Genetic Algorithm‐Optimized Recurrent Neural Network for Channel Estimation in RIS‐Aided MIMO System With Rake Receiver
Journal of Electrical and Computer Engineering · 2026-01 · 2026 (1)
DOI: 10.1155/jece/2714374 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Electrical and Computer Engineering, online erschienen am 05.07.2026. Verfasst von Kevin M. Mwongera und 2 weiteren Autor:innen. Die Arbeit nennt 52 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The integration of reconfigurable intelligent surfaces (RIS) into multiple‐input multiple‐output (MIMO) systems offers transformative potential for enhancing spectral efficiency and mitigating channel impairments. However, efficient and accurate channel estimation (CE) remains a critical challenge due to multipath propagation, high dimensionality, and the passive nature of RIS elements. This paper proposes a hybrid framework integrating a genetic algorithm (GA)‐optimized recurrent neural network (RNN) with a rake receiver (RR) to address these RIS‐aided system challenges. The proposed GA–RNN‐Rake has been applied to an end‐to‐end RIS‐assisted MIMO system developed under Rayleigh fading model in terms of a composite CE matrix. The matrix integrates three components: the direct link, the transmitter‐to‐RIS link, and the RIS‐to‐receiver link, with the passive RIS reflection constraints matrix. At the receiver, the RR resolves multipath signals using finger correlators and a combiner. RIS provides intelligent manipulation of wireless channels by reflecting signals in desired directions. The RNN architecture exploits the temporal correlations in time‐varying channel states. The GA optimizes RNN to automatically tune hyperparameters, including the number of hidden units, layer depth, and learning rate. The RR further enhances signal reception and robustness through coherent combination of multipath components to reduce intersymbol interference. Simulations in MATLAB and Simulink indicate that the GA‐optimized RNN estimator performs better than the traditional methods in terms of reduced mean squared error (MSE) and bit error rate (BER). The GA–RNN achieves a 25% MSE reduction over conventional RNNs and a 40% improvement over least squares (LS) estimation at 30 dB signal‐to‐noise ratio (SNR). The GA–RNN maintains stable CE performance under user mobility up to 120 k m / h with less than 5% MSE degradation and 25% faster convergence. The reported MSE and BER values have 95% confidence intervals (CI) and have been determined from multiple simulation runs, up to 50, to verify the trends in average MSE and BER values. However, this was achieved at a slightly increased computational complexity with the GA–RNN. The proposed system promises reliable CE for the next‐generation RIS‐aided MIMO systems, offering improved spectral efficiency, robustness, and reliability. Future work should deploy wideband CE methods and practical data in real systems to reduce complexity by 40 − 60% while maintaining estimation accuracy within ±2% MSE degradation.
Zitieren
APA
Mwongera, K. M., Langat, P. K., & Manene, F. M. (2026). Genetic Algorithm‐Optimized Recurrent Neural Network for Channel Estimation in RIS‐Aided MIMO System With Rake Receiver. Journal of Electrical and Computer Engineering, 2026(1). https://doi.org/10.1155/jece/2714374
Vancouver
Mwongera KM, Langat PK, Manene FM. Genetic Algorithm‐Optimized Recurrent Neural Network for Channel Estimation in RIS‐Aided MIMO System With Rake Receiver. Journal of Electrical and Computer Engineering. 2026;2026(1). doi:10.1155/jece/2714374
BibTeX
@article{Mwongera2026,
author = {Mwongera, Kevin M. and Langat, Philip K. and Manene, Franklin M.},
title = {Genetic Algorithm‐Optimized Recurrent Neural Network for Channel Estimation in RIS‐Aided MIMO System With Rake Receiver},
journal = {Journal of Electrical and Computer Engineering},
year = {2026},
volume = {2026},
number = {1},
doi = {10.1155/jece/2714374}
}
EUVIMED