In Key Project 1: Design and Evaluation of Ergonomic Hybrid Work Systems Using Interpretable AI Models, using the development of an assembly workstation in the electronics industry as an example, a camera-based assistance system is being developed that captures and analyses body postures directly at the workstation. 

The assistance system developed uses standard webcams and edge AI to record body postures anonymously in real time. A neural network is used to extract joint points from the 2D camera images. The 3D skeletal data reconstructed from this forms the basis for an ergonomic assessment using transparent, rule-based methods (the RULA method). Based on this data, the system identifies ergonomically unfavourable postures and provides employees with direct feedback via a monitor. This enables incorrect postural strain to be identified and corrected at an early stage. 

This enables a transparent analysis of posture whilst at the same time protecting employees’ privacy.