Conceptualisation of a multimodal, non-intrusive, generative AI-based assistive system for assembly

Abstract

The transition to Industry 5.0 highlights the necessity for human-centric and adaptive manufacturing systems. This study conceptualises a multimodal, generative AI-based assistive system for assembly designed to deliver real-time error detection and adaptive guidance tailored to diverse operator profiles. The system improves human-machine interaction by issuing preventive warnings to the operator prior to critical tasks, detecting assembly errors, providing multimodal corrective instructions during operations, and deploying robotic interventions when operator-driven corrections prove inadequate. Preliminary laboratory-scale implementation results show the system capability in mitigating assembly errors through dynamic assistive technology selection and iterative feedback learning.