Visible light communication-channel-adaptive digital pre-equalization scheme based on a deep learning model

Abstract

To meet the anticipated future demand for optical wireless high data rates, our study proposed a deep learning model-based adaptive digital pre-equalization scheme for visible light communication (VLC) channels is proposed to increase the data rate. We established the deep learning model to predict pre-equalization parameter (PEP) which is adaptive for various VLC channels. The proposed system is flexible and adaptive for commercial light emitting diode with different channel conditions. It improves the bandwidth of different VLC channels up to 125 MHz and has been validated on a 250-Mbps testbed. Based on the experimental results, the PEP update time is at most 5.1 s, and the bit error ratio is always <1 × 10 − 6 while using the PEP in the deep learning network. Meanwhile, the predicted and optimal values of the PEP correspond at 250 Mbps, and the maximum error between the predicted and optimal values of the PEP are only 7 for various channel conditions, including optical power, data rate, send distance, and send angle. The adaptive pre-equalization capability of the proposed system would be a universal digital solution for high-speed access in 6G scenarios combined with the VLC spectrum.