Work in Process - motion capture of work activities

• Applicants: Professor Dr.-Ing. Martin Schmauder; Professor Dr. Markus Wacker
• Subject Area: Human Factors, Ergonomics, Human-Machine Systems
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 542857478

Human postures and body movements are decisive characteristics in the assessment and design of physical workplaces. By applying the findings of ergonomics, occupational safety and process design, the strain on employees can be optimized and their health and performance can be preserved in the long term. Ergonomic digital human models offer a helpful tool for this, as they can be used to simulate and analyze health risks and physical stresses during a virtual planning of work systems. Predefined conditions and formal rules are usually implemented for this purpose. However, this approach cannot adequately model the individuality of people and the complexity of work performance. As a result, digital human models often lack in validity and sometimes even usability. Machine learning algorithms, on the other hand, make it possible to integrate knowledge-based models into software. This makes it possible to exploit the potential of digital human models to a greater extent, as complex ergonomics procedures can also be integrated, which, for example, require empirical knowledge for application. Modeling implicit knowledge using machine learning requires a suitable data set that can be used as the basis for training. Since the content of the training data determines the generalization capability of the models, there is a need to capture a relevant spectrum of physical stress types performed by a subject collective with comprehensive anthropometric diversity. However, currently available data collections with motion recordings of work activities only provide part of the necessary scope. There is a need to record the missing data and combine it with the available data in a standardized format. The data set must then be supplemented with ergonomic expertise on work-related postures in an annotation process. The model learned with the training data can be integrated into digital human models for concrete application, enabling automated classification of postures in accordance with the Key Indicator Method (KIM) of the Federal Institute for Occupational Safety and Health. The data set created can also be used in future as a reference data set for the development of machine learning methods in the context of occupational science research. The planned project will thus close a significant gap in the development of a new generation of software tools for digital ergonomics.

• DFG Programme: Research Grants