LSTM-based Channel Estimator for Optical IRS-Assisted non-Linear VLC Systems

Visible Light Communication (VLC) emerges as an energy-efficient, eco-friendly, and economically viable technology for advanced high-speed communication systems, complementing traditional radio-based technologies. Nevertheless, the effectiveness of VLC systems is hindered by challenges such as ambient noise, non-linear characteristics of light-emitting diodes (LEDs), and signal loss in VLC transmission due to the absence of line-of-sight (LoS) link between the transmitter and receiver because of blockages by humans/users or objects in the environment. To address the problem of dead zones, an optical intelligent reflected surface (IRS) is proposed in the literature, which allows to create a multi-path channel to provide coverage to users in the blockage. This work considers an indoor multi-user IRS-aided system, with mirror array (MA) as IRS elements. For channel estimation, deep learning-based methods are becoming popular. These methods either have high computing complexity or perform poorly in complex mobile-multipath circumstances. In this manuscript, the constraints are alleviated through the utilization of an implicit channel estimation algorithm based on Long Short-Term Memory (LSTM). We propose an LSTM-based implicit channel estimation and symbol detection algorithm for a non-linear VLC system. Simulations show superior performance of the proposed LSTM-based receivers over traditional minimum mean square error and zero-forcing-based algorithms.