The research project nxtAIM – NXT GEN AI METHODS develops generative AI methods for the next generation of autonomous driving functions and systems. It focuses on foundation models and world models that represent traffic situations and their evolution over time, while generating realistic sensor data and scenarios. These models provide a basis for simulation, training, and validation, enabling targeted investigation of rare and safety-critical situations.
Another key focus is federated learning of generative models using distributed datasets. This approach aims to enable joint training on data from multiple partners while keeping the data with the respective partners. The project’s objective is to develop generative AI, world models, and foundation models to support the development of autonomous vehicles.