Integrating physiological markers and environmental factors for thermal comfort in moving vehicles

Abstract

This study develops a predictive model for passenger thermal comfort in moving vehicles by integrating physiological signals and environmental parameters. The experiments were conducted inside a moving vehicle under real road conditions to ensure the accuracy and applicability of the findings. A total of 60 field experiments were conducted during summer with 30 male participants (aged 19–39) under three cooling scenarios, combining ventilated seats and air conditioning. The non-uniform exposure conditions included upper air temperature variations from 23 °C to 33 °C and solar radiation fluctuations up to 200 W/m². Physiological signals, such as facial and wrist skin temperatures and heart rate variability (HRV), were continuously recorded. The analysis identified nose skin temperature (r = 0.632, p < .001), High(1000 mm above the vehicle floor) air temperature, and body fat percentage as key predictors of thermal preference. A machine learning model was trained using decision tree-based algorithms (Random Forest, XGBoost, CatBoost, and LGBM), achieving 90 % accuracy in predicting passenger thermal preference. The model's explainability, assessed using SHAP values, confirmed the dominance of nose skin temperature and upper air temperature in influencing thermal perception. By integrating real-time physiological data with adaptive climate control, this study provides a data-driven approach to enhancing passenger comfort. Future research should expand demographic diversity, seasonal testing, and modeling of additional physiological markers to improve applicability.