Improving Ergonomic Training Using Augmented Reality Feedback

Abstract

Work-related musculoskeletal disorders are costly to economies and society, and are the most common work-related health problems. In the European Union and the United States, 30–40% of workers reported work-related musculoskeletal disorders at least in the last decade. Effective ergonomics interventions can be difficult and expensive to implement, leaving administrative measures such as rest breaks, job rotation, and ergonomic training as the main interventions. However, traditional ergonomic training often lacks effectiveness due to design flaws. With advancements in technology, motion capture and pose estimation technologies based on Augmented Reality and Virtual Reality have been proposed as solutions, but these still have limitations such as they are not participative and are not conducted in the actual workplace. This study aimed to evaluate the effectiveness and usability of an AR pose estimation prototype intended for use as real-time visual feedback in ergonomic training at the workplace. Mediapipe Pose, a computer vision solution from Google, was used to project holograms on the bodies of participants based on a custom Rapid Upper Limb Assessment calculation on shoulders, elbows, neck and back. The participants performed a simulated task and were randomly assigned to one of two groups for practical training. AR Group (n = 5) received AR pose estimation ET with a big screen to provide visual feedback to detect postural exposures and improve technique. The Control group (n = 4) received feedback from an ergonomics specialist. The effectiveness of each ET intervention was measured using a custom time-weighted Rapid Upper Limb Assessment-based score showing positive results from both interventions, the AR pose intervention and the Control Group. The usability of the AR prototype was evaluated using the Post-study System Usability Questionnaire and Bipolar Laddering which resulted in both positive and negative points which are going to be used as input for future improvement of the AR prototype. The results of the study showed promising results in terms of the AR pose estimation prototype being effective in transferring knowledge into behavior. Overall, the AR group showed a greater learning effect than the control group. However, these results have to be taken with caution in relation to the number of participants. The study may contribute to enhancing the outcomes of ET and expanding the field of study on its potential impact on health and safety culture in organizations. The improved prototype will be tested in a real occupational context to further assess its effectiveness.