DART: Distributed AI for Robust and Trustworthy Autonomous Vehicles

DART advances distributed AI and federated learning for autonomous vehicles without sharing sensitive data. Through collaboration between RISE, Zenseact, and NVIDIA FLARE, the project develops robust, trustworthy, and privacy-preserving AI solutions that support safer and more sustainable transportation systems.

About the Project

DART (Distributed AI for Robust and Trustworthy Autonomous Vehicles) develops new methods for distributed AI and federated learning for autonomous vehicles. The project enables AI models to be trained across multiple organizations without sharing raw data, strengthening privacy, cybersecurity, and regulatory compliance.

Why is the project needed?

Autonomous vehicles generate large amounts of sensor data that are essential for developing safe and trustworthy AI systems. However, much of this data is sensitive, proprietary, or safety-critical, making centralized data sharing difficult. DART addresses this challenge through distributed AI, allowing models to be trained locally while knowledge is shared across organizations.

Solution and Innovation

The project combines federated learning, self-supervised learning, and knowledge distillation to develop robust and trustworthy AI models for autonomous vehicles. In collaboration with NVIDIA FLARE, the project evaluates advanced technologies for secure, scalable, and privacy-preserving distributed AI training.

Benefits for Society and Industry

By enabling AI development without sharing sensitive data, the project contributes to safer autonomous transportation systems, more efficient use of data resources, and reduced environmental impact from large-scale data transfer. The results are also applicable to sectors such as manufacturing, energy, and smart cities.

Project Partners

The project is carried out by RISE, Zenseact, and NVIDIA FLARE. Together, the partners combine expertise in distributed AI, autonomous driving, and privacy-preserving machine learning