Machine Learning and Reinforcement Learning for Load Balancing and Handover Management in Hybrid LiFi/WiFi Networks: A Comprehensive Survey

Abstract

Hybrid Light Fidelity (LiFi) and Wireless Fidelity (WiFi) networks combine two links with different strengths. LiFi can support high indoor data rates, while WiFi provides wider radio coverage when the optical link is weak or blocked. The main difficulty is that users do not remain in fixed or ideal positions. As users move, traffic changes, and access points (APs) become unevenly loaded, the network has to decide when to keep a user on the current link and when to switch the user to another AP. This review discusses machine learning (ML) and reinforcement learning (RL) methods used for AP selection, load balancing, and handover management in hybrid LiFi/WiFi networks. ML methods usually learn from channel quality, traffic load, user demand, and mobility information, whereas RL methods are used when a switching decision has later ef-fects on throughput, load balance, or handover frequency. The reviewed studies indicate that learning-based methods can perform better than fixed-threshold or rule-based schemes in dynamic indoor settings. Even so, their results are still affected by the training data, network layout, reward design, and coordination among APs. More practical evaluation settings and lighter learning models are therefore needed before these methods can be used widely in real indoor deployments.