Publicación:

A human-centered cognitive support framework for manual industrial assembly: integrating perception, agentic reasoning, and augmented reality guidance

Resumen

Manual industrial assembly remains essential in high-variety and customized production, but increasing product and process complexity places substantial cognitive demands on operators. Existing assistance systems, particularly augmented reality-based solutions, improve instruction visualization and task guidance, yet they often remain weakly connected to the real assembly state and limited in reasoning, adaptation, and decision support. This paper proposes a human-centered cognitive support framework for manual industrial assembly that integrates perception, agentic reasoning, knowledge grounding, and augmented reality guidance. The study is informed by a systematic search, bibliometric overview, and literature analysis of 129 Scopus-indexed documents. The analysis shows that augmented reality dominates current cognitive support approaches, while AI-based methods are increasingly used for object recognition, contextual interpretation, adaptive information delivery, and error detection. However, perception, reasoning, guidance, and knowledge grounding are still commonly treated as isolated functions. Based on these findings, 11 review-derived design requirements are formulated and used to develop a conceptual perception-cognition-guidance framework that represents cognitive support as a closed human-centered loop grounded in procedures, constraints, rules, and memory. The framework is then translated into a layered implementation architecture comprising physical, perception, cognitive, guidance, knowledge, and application layers. Structured data contracts clarify how sensory and interaction data can be transformed into perception evidence, structured assembly states, cognitive support decisions, and device-specific guidance commands. A toy-train assembly demonstrator illustrates how procedural state modeling, multi-camera perception, YOLO11-based object detection, projector-based guidance, and contract-based data exchange can connect physical assembly events with structured reasoning and operator-facing feedback.