Wang, Jiawei; Olugbade, Temitayo; Tom, Davison; Berthouze, Nadia; (2025) Detecting Car Passenger Comfort Levels through Physiological Sensing and Machine Learning. In: Proceedings of the International Conference on Affective Computing and Intelligent Interaction 2025. (In press).

Text ACII2025_CR_44.pdf - Accepted Version Access restricted to UCL open access staff until 20 July 2026. Download (670kB)

Abstract

As autonomous vehicles (AVs) evolve, ensuring passenger comfort becomes a key priority, particularly through intelligent in-car climate regulation. Most studies focus on thermal sensation using proxy metrics like the predicted mean vote. In contrast, we directly predict thermal comfort—a broader and more complex state influenced by physiological and psychological factors. Using the UBComfort dataset, which includes selfreported thermal comfort labels and multimodal physiological data from semi-ecological vehicle settings, we evaluate the effectiveness of individual modalities and fusion approaches. Our results show that electrodermal activity, and to a certain extent skin temperature, is the most informative single signal for thermal comfort prediction. Furthermore, decision-level fusion outperformed unimodal models and feature-level fusion, achieving F1 score of 0.73 in binary classification. These findings provide empirical insights supporting future efforts toward personalized comfort systems in AVs.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item