This project brings together the power of the EOSC Federation to deliver a fully reproducible workflow for generating urban mobility datasets and training AI models that predict handovers for connected vehicles, a challenge at the heart of next-generation transportation systems.
By tapping into federated resources such as the EOSC Node Poland for data storage and processing, and GPU infrastructure in Skopje for AI training, the research demonstrates how geographically distributed, heterogeneous resources can be seamlessly woven into a single, coherent scientific environment.
At its core, the workflow pairs SUMO traffic simulation with OMNeT++ 5G network modelling to produce rich, fine-grained datasets capturing both vehicle telemetry and network performance metrics. These datasets then fuel AI models trained to anticipate which cell a vehicle will connect to next, based on its historical mobility patterns, a capability with real implications for smoother, more reliable connected driving experiences.
Every output, datasets, AI models, and training scripts, is published under FAIR and Open Science principles, making the work fully accessible and reusable by the broader research community. Beyond the data, the project will contribute structured training materials to the EOSC Academy, fostering peer-to-peer learning and empowering other teams to replicate these workflows across domains like intelligent transportation and mobile communications.
Ultimately, this initiative is a concrete demonstration of what the EOSC Federation makes possible: complex, interdisciplinary research that no single institution could efficiently tackle alone.
The project supports EU Open Science priorities and EOSC Federation objectives by delivering FAIR datasets, open-source AI methods, and a reproducible scientific workflow executable across federated infrastructures. By publishing data, code, and documentation openly and with persistent identifiers, it contributes to the practical implementation of FAIR principles and strengthens EOSC as a trusted environment for cross-border, reproducible research.
Benefits for researchers are immediate and concrete: access to realistic urban mobility and 5G network datasets, plus a reference AI model for predicting a vehicle’s next serving cell. This enables benchmarking, method comparison, and rapid experimentation without requiring complex local setups. The published workflow allows others to reproduce results, adapt scenarios to new cities or traffic densities, and extend the model for related research.
Participating organisations each gain something tangible. EOSC Node Poland strengthens its portfolio by hosting a visible, high-demand scientific workflow and gains operational experience for similar projects. FINKI validates how specialised GPU infrastructure can complement Node services and contribute reusable AI pipelines. UMH demonstrates practical cross-border collaboration by integrating distributed resources into a single reproducible workflow and producing publishable results and documentation. For the wider EOSC community, the consortium delivers a tested reference workflow and a concrete example of federation in practice.
The project enhances FAIRness through rich metadata, clear licensing, and thorough documentation, while executable notebooks and open repositories with persistent identifiers ensure long-term interoperability, reproducibility, and accessibility through EOSC channels.