Hybrid Transformer-Based Channel Estimation for Massive Multiple-Input Multiple-Output Systems Using Low-Rank Adaptation Optimization
AI, Computer Science and Robotics Technology · 09.07.2026 · 5
DOI: 10.5772/acrt.20250087 ↗Kurz gefasst
Fachartikel in der Zeitschrift AI, Computer Science and Robotics Technology, online erschienen am 09.07.2026. Verfasst von Tanisha Fairuz und 3 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The escalating demands of 5G and future 6G networks, especially in massive MIMO systems, require highly accurate and efficient channel estimation. However, traditional methods face significant limitations due to high dimensionality, dynamic channel conditions, and prohibitive computational complexity, creating a significant bottleneck for practical deployment. To address these challenges, this paper proposes a novel hybrid deep learning framework that synergistically integrates a Transformer-based architecture with classical minimum mean square error (MMSE) techniques. Our proposed model leverages the Transformer’s ability to capture complex spatial and temporal correlations in the channel state information, which is important in dealing with large antenna arrays. The integration of an MMSE component enhances robustness against noise, especially in low signal-to-noise ratio environments where purely data-driven models might falter due to a lack of statistical priors. In order to enhance computation efficiency, Low-Rank Adaptation (LoRA) is incorporated into the hybrid model as an architectural optimization, introducing low-rank structures that reduce parameter redundancy without degrading estimation accuracy. Evaluated on simulated Rayleigh fading channels, the proposed hybrid framework achieves a test mean squared error (MSE) of 0.1233 and a Pearson correlation coefficient of 0.9377, while the LoRA-enhanced model attains a comparable MSE of 0.1164 with a Pearson correlation of 0.9415. Both models perform better than convolutional neural network and gated recurrent unit baselines, and achieve performance comparable to a Transformer-only model, demonstrating the feasibility of hybrid AI-based signal processing designs for next-generation wireless channel estimation.
Zitieren
APA
Fairuz, T., Noor, N. B., Al-Rizzo, H., & Milanova, M. (2026). Hybrid Transformer-Based Channel Estimation for Massive Multiple-Input Multiple-Output Systems Using Low-Rank Adaptation Optimization. AI, Computer Science and Robotics Technology, 5. https://doi.org/10.5772/acrt.20250087
Vancouver
Fairuz T, Noor NB, Al-Rizzo H, Milanova M. Hybrid Transformer-Based Channel Estimation for Massive Multiple-Input Multiple-Output Systems Using Low-Rank Adaptation Optimization. AI, Computer Science and Robotics Technology. 2026;5. doi:10.5772/acrt.20250087
BibTeX
@article{Fairuz2026,
author = {Fairuz, Tanisha and Noor, Nahiyan Bin and Al-Rizzo, Hussain and Milanova, Mariofanna},
title = {Hybrid Transformer-Based Channel Estimation for Massive Multiple-Input Multiple-Output Systems Using Low-Rank Adaptation Optimization},
journal = {AI, Computer Science and Robotics Technology},
year = {2026},
volume = {5},
doi = {10.5772/acrt.20250087}
}
EUVIMED