Graph Neural Networks for O-RAN Mobility Management: A Link Prediction Approach

• DOI: 10.48550/arxiv.2502.02170
• Publication Date: 2025-02-04
• Authors: Aurelio Bermúdez; Miquel Farreras; Milan Groshev; José Antonio Jimé...; Isabel de-la-Bandera; Raquel Barco
• Keywords: Link (geometry); Ran

Mobility performance has been a key focus in cellular networks up to 5G. To enhance handover (HO) performance, 3GPP introduced Conditional Handover (CHO) and Layer 1/Layer 2 Triggered (LTM) mechanisms While these reactive HO strategies address the trade-off between failures (HOF) ping-pong effects, they often result inefficient radio resource utilization due additional preparations. overcome challenges, this article proposes proactive framework for mobility management O-RAN, leveraging user-cell link predictions identify optimal target cell HO. We explore various categories of Graph Neural Networks (GNNs) prediction analyze complexity applying them domain. Two GNN models are compared using real-world dataset, with experimental results demonstrating their ability capture dynamic graph-structured nature networks. Finally, we present insights from our study outline future steps enable integration GNN-based 6G