Spatial Augmented Reality in Manual Assembly: An Empirical Investigation of Its Effects on Assembly Performance and Cognitive Ergonomics

Bugra Alkan, Naimul Hasan, Louie Webb, Malarvizhi Kaniappan Chinnathai

October 2025

Abstract

Manual assembly systems are becoming increasingly complex due to mass personalisation, often leading to higher risks of assembly and picking errors, reduced quality, and greater cognitive demands on operators. Spatial Augmented Reality (SAR) systems provide hands-free, real-time guidance projected directly onto the workspace, and may overcome limitations of wearable or handheld AR devices, particularly in static assembly settings. This study aimed to evaluate the effectiveness of a SAR system prototype in supporting component picking and assembly tasks of varying complexity compared with traditional paper-based work instructions. We evaluated the SAR prototype against paper instructions in a within-subjects study (𝑁 = 16) using simple, medium, and hard molecular assemblies, and examined the effect of time pressure. Performancemetricswere task time, assembly errors, and picking errors; subjective workload (NASA-TLX) and user feedback were collected, and EEG frontal alpha power indexed cognitive load. Under untimed conditions, SAR reduced task time by a median 14.3% (paired Wilcoxon, 𝑊 = 573.0, 𝑝 = .0018) and lowered errors (assembly 33.0%, 𝑊 = 106.5, 𝑝 = .028; picking 30.9%, 𝑊 = 313.5, 𝑝 = .0047) versus paper. Effort and Frustration were significantly lower with SAR (𝑝 = .0006 and 𝑝 = .020), whereas Mental and Temporal Demand trended lower but were not significant (𝑝 = .066 and 𝑝 = .122). Increasing task complexity elevated workload and errors (Friedman, 𝑝 < .05 on most measures). Moreover, time pressure increased mental/temporal demand and effort and reduced perceived performance (all 𝑝 < .05) while SAR provided modest improvements. EEG analysis indicated a modest suppression of frontal alpha power with increasing task complexity and time pressure, consistent with higher cognitive load, while differences between instruction modalities were limited. User feedback indicated high satisfaction and a strong preference for SAR. Overall, SAR improved user experience and key performance outcomes in manual assembly, supporting its potential as an adaptive, human-centric aid for Industry 5.0 settings.

Type

Preprint