Embedded AI

The research focus “Embedded AI” deals with the use of artificial intelligence in the design of (embedded) electronic systems as well as the design of intelligent electronic systems, in particular autonomous systems. However, lightweight implementations of such embedded autonomous systems present researchers and developers with major challenges that have not yet been adequately solved in terms of data volumes, storage and processing performance, as well as the correctness, safety and security of such intelligent systems.

Due to the high costs, size and relatively high energy consumption, known techniques for implementing machine learning algorithms cannot therefore be used in everyday objects (Internet of Things, IoT), e.g. an intelligent rolling bearing, an adaptive valve or a hearing aid that adapts itself to the wearer. It is important to break new ground here so that machine learning is also possible on small, embedded systems. In the area of analysis and verification, there is also a considerable need for interdisciplinary research to investigate the role and integration of machine learning methods in established methods of signal processing, control engineering and system design. Another branch of research is aimed at assuring provable quality criteria for properties such as robustness, fault tolerance, safety and real-time of learning systems that cannot be statically verified. Guaranteeing the privacy of data and models is also a major challenge for numerous fields of application. In design automation, there are also fundamental questions regarding the support and integration of machine learning (ML) and symbolic AI methods, especially for IoT devices.