HEBE Optimized Mob-LSTM for Channel Estimation in RIS-Assisted mmWave MIMO System

Abstract

Channel acquisitions are one of the most significant challenges to implementing reconfigurable intelligent surface (RIS)–assisted wireless networks. Basically, the base station (BS) and the mobile station (MS) are connected to one another via the RIS. However, accurate channel state information for each individual channel is required for the RIS to perform at its highest level. Therefore, effective execution of superresolution channel estimation (CE) at the BS to RIS, RIS to MS, and composed channel is necessary. Hence, this research proposed the MobileNet–long short-term memory (Mob-LSTM) technique for the RIS-aided mmWave MIMO system in order to provide an accurate CE model. In this research, three types of channels were initially developed: BS to RIS, RIS to MS, and composed channel. After that, these three types of channel parameters are estimated with the aid of the proposed Mob-LSTM model. Additionally, this research utilized a sequential weighting method, namely, a hybrid extended bald eagle (HEBE) optimizer, for fine-tuning the hyperparameters of the Mob-LSTM. Furthermore, the proposed research is implemented and examined using the MATLAB tool. In the simulation scenario, the proposed method can outperform the various existing approaches in terms of normalized mean square error (NMSE) and mean square error (MSE). Additionally, four different scenarios have been used to assess the proposed approach's efficiency: path gain analysis and convergence analysis of Mob-LSTM, MSE, and NMSE measures. According to the simulation outcomes, the suggested method attains a lower NMSE value of −52.53 and exceeds the existing techniques with high-CE effectiveness.