DART: Distribuerad AI för Robusta och Tillförlitliga autonoma fordon

Purpose and goal

The project aims to establish a Sweden–US collaboration on distributed and privacy-preserving AI for autonomous systems. Its goal is to enable robust and trustworthy AI model development across multiple training nodes without sharing sensitive or proprietary data. By combining Sweden’s industrial experience with US advanced distributed-AI technology, the project lays the foundation for secure, large-scale AI innovation in safety-critical domains.

Expected effects and result

The project will deliver a jointly defined technical concept for distributed AI and federated learning in an autonomous vehicle use case. Expected effects include improved model robustness through learning from diverse data sources, stronger privacy and data protection, and a clear roadmap for long-term Sweden–US collaboration. The results will also be transferable to other sectors such as manufacturing, energy, and smart cities.

Planned approach and implementation

The project is implemented through joint technical preparation, on-site discussion in the United States, and structured follow-up activities. Swedish partners prepare data, models, and baseline workflows, which are jointly evaluated and refined with the US partner. The collaboration focuses on experiment design, privacy and security mechanisms, and documentation, resulting in a validated concept and a roadmap for further implementation.