Microelectronics for AI - Neuromorphic Hardware

In the joint project "Microelectronics for AI - Neuromorphic Hardware", three business-related research institutes of the Innovationsallianz Baden-Württemberg are jointly researching and developing adaptive as well as secure and energy-efficient AI chips, which are considered to be an elementary building block for Industry 4.0 and the Internet of Things. The project is being implemented under the coordination of the Institut für Mikroelektronik Stuttgart (IMS CHIPS) with the participation of the Research Center for Information Technology (FZI) and Hahn-Schickard. An industry advisory board will accompany the project.

For Industry 4.0 and the Internet of Things to have an impact, it is crucial that data cannot only be processed in the cloud as Big Data. Rather, for many applications it is necessary that intelligent data acquisition, signal processing and data reduction already take place in the local microsystem, e.g. at the sensor interface. The research project therefore aims to develop practical and secure chip solutions for resource-efficient on-site processing of data (so-called edge computing) using neuromorphic AI chips. Due to their special chip architecture, these AI chips are capable of learning. The use of certain AI hardware can drastically reduce energy consumption, which is of enormous importance given the large number of such intelligent microsystems in autonomous systems.

The researchers are taking security-relevant aspects into account both in the chip architecture and in the technology routes that such a chip must pass through during its manufacture after the design stage. At the end of the project, a chip architecture for the neuromorphic processor flanked by specially developed solutions for adapting sensors and with secure communication interfaces will be available to demonstrate its performance. In a demonstrator, the functional verification of an industrial product example is to be provided using the application example of intelligent condition monitoring. On the one hand, this allows intelligent automated final inspection during production, and on the other hand, it allows continuous self-monitoring during operation. Assemblies equipped with "intelligence" in this way are the basis for autonomous and networked applications that are increasingly being used in Industry 4.0 and the Internet of Things.

Expressions of interest from well-known small and medium-sized companies from industry for the project advisory board have already been submitted to the research institutions. Companies also have the opportunity to join the industry advisory board during the project period. In addition, networking with other AI-relevant research initiatives and networks in the state is being sought.