Human thermoregulation is a complex process involving multiple physiological systems that work together to maintain core temperature through mechanisms like sweating, shivering, and vasodilation. This dynamic balance reflects interactions between the body and its environment, emphasizing that thermal experience emerges across biological and ecological scales. Yet, thermoregulation has limits. Factors such as heat, humidity, and poor ventilation can overwhelm the body’s cooling mechanisms, posing risks especially for vulnerable populations. As climate change alters microclimates and ecosystems, these shifts challenge the body’s adaptive capacity and highlight the need for responsive environmental design.

Real-time physiological and environmental data collected through a custom wearable device at the Arnold Arboretum in Boston were used to model human–microclimate interactions. The system recorded biometric signals, including heart rate, skin temperature, GSR, and blood oxygen, alongside environmental variables like temperature, humidity, CO₂, and light intensity under varying forest canopy conditions. Machine learning models trained on this dataset revealed how subtle microclimatic shifts, such as changes in canopy density or humidity, correlate with physiological stress markers. This data-driven framework enables prediction of comfort states and supports the design of adaptive systems that respond dynamically to both human and environmental conditions.

This study develops a data-driven framework for optimizing and personalizing outdoor comfort through predictive modeling and adaptive decision-making. Using physiological and environmental data, machine learning models forecast future microclimate conditions and human physiological responses. Gradient boosting and LSTM models predict localized temperature, humidity, and light variations. Subsequent models estimate how these microclimates influence physiological stress markers, revealing greater comfort and resilience in canopy-shaded areas. Building on these predictions, classification and regression models recommend real-time adaptive actions, such as resting or hydrating, and optimize route selection to minimize stress and maximize comfort. Through multimodal data fusion and integrated system design, the framework demonstrates how AI can enable personalized, context-aware guidance for physiological well-being and environmental adaptation.

Using supervised machine learning, the system classifies real-time physiological and environmental data into comfort states and delivers personalized feedback to guide thermoregulatory actions like resting, hydrating, or seeking shade. Beyond monitoring, unsupervised models integrate biometric, spatial, and environmental data to generate adaptive route recommendations that anticipate future comfort conditions. Field tests showed that ML optimized routes improved physiological stability and reduced stress indicators. Future work aims to expand datasets and refine predictive accuracy for more robust, personalized environmental adaptation.