Project ANDANTE: Analog Accelerator for Inference at the Edge

Edge computing is considered a key for new IoT applications. In order to bring artificial intelligence into future edge products, Fraunhofer EMFT researchers are working together with Fraunhofer IIS and Fraunhofer IPMS as part of the EU ANDANTE project to develop innovative mixed-signal artificial neural network (ANN) accelerator with computation-in-memory (CIM) ability. These are intended to enable the construction of solid hardware and software platforms for AI application development. 

The resulting IoT devices are expected to combine excellent energy efficiency with robust neuromorphic computing capabilities. Through an efficient collaboration between major European manufacturing facilities, chip design companies, system houses, application development companies and research partners, the project will build and expand the European ecosystem around the definition, development, manufacturing and application of neuromorphic integrated circuits.

Fraunhofer EMFT is contributing its studies on AI devices, methods and tools for creating a flexible yet efficient mixed-signal ANN circuit architecture to the project work. In addition, the Munich researchers are developing tools that support resource-conscious planning of an ANN model for the available hardware in the edge products. The focus is on the aspects of computational accuracy, data throughput and performance trade-offs. Furthermore, various circuit blocks for neural networks of will be developed at EMFT, with a focus on high configurability and low power consumption.

The overall project is funded within the European ECSEL initiative under the grant number 876925, the sub-project is additionally funded by the German Federal Ministry of Education and Research BMBF under the grant number 16MEE0117.