Cellular-Satellite Multi-Connectivity with Link Activation Based on Random Forest Classifier

In the context of the Horizon Europe COMMECT project, we seek to develop a multi-connectivity solution that intelligently integrates cellular and satellite networks for the purpose of monitoring livestock transport in rural regions, where 5G coverage is limited. To achieve seamless connection in the multi-connectivity solution, we use machine learning (ML) based on a Random Forest (RF) classifier to efficiently integrate 5G and satellite links. The binary output of the classifier is used to activate the satellite link, in addition to the cellular link, to ensure uninterrupted connection according to the targeted performance criteria, often necessary in rural areas with limited 5G coverage. The input to the ML model are radio-related key performance indicators (KPIs). In our emulation, using experimental data, we demonstrate that our proposed solution fulfills the seamless connectivity requirements of the COMMECT project in most cases through the combined utilization of cellular networks and satellite links. This is achievable because the RF classifier, based on pre-processed radio KPIs, can accurately predict when the cellular network is unable to provide satisfactory service with a success rate of 94.3 %. The proposed solution achieves the level of application throughput required to support monitoring of livestock transport, thereby advancing communication systems for various scenarios of limited 5G connectivity.