AI-Disco - Edge-Cloud AI for Distributed Sensing and Data Processing

Fraunhofer IPMS is contributing its expertise in intelligent, AI-based sensor technology, and neuromorphic hardware to the AI-DISCO joint project. The aim is to develop energy-efficient, AI-enabled hardware components and practical demonstrators for industry, agriculture, and medicine, with a strong focus on Cottbus and the Lusatia region. In the AI-DISCO project, Fraunhofer IPMS is thus making a significant contribution to transferring AI-enabled electronics and intelligent sensor technology from research into market-ready applications – while at the same time strengthening Lusatia as a high-tech region for AI and microelectronics.

Intelligent sensor systems and edge integration

In addition to processor development, Fraunhofer IPMS is responsible for key work on intelligent sensor systems that are directly linked to the AI‑DISCO platform:

• Interfaces between sensor technology and edge AI: Design and implementation of hardware and software interfaces that feed sensor data into AI processing with low latency and high robustness.
• Integration in edge nodes: Adaptation of sensor systems to reconfigurable edge hardware (AI-DISCO node) and preparation for use in real-world applications.

Demonstrators for agriculture and medical technology

To demonstrate the practical applicability of the new technologies, Fraunhofer IPMS is developing two exemplary demonstrators:

1. Agriculture: Mobile NIR spectrometer with edge AI

• Miniaturization of a near-infrared spectrometer based on MEMS technology
• Development of a handy, battery-powered measuring system for determining the ripeness of fruit directly in the field
• Collection of extensive training data and development of AI-based evaluation algorithms on the AI-DISCO-Edge platform
• Field tests and optimization in a real agricultural environment

2. Medical technology: Ultrasound eye tracking in glasses

• Integration of CMUT ultrasonic sensors and control electronics into a glasses-based system
• Real-time recording and evaluation of eye movements (eye tracking)
• Development of AI models for robust gaze direction and saccade detection
• Preparation of applications, e.g., for contactless human-machine interfaces and neurodiagnostic issues

Neuromorphic AI processors and in-memory computing

The core task of Fraunhofer IPMS is to research and implement novel neuromorphic hardware architectures in the work package “Advanced Neuromorphic AI Processor”:

• Design of FeRAM-based in-memory computing architectures: Development of mixed-signal computing cores that process data directly in memory, thereby combining computing and memory accesses
• Physical design in 22 nm technology: Physical implementation and post-layout simulations of FeRAM-based accelerators in modern FDSOI technology.
• Characterization and optimization: Setting up test platforms and measurement systems to evaluate and improve the speed, energy consumption, and reliability of the new components..

The IPMS thus creates the hardware basis for particularly energy-efficient edge AI computing and neuromorphic computing concepts.