Conceptual Design of ML-Based Handover Models for UAV Networks in Disaster Scenarios

Unmanned Aerial Vehicles (UAVs) play a critical role in enhancing communication resilience during natural disasters when Ground-Based Stations (GBSs) are damaged or inaccessible. Acting as Flying Base Stations (FBSs), UAVs can restore coverage and support emergency communications. However, their integration into 5G and 6G networks presents challenges, particularly with seamless handover due to dynamic mobility, frequent transitions, and velocity. These factors often lead to handover failures, ping-pong effects, and degraded Quality of Link (QoL). This paper presents the design of an Aerial Seamless Handover Framework (ASHF), built upon the proposed New Differential Handover Optimization (NDHO). NDHO utilizes a variational mathematical model optimized using the Euler-Lagrange method. The model minimizes a cost functional incorporating SINR, bandwidth availability, and UE velocity to dynamically adjust Time-to-Trigger (TTT) values, ensuring adaptive handover control. Simulation outputs from NDHO are intended to be used as input data to train machine learning models within the ASHF framework. Regression models will predict UAV trajectories, while classification models will support real-time handover decisions, determining ideal handover types and timing. This paper outlines a conceptual hybrid framework aimed at reducing unnecessary handovers and enhancing network stability, with preliminary results supporting its feasibility.