KI embedded KI-Grundlagenentwicklung für Embedded-Systems mit Leitanwendungen in Virtueller Sensorik und Brennstoffzellenregelung


The overall goal of the project is a novel method for the development, modeling and control of powertrain control  systems based on AI technology. In this context, OFFIS is developing a compiler for novel data flow processor architectures.

For these processors, a compiler should translate control algorithms from a mathematical-algorithmic domain-specific language (DSL) into a configuration for this processor, which corresponds to a program for classical processors. In contrast to other processors, the data flow architecture does not process sequential instructions, but can process data streams in a highly parallelized manner. As a result, completely new data flow-based optimization methods can be explored.

The technologies used here focus on compiler design and high level synthesis. Application areas are artificial intelligence and conventional signal processing algorithms.


External Leader

Robert Bosch GmbH, Bosch Center for Artificial Intelligence

Scientific Director

A Flexible Graph Language for a Model-Based Semi-Automatic CGRA Compilation Flow

Böseler, Felix and Walter, Jörg; Forum on specification & Design Languages (FDL) 2023; 009 / 2023

Enabling Flexible Model-Based Manually Controllable Compilation for CGRAs (PhD Forum - unpublished)

Böseler, Felix; Forum on specification & Design Languages (FDL) 2023; 009 / 2023

A Rate-Parametric Dataflow Language for a Manual Controllable CGRA Compilation Flow

Böseler, Felix and Walter, Jörg; Methoden und Beschreibungssprachen zur Modellierung und Verifikation von Schaltungen und Systemen (MBMV) 2024; 002 / 2024

Itemis AG
PLS Programmierbare Logik & Systeme GmbH
Robert Bosch AG, Abteilung: Powertrain Solutions - Engineering Microcontrollers and Memories (PS-EC/EHM6)
Fraunhofer ISE, Abteilung: Brennstoffzellensysteme
Otto von Guericke Universität Magdeburg, Institut für Automatisierungstechnik


Start: 01.09.2021
End: 31.08.2024

Source of funding

Related projects


An energy-efficient AI network of elementary lookup tables