E.I. Glushankov1, Q.C. Pham2

1,2 The Bonch-Bruevich Saint-Petersburg State University of Telecommunications (St. Petersburg, Russia)

1 glushankov.ei@sut.ru; 2 fam.kk@sut.ru

Formulation of the problem. Currently, reducing the computational complexity of linear precoding algorithms in Multi-User Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing (MU-MIMO-OFDM) wireless communication systems remains a critical challenge. This is driven by the high computational overhead associated with high-dimensional channel matrix inversion, which scales cubically with the increasing number of antennas and users in each OFDM subcarrier, thereby hindering the practical deployment of such systems. Consequently, a comprehensive approach is required to universally replace resource-intensive matrix inversions with the iterative Preconditioned Conjugate Gradient (PCG) method for both channel estimation and precoding weight calculation. To achieve this, a predictive neural network model must be developed and deployed to determine the optimal number of iterations. This model will adaptively tune the iteration parameter based on instantaneous channel conditions, thereby eliminating the need for computationally expensive brute-force search procedures.

Objective. To propose a method for mitigating the computational complexity of core signal processing algorithms in MU-MIMO-OFDM systems through adaptive iteration control in the PCG method utilizing neural network-based prediction.

Results. This study addresses the problem of reducing the computational complexity of linear precoding algorithms and Linear Minimum Mean Square Error channel estimation in MU-MIMO-OFDM systems. A unified approach is proposed and investigated, which substitutes the computationally expensive matrix inversion operation with an adaptive iterative procedure based on the PCG method. A predictive neural network model has been developed and utilized for dynamic control over the number of iterations. It is demonstrated that the proposed intelligent mechanism accurately determines the minimum required number of iterations based on the current channel state. This achieves performance on par with classical direct matrix inversion techniques while operating at significantly lower and dynamically scalable computational costs. Furthermore, it maintains transmission quality comparable to conventional direct inversion methods, particularly in the medium-to-high signal-to-noise ratio regimes.

Practical significance. The developed method ensures the scalability of MU-MIMO-OFDM systems, paving the way for their efficient practical implementation in scenarios characterized by a massive number of users and antennas. Replacing a deterministic, resource-intensive operation with an adaptive iterative process governed by predictive iteration counts serves as a key enabler for constructing high-throughput, next-generation communication systems that demand real-time resource management.

Glushankov E.I., Pham C.Q. An investigation into methods for reducing the computational complexity of signal processing algorithms in MU-MIMO-OFDM systems via machine learning // Radiotekhnika. 2026. V. 90. № 8. P. 69−81. DOI: https://doi.org/10.18127/j00338486-202608-07

1. Kansal L., Berra S., Mounir M., Miglani R., Dinis R., Rabie K. Performance analysis of massive MIMO-OFDM system incorporated with various transforms for image communication in 5G systems. Electronics. 2022. V. 11. № 4. Art. no. 621. DOI: 10.3390/electronics11040621.
2. Krejndelin V.B., Smirnov A.E., Ben Rezheb T.B.K. Effektivnost' metodov obrabotki signalov v sistemah MU-MIMO vysokih poryadkov. T-Comm: Telekommunikacii i transport. 2016. T. 10. № 12. S. 2430 (in Russian).
3. Fatema N., Hua G., Xiang Y., Peng D., Natgunanathan I. Massive MIMO linear precoding: a survey. IEEE Systems Journal. 2018. V. 12. № 4. Р. 39203931. DOI: 10.1109/JSYST.2017.2776401.
4. Albreem M.A. Efficient iterative massive MIMO detectors based on iterative matrix inversion methods. Design Methodologies and Tools for 5G Network Development and Application. Ed. by P. Suresh, G. Vairavel, U. Saravanakumar. Hershey: IGI Global. 2021. P. 175195. DOI: 10.4018/978-1-7998-4610-9.ch009.
5. Albreem M.A., Alsharif M.H., Kim S. A low complexity near-optimal iterative linear detector for massive MIMO in realistic radio channels of 5G communication systems. Entropy. 2020. V. 22. № 4. Art. no. 388. DOI: 10.3390/e22040388.
6. Pham C.Q., Glushankov E.I., Sorotskiy V.A. Hybrid MIMO detectors with quasi-orthogonal space-time block coding and adaptive complexity management based on deep learning. Radiotekhnika. 2026. V. 90. № 3. P. 53−65. DOI: 10.18127/j00338486-202603-05 (In Russian).
7. Averina L.I., Tokarev P.A. Metody predvaritel'nogo kodirovaniya dlya sistem svyazi s tekhnologiej massive MIMO. Vestnik Voronezhskogo gos. un-ta. Ser. Fizika. Matematika. 2021. № 3. S. 514.
8. Kalachikov A.A. Analiz harakteristik algoritmov prekodirovaniya signalov v sisteme MU-MIMO s gruppirovaniem abonentov. Trudy uchebnyh zavedenij svyazi. 2023. T. 9. № 2. S. 6571. DOI: 10.31854/1813-324X-2023-9-2-65-71 (in Russian).
9. Subitha D., Vani R. Analysis of linear precoding techniques for massive MIMO-OFDM systems under various scenarios. IOP Conference Series: Materials Science and Engineering. 2021. V. 1084. Art. no. 012053. DOI: 10.1088/1757-899X/1084/1/012053.
10. Zu K., De Lamare R.C., Haardt M. Multi-branch tomlinson-harashima precoding design for MU-MIMO systems: theory and algorithms. IEEE Transactions on Communications. 2014. V. 62. № 3. P. 939951. DOI: 10.1109/TCOMM.2014.012514.130241.
11. Jiaxun L., Pingyi F. Precoding with adaptive subcarrier grouping in MIMO-OFDM systems. 2014 International Workshop on High Mobility Wireless Communications. Beijing. China. 2014. P. 103108. DOI: 10.1109/HMWC.2014.7000223.
12. Le H.A., Chien T.V., Nguyen T.H., Jeon G. Machine learning-based 5G-and-beyond channel estimation for MIMO-OFDM communication systems. Sensors. 2021. V. 21. № 14. Art. no. 4861. DOI: 10.3390/s21144861.
13. McNamara D.P., Beach M.A., Fletcher P.N. Spatial correlation in indoor MIMO channels. Proceedings of the 13th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC). Lisbon: IEEE. 2002. V. 1. P. 290294. DOI: 10.1109/PIMRC.2002.1046707.
14. Ferreira A., Gaspar G., Carvalho P., Dinis R., Moghavvemi M. A low complexity channel estimation scheme for Massive MIMO systems. Proceedings of the 2017 13th International Wireless Communications and Mobile Computing Conference (IWCMC). Valencia: IEEE. 2017. P. 234239. DOI: 10.1109/IWCMC.2017.7986292.
15. Yue Y., Zhang C., Zhang S., You X. A fast-convergent pre-conditioned conjugate gradient detection for massive MIMO uplink. Proceedings of the 2016 IEEE International Conference on Digital Signal Processing (DSP). Beijing: IEEE. 2016. P. 331335. DOI: 10.1109/ICDSP.2016.7868573.
16. Schmid L., Sulz D., Schmalen L. Fast and Robust Expectation Propagation MIMO Detection via Preconditioned Conjugated Gradient. Proceedings of the 2024 58th Asilomar Conference on Signals, Systems, and Computers. IEEE. 2024. P. 16031608. DOI: 10.1109/IEEECONF60004.2024.10942808.
17. Zhou Z., Niu Y., Wan B., Zhou W. Anti-Jamming Communication Using Imitation Learning. Entropy. 2023. V. 25. № 11. Art. no. 1547. DOI: 10.3390/e25111547.
18. Gavenski N., Meneguzzi F., Luck M., Rodrigues O. A Survey of Imitation Learning Methods, Environments and Metrics. arXiv preprint. 2024. arXiv:2404.19456. DOI: 10.48550/arXiv.2404.19456.