Holistic Connectivity-Oriented GCN-LSTM for Pro-Active Handover in 5G Vehicular Networks

Connected vehicles play a crucial role in enhancing safety and comfort within Intelligent Transportation Systems. The 5G technology offers high bandwidth, ultra-low latency, and high reliability with small cell densification, which is essential for efficient data sharing among connected vehicles. However, the limited range of 5G cells presents challenges, including frequent handovers (HO), which can lead to unnecessary HOs, HO failures, ping-pong effects, and packet loss, ultimately degrading user experience. To address these challenges, a proactive handover decision strategy based on a comprehensive analysis of network conditions is necessary to ensure stable connections. We introduce a holistic connectivity-oriented hybrid spatio-temporal graph-oriented approach, HCO-PHD, which leverages Graph Convolutional Networks and Long Short-Term Memory to capture spatial dependencies and temporal patterns across the network, proactively enabling HO decision-making. HCO-PHD enhances connection stability by dynamically assessing real-time network changes and making informed HO decisions in highly mobile and ultra-dense environments. Simulation results demonstrate that HCO-PHD significantly improves the efficiency of HO decision-making, ensuring stable connectivity and minimizing disruptions caused by HOs, outperforming existing approaches.