Predictive modeling for VLC systems: an artificial neural networks approach to real-world performance

Communication, involving voice, video, text, and data, requires expanding bandwidth capacity. Optical wireless communication (OWC), particularly free space optics and visible light communication (VLC), offers higher data transfer rates by using optical signals instead of physical cables. The performance of a VLC system is largely influenced by factors like link distance, transmitter and irradiance angles, detection area, optical concentration, and incidence angle. However, accurately predicting system performance in real-world conditions is difficult. The same components may perform optimally in one setup but fail in others, resulting in inconsistent outcomes across different link lengths and receiver configurations. To overcome this challenge, we propose a machine learning-based prediction model that evaluates system feasibility based on these known parameters. To address this, a machine learning-based model is proposed, using synthetic data from Optisystem and training various algorithms. An artificial neural network (ANN) achieved 95% validation accuracy, with a 96% ROC score, proving effective in classifying VLC outcomes. ANNs are ideal for handling complex, nonlinear data, making them suitable for real-time VLC system predictions.