Prediction-model-assisted reinforcement learning algorithm for handover decision-making in hybrid LiFi and WiFi networks

The handover process in hybrid light fidelity (LiFi) and wireless fidelity (WiFi) networks (HLWNets) is very challenging due to the short area covered by LiFi access points and the coverage overlap between LiFi and WiFi networks, which introduce frequent horizontal and vertical handovers, respectively. Different handover schemes have been proposed to reduce the handover rate in HLWNets, among which handover skipping (HS) techniques stand out. However, existing solutions are still inefficient or require knowledge that is not available in practice, such as the exact user's trajectory or the network topology. In this work, a novel machine learning-based handover scheme is proposed to overcome the limitations of previous HS works. Specifically, we have designed a classification model to predict the type of user's trajectory and assist a reinforcement learning (RL) algorithm to make handover decisions that are automatically adapted to new network conditions. The proposed scheme is called RL-HO, and we compare its performance against the standard handover scheme of long-term evolution (STD-LTE) and the so-called smart handover (Smart HO) algorithm. We show that our proposed RL-HO scheme improves the network throughput by 146% and 59% compared to STD-LTE and Smart HO, respectively. We make our simulator code publicly available to the research community.