Development and innovation

A central focus of the AI Innovation Center is direct cooperation with industrial companies: Feasibility studies (Quick Checks) and projects to develop initial prototypes (Exploring Projects) of AI applications. In addition, consortium projects and innovation networks are aimed at cross-company exchange.

Machine learning with embedded devices

AI accelerator chips enable lower latencies and higher throughput rates for industrial applications. An analysis of the Hailo-8 AI chip shows great potential for increasing efficiency in machine control.

NeuraConnect: Outsourcing compute-intensive operations to a 5G edge cloud

Outsourcing of compute-intensive operations to a 5G edge cloud.

Packing planning shipping carton

Automatic model-free specification of packing patterns for shipping cartons.
Automation of the packaging process for various objects.

Peg-Men: Flexible assembly under disturbance variables

Flexible assembly of sensitive valve needles is optimized using deep reinforcement learning and physics simulation. The aim is to automate repetitive processes in order to increase efficiency and precision.

Damage evaluation of rockfalls with machine learning

The semi-automated labeling tool can identify objects in images and draw bounding boxes around the defective areas using Segment Anything. Although the dataset is limited, the object detection model has shown promising results and is expected to be further improved with a larger dataset.

Transfer Learning Basic Network Examination

The aim is to investigate transfer learning and optical character recognition to improve the recognition of small objects, adapting existing models to new use cases.