Chong, PeterUr Rehman, SaeedShakir, Ullah2025-11-192025-11-192025http://hdl.handle.net/10292/20142Mobile data traffic has increased exponentially over the last few years, from around 50 exabytes (EB) per year in 2018 to 240 EB in 2024, and is expected to further increase to around 560 EB by 2029. This increase can be attributed to the data generated by smartphones and the heavy deployment of Internet of Things (IoT) devices. Approximately 80% of mobile traffic originates from indoor devices connecting to outdoor base stations. This significantly impacts the outdoor mobile network's performance. Therefore, the Third Generation Partnership Project (3GPP) and other mobile standardization bodies are strongly encouraging mobile traffic to be offloaded to the nearest indoor WiFi network. However, WiFi has limited bandwidth, and this additional mobile traffic offload, along with the existing indoor data traffic, can result in drastic degradation in performance. To meet this growing traffic demand, Light Fidelity (LiFi), which utilises visible light for communication (VLC), has emerged as a promising candidate for an additional networking technology. LiFi offers several advantages, including higher data rates due to its operation at higher frequencies and increased security, as light can be contained within a room, thereby reducing the chances of eavesdropping. LiFi, however, lacks robustness in mobility and has a limited coverage area due to higher free space and penetration losses. For this reason, the chances of service disruptions increase, thereby leading to more frequent handovers, inefficient resource utilisation, and QoS degradation. These challenges are unique to LiFi compared to WiFi, and as such, require the necessary design and modelling effort. Therefore, this thesis provides a comprehensive modelling and implementation of the LiFi physical layer and medium access control (MAC) layer in Network Simulator 3 (ns3). At the physical layer, appropriate modulation schemes, mobility, and necessary performance metrics are modelled and implemented. The MAC layer provides a comprehensive TDMA-based design and modelling, including framing for user association and dis-association, as well as mobility-aware resource sharing. The limited coverage and mobility robustness issues necessitate that LiFi be deployed in a hybrid arrangement with WiFi and other cellular technologies to provide a comprehensive networking system. The critical aspect of integrating LiFi and WiFi networks is user association and handovers, which are initiated when user devices encounter outages and/or experience degradation in other performance metrics. However, these are challenging problems in LiFi/WiFi hybrid networks, as the handover decision-making algorithms during the association process favour WIFI, because it offers consistent signal strength over a larger area. This can result in inefficient resource utilisation, an increase in delay, and a decrease in throughput. To address this issue, this thesis provides two handover solutions. The first one is a QoS-enabled handover process that utilises multiple criteria, including user satisfaction, packet delay, and signal strength, during the network selection process, which can lead to improved performance compared to traditional handover solutions. The proposed solution is also provided with a centralised controller, capable of initiating a handover when some specific conditions are met. However, since it is reactive in nature, the handover starts only in response to channel outage events, which can lead to unnecessary packet loss. The second solution offers a proactive approach that initiates the handover in an anticipatory manner by utilising machine learning to predict the handover conditions. It is also equipped with QoS-enabled network selection and prepares multiple targets in advance to ensure seamless transition and unnecessary packet loss. The proposed solution significantly improves upon the reactive handover approach in terms of throughput and user satisfaction. Both the handover solutions have been implemented in ns-3, and the simulation results show that the proposed handover architectures outperform the benchmark algorithms.enThesisOpenAccess