In the nxtAIM research project, dSPACE is working with 20 partners from industry and research to overcome the limits of general generative AI models and develop a generative AI model for the automotive industry. The project combines the know-how of leading experts, big data from the automotive industry, and the computing power of the supercomputer in Jülich.
In the world of tomorrow's autonomous mobility, people and machines will share the traffic space. They will interact and cooperate. However, the necessary level of cognitive abilities of autonomous road users can only be achieved through the comprehensive use of machine learning methods. In terms of scalability, transferability, and traceability of the developed data-based driving functionality, however, there are still hurdles on the road to higher levels of autonomy. The limits result from the current system architecture and the machine learning methods used.
Generative AI offers an alternative approach and has impressively demonstrated its capabilities in recent applications such as large language models or text-to-image generators. By using generative methods and the resulting development of base models, nxtAIM is attempting to initiate a paradigm shift in system architecture and AI methodology.
Today's system architectures are based on linear, unidirectional information processing along the effect chain, consisting of perception, environment modeling, prediction, and planning as well as subsequent implementation. Generative methods are used to create a feedback channel and information processing is expanded bidirectionally.
Hitarth Bhatt and Sven Burdorf answer a few questions about the project status.
What qualifies dSPACE to be part of the project consortium?
Hitarth Bhatt: We are already the market leader for simulation and validation tools in the automotive sector, and we are well on the way to becoming the market leader for AI in our domain.
We have already developed AI product functions to significantly increase customer benefits. One example is the provision of AI-supported scenario recognition in the tool chain for data-driven development. We are further strengthening our expertise by participating in research projects, for example, in the field of AI data tooling. And the dSPACE sensor vehicle with high-precision, reliable reference sensors lets us collect data that enables the training and development of AI product functions.
What tasks does dSPACE perform in the nxtAIM project?
Sven Burdorf: dSPACE is responsible for the subject of “Domain Transfer - Using Generative AI to Augment Images”. We are also involved in data consolidation. Here, we plan to use data from the dSPACE sensor vehicle to train the basic automotive generative AI model. Another important component of our work is the evaluation of the models developed in the project with regard to the realism and diversity of the data generated.
How do dSPACE customers benefit from this?
Hitarth Bhatt: nxtAIM will provide the automotive industry with a specific basic generative AI model. On the one hand, we bring our know-how in the field of simulation and validation and on the other hand, we want to gain expertise in order to integrate even more generative AI technologies into dSPACE solutions so that we can support our customers more efficiently in their development projects in the future.
Sven Burdorf
AI Expert, coordinates the dSPACE activities in the project
Hitarth Bhatt
Strategic Project Manager, dSPACE