UA-FlexKI

Study on co-optimization and deployment of neural networks

01/06/2023 to 31/08/2025

The subject of this project is the support Mercedes-Benz AG in the BMBF project MANNHEIM-FlexKI by preparing a study within the scope of a subcontract.

The project MANNHEIM-FlexKI (Flexible AI Deployment and AI Platforms for Embedded, Automotive Applications) aims at breaking the HW dependency of AI applications and to explore an open reference architecture for the deployment of AI and DSP applications, which allows to port these AI applications easily to a new HW platform (so-called retargeting). The FlexKI retargeting process provide significant economic advantages for German companies to deal with future challenges in the development of AI systems.

FlexKI pursues two development paths here: 1) The flexible deployment of networked AI applications on high-performance, heterogeneous commercial-off-the-shelf hardware platforms (COTS-HW platform). 2) A hardware/software co-design approach for deploying to a new customized, energy-efficient AI hardware platform. Both paths avoid vendor lock-in and enable fast and automatic migration of AI applications to other hardware platforms in case specific hardware components are not available due to protectionism, blacklisting or supply shortages (chip crisis). In summary, the project strengthens all levels of the automotive value chain in Germany and establishes a path for future standardization in the field of AI deployment.