-driven end-to-end joint optimization" for CSI compression and feedback. However, these AI/ML methods have several problems such as complex plans, inconsistent assessment criteria, and difficulties in implementing standardization. This article takes reducing feedback overhead as the main line, classifies the existing schemes into two categories: autoencoders and Transformer architectures, and comprehensive overview of methodological principles, strengths, and weaknesses of each type of method. In addition, the article links the latest research results to the standard progress of 3GPP Rel-18/19/20 and points out the gap between the existing work and the standard. Also, this article offers multi-objective optimization suggestions and a deployment plan by combining parameter size, inference latency, energy consumption. Finally, the article proposes generalization ability‌, lightweight design, and the capability boundaries of large models as future research directions, with the aim of promoting the early practical application of AI/ML-based CSI feedback technology.

Keywords: FDD Massive MIMO , CSI , Autoencoder , Transformer

Article Information

Received: 9 September 2026

Published: 13 October 2026

DOI: https://doi.org/10.54254/2755-2721/2026.37637

Article Type: Research Article

Cite this article

Jiang,Y.;Wang,X. (2026). FDD Massive MIMO Channel State Information Compression and Reporting Technology Based on AI/ML—Take Reducing Feedback Overhead as the Main Line. Applied and Computational Engineering,271,225-231.

Data availability

The datasets used and/or analyzed during the current study will be available from the authors upon reasonable request.

Disclaimer/Publisher's Note

The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of EWA Publishing and/or the editor(s). EWA Publishing and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

About this Volume

Volume Title: ACE Vol.271

Part of Series: Applied and Computational Engineering

ISSN: 2755-2721 (Print) / 2755-273X (Online)

Copyright & License

[1]Liu, X., Gao, S., Liu, B., et al. (2026). WiFo-CF: Wireless foundation model for CSI feedback.IEEE Transactions on Wireless Communications, 25, 15039–15053.

[2]Zhaohua, L., Chenyang, Y., Jie, W., et al. (2026). Deep CSI compression and feedback for massive MIMO: A survey.ZTE Communications, 24(1).

[3]Guo, J., Wen, C. K., Jin, S., et al. (2022). AI for CSI feedback enhancement in 5G-advanced.IEEE Wireless Communications, 31(3), 169–176.

[4]Xiao, H., Wang, Z., Tian, W., et al. (2021). AI enlightens wireless communication: Analyses, solutions and opportunities on CSI feedback.China Communications, 18(11), 104–116.

[5]Radwan, A. Y., Muhammad, F. S., Baker, M., et al. (2026). Contrastive predictive coding with compression for enhanced channel state feedback in wireless networks.IEEE Transactions on Neural Networks and Learning Systems.

[6]Zhuang, J., Wang, Y., Hou, H., et al. (2025). OLAMCF: Offline large AI models enhanced CSI feedback in FDD massive MIMO systems. InGLOBECOM 2025 - 2025 IEEE Global Communications Conference(pp. 3517–3522). IEEE.

[7]Khan, F. M. A., Hallaq, M., Abou-Zeid, H., et al. (2026). Model compression for sustainable AI in xG wireless networks: Recent advances, challenges, and future directions.IEEE Communications Surveys & Tutorials.

[8]Zhai, Q., & Huang, J. (2026). Design of a low-power constraint communication AI fusion model compression algorithm.Neural Computing and Applications, 38(12), 486.

[9]Li, J., Li, R., Zhao, Z., et al. (2025). Wireless digital twin network enabled model selection for AI based CSI compression. In2025 IEEE Globecom Workshops (GC Wkshps)(pp. 1863–1868). IEEE.

[10]Shehzad, M. K., Rose, L., & Assaad, M. (2021). Dealing with CSI compression to reduce losses and overhead: An artificial intelligence approach. In2021 IEEE International Conference on Communications Workshops (ICC Workshops)(pp. 1–6). IEEE.

[11]Huang, C.-H., Wen, C.-K., & Li, G. Y. (2025). AI/ML life cycle management for interoperable AI native RAN (arXiv:2507.18538).arXiv.

[12]Gao, Y., Wu, X., Jiang, J., Hu, B., Du, J., Ye, Q., Zhang, S., Yu, F. R., & Xu, S. (2026). AI/ML for mobile networks: Current status in Rel. 19 and challenges ahead (arXiv:2603.14317).arXiv.