Traditional methods for personalized thermal comfort measurement rely on subjective questionnaires, which are often impractical to collect in real-world contexts due to their intrusiveness, discontinuity, and dependency on user. To overcome these limitations, this study proposes an innovative methodology for the assessment of thermal comfort measuring solely physiological data, thereby eliminating the need for feedback collection. A non-intrusive sensing system comprising a wearable smartwatch, an infrared temperature sensor, and a social robot, has been employed to measure occupants’ skin temperature and heart rate variability under varying indoor thermal conditions. K-means clustering was applied to unlabeled physiological features. A silhouette score of 0.52 has been obtained with physiological data collected during winter season measurement campaign, revealing that it is possible to discern comfort and discomfort sensation based on solely physiological signals. The results demonstrate the feasibility and robustness of a physiology-based, label-free approach for thermal comfort measurement, which provides a scalable alternative to survey-based methods.

Urban heat exposure is one of the most critical and rapidly intensifying impacts of climate change in metropolitan areas. Reliable measurement of outdoor thermal comfort is therefore essential not only for long-term climate-adaptation planning but also for real-time heat-risk management. However, conventional approaches for measuring the Universal Thermal Climate Index (UTCI) rely either on computationally demanding radiative models or on dense in situ monitoring networks, both of which limit scalability and operational deployment at city scale. This study introduces an innovative data-driven framework for urban-scale measurement of mean radiant temperature (t_r) and UTCI using exclusively open-access reanalysis and real-time meteorological data. A supervised Multi-Layer Perceptron (MLP) model is trained to estimate t_r without requiring detailed urban morphology inputs or demanding radiometric instrumentation. The framework achieves high predictive performance (R² = 0.99; RMSE = 1.4 °C for t_r) and computational efficiency suitable for real-time applications. Another key innovation of this study lies in the comfort–vulnerability integration. The estimated outdoor thermal comfort (i.e., UTCI) is combined with socio-demographic vulnerability measurements within a unified Urban Heat Risk Index (UHRI). Applied to the Besòs district in Barcelona, the framework identifies hotspots where thermal stress and social vulnerability converge, supporting urban resilience and adaptation strategies. A Monte Carlo sensitivity and uncertainty-propagation analysis further showed limited error propagation across the framework, with standard uncertainties of 0.88 °C for MLP-based t_r, 0.67 °C for UTCI, and 0.017 for UHRI.

This study proposes a non-invasive experimental setup that integrates environmental and physiological sensors with personal parameters (age, gender, and Body Mass Index (BMI)) to measure the perceived thermal comfort of occupants via a Machine Learning (ML) framework. An experimental campaign involving 23 participants was conducted across cold (17–20 °C), neutral (22–24 °C), and warm conditions (26–28 °C). Data collected combined environmental measurements, continuous wearable physiological signals, subjective Thermal Sensation Votes (TSVs), and a multidimensional Global Sensation Vote (GSV). Evaluating six distinct datasets via a Leave-One-Subject-Out (LOSO) cross-validation strategy revealed that the value of personal data depends heavily on the underlying sensor architecture. While integrating personal data introduced slight noise in purely environmental linear models (decreasing accuracy by 4.77%), it yielded significant quantitative increments as sensor complexity shifted toward the individual, boosting accuracy by 2.03% in physiological setups and a remarkable 13.6% in bare-minimum sensor networks. Overall predictive performance reached an accuracy of 73% using a Support Vector Machine (SVM-LIN) for environmental configurations, while an XGBoost model achieved a robust 68% accuracy in a highly reduced, low-cost sensor setup by leveraging personal context. The innovative aspect of this research is the simultaneous use of sensors to monitor both physiological and environmental parameters, which, when combined with subjective feedback and personal data, significantly increases the prediction of TSV in practical cases. This integrated approach suggests prospective applications for vulnerable categories whose thermoregulatory responses deviate from standardized models, though it requires future empirical validation among these specific demographics. Overall, the integration of personal parameters into data-driven models contributes to generalizing individual physiological differences and supports the development of more effective, human-centric IoT monitoring systems.

This study presents a methodology for measuring indoor thermal comfort assessment using data measured from environmental sensors and fed into a simplified model. The proposed measurement system enables thermal comfort evaluation using a reduced set of directly measurable mechanical and thermal quantities, namely indoor air temperature, outdoor air temperature, and relative humidity, thereby avoiding the need for full microclimatic instrumentation. To this aim, an optimized simplified Predicted Mean Vote model (sPMVopt) is introduced within a model-based measurement framework. Clothing insulation, which is not directly measurable in operational buildings, is estimated dynamically using a Generalized Additive Model (GAM) to capture the non-linear relationship between environmental variables and occupants’ clothing behavior. The resulting indirect clothing insulation estimates are subsequently approximated by third-degree polynomial functions, yielding an analytical formulation suitable for real-time implementation. Based on the distribution of the estimated insulation values, four refined clothing insulation ranges are identified, and for each range climate-specific regression coefficients are calibrated using data from the ASHRAE Global Thermal Comfort Database II, considering three representative European climate zones (France, Portugal, and Sweden). The proposed measurement system is validated against the reference PMV defined by ISO 7730 and tested in a real office environment. Results show that sPMVopt achieves mean absolute error (MAE) and mean squared error (MSE) values of 0.19 and 0.06, respectively, corresponding to improvements of 13% and 17% compared to existing reference methodologies (e.g. PMV). A sensitivity analysis also shows that air temperature is the input variable that causes a higher deviation sPMVopt. Overall, the study demonstrates that reliable thermal comfort measurement can be achieved through a simplified sensing setup and a validated model-based estimation process, enabling scalable and real-time comfort monitoring for building diagnostics, HVAC commissioning, and smart-building applications.