DARIUS
Federated Learning and Edge AI for Smart Urban Planning
DARIUS is an innovative project that introduces a Federated Learning (FL) framework with edge computing capabilities, with the goal of transforming urban planning in European smart cities.
Detailed project description
Unlike traditional cloud platforms, which centralize raw data in central systems, DARIUS maintains data sovereignty by training artificial intelligence models locally via municipal browsers, using WebAssembly and accelerated by WebGPU. This collaborative approach among cities allows municipalities to share knowledge and intelligence without exchanging sensitive data.
A key differentiating feature of DARIUS is its “what-if” simulation engine, which enables urban planners to quantify the impacts of changes to infrastructure before they are physically implemented. Such impacts may include reducing CO₂ emissions, alleviating traffic congestion, and improving the time it takes to find a parking space.
Integrated into DOTSOFT’s NOON IoT platform, DARIUS shifts the focus from historical dashboards to data-driven and predictive planning, covering areas such as mobility, air quality, and waste management.
In line with the European Green Deal and the EU Regulation on Artificial Intelligence, the project promotes reliable, low-latency Edge AI. By bridging the gap between research and the market at Technology Readiness Level (TRL) 4–5, DARIUS provides a scalable solution for more than 300 existing facilities, enhancing data sharing and sustainable urban development, without the high cost of data centralization.
Type and scope of work provided
- Federated AI Framework: Development of a browser-based federated learning system using WebAssembly and WebGPU to enable collaborative training between cities without transferring raw data.
- “What-If” Simulation Engine: Development of a predictive tool to quantify the outcomes of urban planning—such as CO₂ reduction, traffic congestion levels, and parking availability—prior to physical investments.
- Cross-sector integration: Incorporating datasets on mobility, air quality, waste, and energy into DOTSOFT’s NOON IoT platform.
- Technical Comparative Evaluation: Using the dAIEDGE Virtual Lab to measure the energy efficiency, latency, and accuracy of models on real edge devices, such as the Jetson Orin.
- Privacy and Compliance: Ensuring data sovereignty and alignment with the EU Artificial Intelligence Regulation through on-device computing and low-latency inference.
- Pilot validation: Demonstration of the solution in real-world municipal environments, with the aim of ensuring scalability and market readiness.
Funded by the European Union. The views and opinions expressed are those of the authors alone and do not necessarily reflect the views of the European Union or the European Commission. Neither the European Union nor the relevant funding authority can be held responsible for them.
Ημερομηνία:
January 2026
