RIS-Aided MISO Channel Estimation Using Fuzzy Embedded Recurrent Neural Network and Binary Kepler Optimization Algorithm

Multi-antenna wireless systems enhanced by reconfigurable intelligent surfaces (RISs) offer improved spectral and energy efficiency. RIS improves coverage and energy efficiency, but accurate channel estimation is challenging. The least-squares (LS) strategy is sub-optimal, while the MMSE estimator is difficult due to nonlinearity and non-Gaussianity. To overcome these issues, RIS-Aided MISO Channel Estimation using Fuzzy Embedded Recurrent Neural Network and Binary Kepler Optimization Algorithm (RIS-MISO-CE-FERNN-BKOA) is proposed. Initially, the Linear Minimum Mean Square Error (LMMSE) estimator, optimized with BKOA for RIS phase shifts, achieved higher accuracy than the LS approach. To further enhance the efficiency and better approximate the globally optimal MMSE channel estimator, Fuzzy Embedded Recurrent Neural Network (FERNN) is proposed. The RIS-MISO-CE-FERNN-BKOA method attain 34.56%, 25.63%, and 18.89% higher accuracy; 28.63%, 25.41%, and 19.23% lower MMSE; and 33.56%, 29.78%, and 25.74% higher SNR when analyzed with the existing techniques. The proposed technique achieves better accuracy when compared with the conventional models, making it a robust solution for RIS-assisted MISO communication systems.