Deep learning applications in visible light communication networks: Current trends, challenges and future outlook

Abstract

Deep learning (DL) algorithms have demonstrated remarkable performance across diverse fields, owing to their capability to capture complex relationships between variables that are hard to model using traditional machine learning techniques or analytical approaches. In visible light communication (VLC) systems, DL algorithms have increasingly attracted attention for their ability to develop robust and efficient solutions to diverse system design objectives. These include mitigating the nonlinear distortion arising from optoelectronic devices as well as interference from multipath propagation. Although existing reports on applications of DL in VLC have covered some interesting areas, emerging research continues to uncover new directions and challenges that must be addressed to align with the objectives of next-generation communication networks. Motivated by this, this article aims to provide an extensive review of up-to-date applications of DL algorithms in VLC systems. To this end, we first present a concise overview of the key features of DL algorithms relevant to VLC systems to establish the necessary background knowledge in this research area. In this context, the algorithms are grouped into three main categories: discriminative, generative, and deep reinforcement learning models. Building on this classification, our discussion is structured around current application trends encompassing areas such as modulation design and optimization, channel estimation and equalization, positioning systems, healthcare systems, secure communication, among others. In the concluding sections, we highlight the key challenges that affect the performance and practical deployment of DL techniques in VLC systems and outline several open research directions for future investigation.