BMBF’s Energy-efficient AI System competition

© © MEV Verlag
The research field of artificial intelligence aims to imitate human perception and human action by machines.

In order for artificial intelligence (AI) to create value and also to find its way into mobile and safety-critical applications, work must be done to improve energy efficiency. This is because the energy consumption required to run today’s AI systems has so far been too high for many applications. The innovation competition Energy-efficient AI System organized by Germany’s Federal Ministry of Education and Research (BMBF) aims to tackle this problem and supports universities and research institutions in developing solutions.

The winners of the pilot Energy-efficient AI System competition can look forward to follow-up projects in which they can further develop their ideas. With the ADELIA and LO3-ML projects, Fraunhofer IIS and Fraunhofer IPMS are also part of the competition.

ADELIA project:

Experts from Fraunhofer IIS and Fraunhofer IPMS have joined forces across institutes to compete in the competition category FDSOI Technology (22FDX) from GLOBALFOUNDRIES. Their activities are coordinated by Fraunhofer IIS. The proposed approach is based on an energy-efficient crossbar analog accelerator implementation. The potential applications range from energy-efficient ASICs for human-machine communication to disease prediction with medical wearables.

LO3-ML project:

The energy efficiency of the chip is set to be increased both through new technologies and through innovative computer architecture. To realize this, Friedrich-Alexander University Erlangen-Nürnberg (FAU) and Fraunhofer IIS are joining forces in a second project. The project will further expand collaboration between FAU and Fraunhofer IIS in the areas of analog and digital circuit technology and system design.

As an innovation driver in the field of textile-integrated smart sensors and algorithms, Fraunhofer IIS, together with industrial partners, will integrate the results of the projects into medical and non-medical applications.

Last modified: