Towards an Intelligent and Proactive Multi-Domain Handover for 5G networks

5G networks offer high throughput, ultra-low latency, and flexible service differentiation, enabling support for a wide range of applications. However, mission-critical services—such as public safety—require uninterrupted, resilient, and secure connectivity across heterogeneous network domains. One of the main challenges in this context is multi-domain handover (MDHO), where transitions between different operators or administrative domains may introduce significant delays and service degradation due to heterogeneous policies, fluctuating radio conditions, and complex signaling procedures.

In this project, we propose to develop predictive machine learning (ML) models that proactively trigger MDHO in 5G networks to ensure seamless transitions across operators and network domains. The scientific foundation of these models lies in their ability to analyze real-time and historical network conditions to anticipate handover events, thereby minimizing handover latency and preserving service continuity. The proposed models will be trained on cross-domain datasets collected from our Open RAN testbed and subsequently integrated into a multi-site 5G testbed, leveraging collaborative and federated learning (FL) approaches to enable scalable and privacy-preserving model training.