Deep Learning-Driven Optimal Beam Prediction for Drone Connectivity via Flying-RIS and Base Station

A deep learning model is presented for the cutting-edge wireless communication framework to ensure efficient channel estimation for a drone operating at sub-terahertz frequencies by leveraging Flying-Reconfigurable Intelligent Surfaces (Flying-RIS) in conjunction with a ground Base Station. Terahertz (THz) frequencies are envisioned as key enablers for different 6G applications such as ultra-broadband, but this technology still has its own limitations, for instance, high path loss and limited signal wavelength. Channel estimation is an essential component at such high frequencies to enhance overall system performance. To address this problem, we have employed multi-layer Long Short-Term Memory (LSTM) to handle temporal dependencies and OFDM channel complexities. The performance of our proposed approach is tested on different evaluation parameters in comparison with deep learning models including gated recurrent unit (GRU), convolutional neural networks (CNN), and fully connected networks (FCN). The findings for our proposed multi-layer LSTM are better than those of all other deep learning models across all tested parameters, proving its effectiveness in dynamic wireless communications operating at high frequencies.