The increasing demand for low-latency, high-speed content delivery in Mobile Edge Computing (MEC) environments, particularly in indoor settings, necessitates efficient resource management and intelligent caching strategies. Hybrid networks that combine Visible Light Communication (VLC) and millimeter Wave (mmWave) technologies offer significant advantages in high data rates and precise localization. However, dynamic user mobility, varying content demand, and resource limitations pose challenges for optimizing network performance. This paper proposes a personalized and dynamic caching strategy for software-defined hybrid VLC-mmWave MEC networks. The strategy integrates a transformer-based model for accurate user mobility prediction with Neural Matrix Factorization (NeuMF) for content preference modeling. By forecasting users’ future locations and content demands, the approach enables proactive and personalized caching, optimizing cache hit ratios, reducing latency, and minimizing energy consumption. A Soft Actor-Critic (SAC) reinforcement learning was applied to obtain optimal caching strategy, enhancing caching decisions, ensuring stability and adaptability in a dynamic environment. Extensive simulations demonstrate significant improvements in cache performance, latency reduction, and energy efficiency compared to existing benchmarks. This work provides a scalable solution for improving MEC network efficiency, particularly in indoor environments leveraging hybrid VLC-mmWave technologies.
Kombo et al. (Wed,) studied this question.