Abstract
Air pollution is a critical global health issue, affecting nearly the entire human population and contributing to millions of premature deaths annually. One pathway through which pollution impacts human health is via the autonomic nervous system, measurable through HRV, a sensitive marker of cardiovascular regulation. Prior studies have demonstrated that fine PM2.5, CO₂, and VOCs are associated with HRV reduction, but most rely on large cohort data or lack fine temporal resolution. Here, we present a real-time, multimodal sensing platform that combines wearable physiological monitoring with environmental air quality sensors to assess short-term HRV responses under varying ambient conditions. Using our in-house Zhurek IoT device, we synchronized physiological data with environmental parameters across three contrasting urban and natural settings. Machine learning models, especially XGBoost, accurately classified levels of HRV change (low, moderate, high) based on air quality metrics with up to 86.71% accuracy. This confirms that pollutant levels can predict subtle changes in autonomic function even in healthy young adults. These findings extend prior knowledge by demonstrating that short-term fluctuations in air quality can measurably affect HRV, even in the absence of chronic illness. In contrast to earlier research focusing on long-term exposure or clinical populations, our study highlights the vulnerability of healthy individuals to environmental stressors and shows how machine learning enhances the detection of such effects. The results underscore the utility of combining wearable technology and artificial intelligence in environmental health monitoring. They provide a foundation for personalized risk assessment and targeted public health interventions, especially in urban areas with fluctuating pollution levels.