A preliminary study of a co-embodiment approach for welding training

Co-embodiment enables users to share a single virtual avatar with other users, robots, or autonomous agents. Recent studies explored co-embodiment usage in Virtual Reality and proved its potential in training scenarios as motor skill learning and retention can be improved. This paper investigates co-embodiment applied to high-accuracy tasks, which usually require high spatial and temporal accuracy. A virtual reality, multi-user application for welding tasks has been designed and developed. The application supports two different teaching approaches. In the learning-by-imitation approach, the trainer controls his avatar, whereas the trainee can observe the trainer performing the welding procedure and replicate the task by controlling his avatar. In the co-embodiment approach, both trainer and trainee control the same avatar and should work together to complete the task. An experiment has been proposed to compare the usability of the two approaches. Questionnaires and user interviews have been used to obtain subjective measures of the system's usability. Users' evaluation and feedback provided helpful insight into the proposed system and led to the definition of other possible approaches for designing co-embodiment interfaces for teaching high-accuracy tasks. Future research will be aimed at developing and testing these different paradigms.