After one year, GAIA-X 4 KI is showing initial results with demonstrators and the connection to Gaia-X for AI applications in the automotive sector.
Two minimal examples ("CARLA" and "KI Training Pipeline") explore the foundations for Gaia-X-based AI infrastructures.
Further development of architectures, legal framework conditions and integration in Gaia-X.
After just one year, the GAIA-X 4 KI project has already achieved initial results in the development of distributed system structures for artificial intelligence (AI) applications in the automotive sector. These include the first demonstrators, which the German Aerospace Center (DLR) can use together with the 13 participating companies and research institutions to show the application of the open IT infrastructure Gaia-X. The project team has also identified requirements and components for further project work, such as the definition of a legal and regulatory framework for handling large amounts of data and automated driving functions in the context of Gaia-X. Furthermore, the researchers have designed interfaces and architectures to ensure the connection of the developed AI applications to the European Gaia-X data infrastructure.
Project coordinator Dr Sascha Knake-Langhorst from the DLR Institute of Transportation Systems summarises: "After one year of the project, we now have a better understanding of how Gaia-X can advance automated driving and manufacturing processes. The project has thus laid the foundation for achieving its goal over the next two years: building a data and services ecosystem for training and validating AI for automated driving and vehicle manufacturing."
Last year, the GAIA-X 4 KI project developed two so-called minimal examples in the fields of production and automated connected driving. Minimal examples have a reduced level of complexity compared to the subsequent use cases based on them. They serve to develop a basic understanding of the requirements, processes and methods that a Gaia-X-based AI data infrastructure needs.
The minimal example "CARLA" aims to connect a vehicle automation framework to the open source simulation framework CARLA, which virtually visualises application scenarios for automation development. This enables the researchers to link large amounts of data with complex software services in order to demonstrate cross-provider services with Gaia-X.
In the second minimal example called "KI Training Pipeline", a process chain for training AI functions is being set up and integrated into cloud systems. Once completed, this service can be used across all applications in the project and thus offers a very good example of research into the requirements for an open data and service infrastructure, as made possible by Gaia-X.
"With the help of the two Minimal Examples, we were able to successfully develop the foundations for handling data and services in a Gaia-X-based ecosystem and already implement the first prototype applications," summarises Knake-Langhorst. Through the targeted selection and integration of further Gaia-X core components, the development of which is currently being driven forward across Europe, the project consortium will be able to add further functions to Minimal Examples over the next two years and link them to the Data Space. The researchers will utilise the knowledge acquired for the use cases in the project, which will be further developed in parallel with the activities described above.
In addition to the collaborative work on the Minimal Examples, the GAIA-X 4 AI team researched and developed the technical foundations. This included the conceptualisation and development of infrastructure solutions for the implementation of Gaia-X-compliant functions and services. As part of this, the project participants set up a cloud system, for example, on which the minimal examples described above can be implemented and tested. In addition, the researchers are currently developing a Gaia-X-compliant data space, which will form the basis for the various application examples in the project.
The consortium is also involved in the development of necessary technical Gaia-X main components beyond project boundaries. Examples of this are the Gaia-X Federation Services and the Eclipse Dataspace connector. To this end, the project participants are actively involved in the relevant initiatives and contribute the requirements identified in the project directly to the development activities.
Both the collection of big data by AI applications and the development of automated driving functions as part of the Gaia-X ecosystem entail special legal requirements. The GAIA-X 4 KI project participants are therefore analysing the legal and regulatory issues arising from the project and developing suitable concepts.
Based on the planned use cases and software developments, the researchers have already identified specific legal requirements and created IT security and cyber security concepts. The knowledge gained will serve as a guide for the project participants in the development, concrete implementation and realisation of the demonstrators in the data ecosystem. In addition, the project aims to provide an overview of the basic requirements of the aforementioned legal areas.
The GAIA-X 4 KI project was launched in June 2021, making it the first of the GAIA-X 4 Future Mobility project family in the mobility domain. 14 companies and research institutions are working together on the project. It will run until mid-2024, has a budget of 18 million euros and is funded by the Federal Ministry of Economic Affairs and Climate Action (BMWK)
