Wearable and Low-Cost IoT Air Quality Sensing with AI and XAI: A Review of Computational Framework for Personalized Exposure Assessment

Air pollution monitoring has undergone a significant transformation with the emergence of wearable sensing technologies, low-cost IoT monitoring networks, and AI–driven analytical frameworks. Conventional regulatory monitoring systems, although highly accurate, are spatially sparse and therefore insufficient for capturing the dynamic and heterogeneous nature of individual exposure across diverse microenvironments. Recent advances in distributed sensing and data-driven modeling have enabled a transition from ambient concentration monitoring toward personalized exposure assessment. This review presents a comprehensive synthesis of computational methods that support this transition by integrating wearable and low-cost IoT sensing platforms with ML, DL, and XAI techniques. The study first examines the sensing principles, architectures, and deployment strategies of portable air quality monitoring devices and distributed sensor networks. It then systematically reviews AI-based approaches for environmental data analytics, including classical ML, DNN, transformer-based models, and hybrid ensemble frameworks for pollutant prediction and data fusion. Particular emphasis is placed on computational calibration techniques addressing measurement uncertainty in low-cost sensors, including transfer learning, drift detection, and adaptive recalibration strategies. The review further analyzes the role of XAI methods in enhancing transparency, interpretability, and trustworthiness of AI-enabled environmental monitoring systems. Finally, emerging computational frameworks for personal exposure estimation are discussed, including mobility-aware modeling, multi-source data fusion, and satellite-assisted exposure reconstruction. By synthesizing sensing technologies, computational modeling strategies, and explainability mechanisms within a unified analytical pipeline, this review identifies key research challenges and outlines future directions for scalable, reliable, and interpretable personalized air pollution monitoring systems.