This systematic review evaluates non-intrusive artificial intelligence-based occupant monitoring for Occupant-Centric Control systems, analyzing 82 studies published between 2015 and March 2026 across thermal, indoor air quality, and visual domains. The analysis reveals that deep learning-based computer vision represents the predominant approach, achieving accuracies of 80–100% for physiological parameter estimation under controlled laboratory conditions, and up to 99.3% for occupant detection in field settings. Integrating these technologies into building operations facilitates substantial benefits: energy savings of up to 50% and thermal comfort improvements of 43–73% have been reported in individual experimental studies, though typical field-validated ranges are more modest. Despite these advancements, significant barriers persist, including high implementation costs, privacy concerns, and a persistent scarcity of labeled training data. This paper establishes a comprehensive technical framework for developing responsive, occupant-centric environments that balance human well-being with operational efficiency.
Yun et al. (Wed,) studied this question.