Machine learning-driven psychophysiological analysis in immersive construction environments: a systematic review

Machine learning (ML)-augmented psychophysiological monitoring (PPM) in immersive construction environments (ICEs) holds significant promise for assessing workers’ cognitive and behavioral states. However, the field lacks a comprehensive synthesis to guide its development. This study conducted a systematic review of 20 articles to bridge this gap and map the current landscape. The analysis identified a growing adoption of diverse PPM modalities (e.g. EEG, fNIRS, EDA) integrated with ML algorithms (e.g. SVM, deep learning) in VR/AR simulations. Thematic findings reveal four core research themes: (1) cognitive state prediction (e.g. fatigue, mental workload), (2) real-time hazard identification, (3) training performance evaluation, and (4) occupational health risk assessment. By synthesizing these findings into a structured framework, this review provides scholars and practitioners with a clear understanding of the state-of-the-art and a strategic agenda for future work. The review culminates in a research roadmap that outlines critical theoretical contributions and practical implications. This work provides a foundational framework for future research, highlighting both the potential and the challenges—such as experimental validity and model interpretability—that must be addressed to advance safety and productivity in construction through ML-driven PPM.