GCN-ATO: GCN-Assisted Task Offloading for Multi-Access Edge Computing in Hetnets

The emergence of 5G and the upcoming 6G wireless communication technologies driven the need for innovative network architectures, such as Heterogeneous Networks (HetNets), which integrate diverse components to boost network capacity and performance. Mobile Edge Computing (MEC) offers a promising solution for low-latency computation requirements for different application and use cases at the network edge, but effective task offloading in dynamic HetNet environments remains challenging due to factors like user mobility, channel conditions, and frequent handovers. This paper proposes a Graph Convolutional Network-Assisted Task Offloading (GCN-ATO) algorithm, which leverages a trained GCN model to predict Key Performance Indicators (KPIs) based on current network conditions and user positions, thereby optimizing task offloading decisions. By reducing unnecessary data transmission and enhancing resource utilization, the GCN-ATO algorithm improves energy efficiency and task execution latency. We evaluate our approach using real-world mobility and network traces, demonstrating significant reductions in energy consumption compared to baseline schemes, while preserving task execution deadlines and system performance.