A multi-dimensional evaluation of existing thermal comfort models for vehicle cabins

Driven by increasing demands for vehicle thermal comfort, intelligent HVAC and localized thermal management systems are widely adopted in modern vehicles. Their design relies on thermal comfort models that balance predictive accuracy and practical deployability. This study systematically evaluates 13 widely used vehicle thermal comfort models using a unified framework covering four criteria: theoretical completeness, predictive accuracy, generalizability, and measurement feasibility. Experimental validation was conducted under four representative scenarios, including automatic HVAC operation in winter, HVAC combined with seat heating in winter, automatic HVAC cooling in summer, and HVAC with localized solar exposure in summer. The results show that only four models account for transient dynamics, and most neglect local thermal sensation, limiting their applicability in non-uniform cabins. Models incorporating skin temperature outperform environment-only approaches, achieving prediction errors below 1 RMSE across multiple scenarios. The best-performing hybrid model achieved an RMSE of 0.34 under localized solar exposure, whereas environment-only models often exceeded an RMSE of 1. Hybrid models provide the most balanced trade-off between accuracy, generalizability, and feasibility. These findings highlight the need for future vehicle thermal comfort models that integrate dynamic physiological responses while remaining practical for real-time applications.