AUTOCRAFT: Automated Craftsmanship Skill Transfer

In the current manufacturing industry, many tasks are still handled by human operators due to the craftsmanship skill required for the job. There is a huge shortage of such skilled workers in the Flemish industry. Skilled operators execute tasks fast, efficiently, safely, ergonomically, and with high quality. With growing flexibility demands for production environments, operators need to switch more often between different tasks, resulting in a lower experience and skill in the task and, thus limited output quality and speed. In this project, we aim to extend the understanding of craftsmanship beyond (but not excluding) the explicit knowledge available from engineering documents and obtained from interviews. This additional data will be captured from skilled workers through multi-modal sensing, the information or know-how embedded in motion trajectories, contact forces, body postures, properties of materials and workpieces, process parameter settings, etc. If available, this knowledge will be obtained from different skilled workers to measure their variability while executing tasks. This data will be used to generate a “skill model” and an effective automated skill transfer system with personalized multimodal feedback based on skill levels, e.g., trajectory guidance and correction using Virtual Reality/Augmented Reality (VR/AR), auditive feedback, and tactile feedback. This skill transfer software can also be used to assess the skill level or to measure skill progress by the trainee. The AUTOCRAFT craftsmanship training platform will help to create a more skilled workforce, relax the need for skills of new hires and help to fill in vacancies, decrease the need for available expert trainers, and result in a more flexible workforce. The resulting frameworks of this project will describe a craftsmanship skill in an unambiguous, standardized, and quantified manner, capture a skill, assess an individual’s skill level, model a skill, and transfer a skill. These reusable results will be demonstrated and validated by a lab use case, benchmarked in a real training environment, and demonstrated in three use cases proposed by participating end users (Pfizer, Atlas Copco, Daikin).