Before installing thousands of sensors across a city, what if planners could already know their impact? That question lies at the heart of DARIUS, the project led by Despina Elisabeth Filippidou, winner of the Third Open Call of dAIEDGE.
Developed by DOT-SOFT, DARIUS introduces a federated Edge AI framework that enables municipalities to explore “what-if” scenarios before investing in new infrastructure. By training models locally and preserving data privacy, cities can collaborate to predict how additional sensors may affect traffic flow, parking availability, urban comfort and CO₂ emissions.
At the core of the system is a simulation engine that turns conventional dashboards into predictive decision-support tools. Using graph neural networks and gradient-boosting techniques, DARIUS empowers city planners to evaluate alternative deployment strategies and make evidence-based decisions that support sustainable urban development.
Throughout the programe, Despina worked closely with Foundation for Research and Technology – Hellas (FORTH), whose expertise was instrumental in refining the project’s federated-learning architecture and validation methodology. We sincerely thank FORTH for its dedication as hosting institution.
The dAIEDGE-VLab also played a central role during the seven-month journey, enabling the team to assess not only predictive performance but also latency, energy efficiency and execution on edge hardware. This practical validation helped transform DARIUS from a promising concept into a mature and scalable solution ready for real-world smart-city deployments.
DARIUS demonstrates that the cities of the future will not simply collect data, they will use AI to anticipate the consequences of today’s decisions.
Discover the full story behind DARIUS in Despina’s interview on our YouTube channel.