Introduction to the Problem
As part of the ongoing industrial digitalisation, augmented reality (AR) technology offers unique potential for enhancing industrial operations by providing a real-time blend of digital and physical worlds. The AR allows for superimposing interactive information onto the physical world, merging digital content with the tangible environment. As a consequence, AR can be used to effectively guide workers through intricate tasks, furnish step-by-step instructions, and accentuate potential hazards, particularly within the context of manual assembly. For instance, AR can seamlessly project assembly instructions onto or adjacent to the workpiece, thereby mitigating the necessity for paper manuals and minimising the likelihood of human error during the assembly process. Furthermore, the AR interfaces can be enhanced with computer vision algorithms to provide real-time verification of the assembly progress. However, the current state of AR technology remains relatively immature, necessitating in-depth studies to fully harness its potential and address the challenges associated with its implementation in industrial settings.
Solution
In a typical manual assembly guidance context, the AR system actively updates the displayed visual cues to align with the current stage of the process or any identified issues. A wide range of visualisation options are available for AR-based manual task guidance. Drawing upon existing research, we categorised these visualisations into four primary groups and conducted a crowdsourcing experiment to evaluate their effectiveness in a pilot study.
Initially, we examined assembly rework tasks that receive AR guidance upon identification of an assembly error. Given the diversity of potential visual guide types that can be utilised in AR guidance, this particular use case scenario is well-suited for the crowdsourcing approach, as the traditional in-lab experimental approach could be a lengthy and resource-demanding process.
Furthermore, to ascertain the practical efficacy of AR-guided assembly training in authentic industrial environments, we conducted field studies with manufacturing apprentices tasked with performing manual assembly of engineering components in distinguished controlled and uncontrolled environments. To achieve that, we designed, developed, and deployed tailored AR-guided manual assembly systems that were adaptable to different levels of asset complexity. Our results suggest a substantial increase in task performance within the industrial workshop, characterised by faster turnaround times, reduced error rates, and a subjective improvement in the overall work experience. These counterintuitive findings highlight the need to evaluate AR-based assistance systems under real-world conditions to obtain accurate information about their effectiveness and usability.