Evaluation of thermal comfort in urban commercial space with Vision Language Models-based agent model

Thermal comfort in urban commercial spaces plays a crucial role in influencing both business performance and public well-being. Traditional evaluation methods, which typically depend on field surveys and expert assessments, are often time-consuming and labor-intensive. This study introduces a novel proxy evaluation method based on the Vision Language Models (VLMs) to assess thermal comfort in commercial spaces. Specifically, we developed a VLMs-based agent system (ChatGPT-4o) that simulates eight distinct roles with diverse demographic backgrounds to assess the thermal comfort of Harbin Central Street’s commercial district through visual analysis. To validate the method’s effectiveness, we compared the VLMs assessment results with evaluations from 200 professional volunteers based on a 25% sample (n=143). Additionally, we conducted spatial distribution and thematic analyses to examine evaluation patterns and identify the key factors influencing thermal comfort. The findings reveal the following: First, the VLMs assessments showed a significant correlation with human expert evaluations (r=0.602, p

城市商业空间的热舒适度，对经营绩效与公众福祉均具有至关重要的影响。传统评估方法通常依托实地调研与专家评定，往往耗时耗力。本研究提出一种基于视觉语言模型（Vision Language Models, VLMs）的新型替代评估方法，用于商业空间热舒适度的评估。具体而言，本研究构建了一套基于视觉语言模型的智能体系统（ChatGPT-4o），该系统模拟了8个具备不同人口统计学背景的角色，通过视觉分析对哈尔滨中央大街商业区的热舒适度开展评估。为验证该方法的有效性，本研究基于25%的抽样样本（n=143），将视觉语言模型的评估结果与200名专业志愿者的评定结果进行对比。此外，本研究通过空间分布与主题分析，探究评估模式并识别影响热舒适度的关键因素。研究结果如下：其一，视觉语言模型的评估结果与人类专家评定结果存在显著相关性（r=0.602，p