引用本文:
【打印本页】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 87次   下载 392 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于生成式大模型的老年友好城市设计 ——以杭州市街道为例
马 爽1, 汪碧妍2, 李双金3
1.浙江大学建筑工程学院,研究员;2.浙江大学建筑工程学院,硕士研究生;3.( 通讯作者):广岛大学先进理工学院, 助理教授,lisj@hiroshima-u.ac.jp
摘要:
中国正面临深度老龄化的挑战,建设 老年友好城市已成为迫切目标。本研究利用百度 “文心一格”大模型,结合居民对街道的老年友 好需求生成改造方案,进一步统计分析各群体 的老年友好需求差异及整体特征。研究发现:第 一,生成式大模型可以有效地应用在老年友好 城市设计;第二,女性更关注老年友好街道的环 境设施功能,男性更关注街道形态;受教育水平 越高,老年友好街道的绿化率需求越高;第三, 当前街道的天空开敞度、围墙(栏)、步行道、路 灯、车流和车行道现状与老年友好街道需求之间 仍存在差距。本研究以期为老年友好理念下的城 市规划和设计提供创新方法和实践参考。
关键词:  老年友好  城市设计  大模型  街景图 像  个体属性特征
DOI:10.13791/j.cnki.hsfwest.20240205
分类号:
基金项目:
Designing elderly-friendly cities based on a generative large language model: A case studyof streets in Hangzhou
MA Shuang,WANG Biyan,LI Shuangjin
Abstract:
The global aging population is accelerating, with China’s aging population being prominent. China is expected to become the country with the largest aged population (defined as aged 60 years or older) in the world by 2050. At present, China is actively carrying out the “active aging” action, and relevant documents emphasize the importance of responding to this national strategy from the perspective of urban planning. Analyzing street scale is crucial for constructing elderly-friendly cities, yet there is limited research on evaluating spatial elements of streets and the varied demands for such cities across socio-economic groups. Generative Large Language Models (GLLMs), with their unique advantages such as high generalization ability, multitasking learning, and powerful computing resources, can consider spatial contextual elements and quickly generate planning and designing drawings based on public will, which helps to break the dilemma of planning before evaluation. However, there is currently limited research on the application of GLLMs to guide the construction of elderly friendly cities. Therefore, based on the concept of “active aging”, this study collects 118 effective questionnaires on the survey of street elderly-oriented renovation, which uses street view images within Hangzhou City, China as carriers. The survey respondents are residents of Hangzhou City aged 30 years or older. The questionnaire content includes their individual attributes, whether the street is suitable for elderly-oriented renovation, and the corresponding elderly-oriented renovation needs (including 22 element indicators from street form, environmental facility, road and transportation, and green landscape dimensions). Then, this study selects Baidu “Wenxin Yige”, a multi-modal GLLM, as the automation designing platform, and inputs street view images (image mode) that are suitable for elderly-oriented renovation and corresponding street renovation requirements (text mode), until the corresponding elderly-oriented renovation designing diagrams that satisfy the respondents are generated. Finally, the street view images before and after the elderly-oriented renovation are subjected to semantic segmentation to extract the percentage of environmental elements in each street, and further statistical analysis is conducted on the differences in elderly friendly needs among different groups and their overall characteristics. The results indicate that the application of GLLMs in the design of elderly friendly cities is feasible and effective. Firstly, there are significant differences in the attention points of respondents of different genders, ages, education levels, housing properties, and pension (or income) towards urban environmental elements in streets after elderly-oriented renovation. 1)The male groups place more emphasis on the renovation of elements related street form and environmental quality to improve the recognition of street space. While, the female groups place greater emphasis on the elements related transformation of functional and accessibility, such as more commercial service facilities and urban furniture facilities. 2)The preferences of different groups with different levels of education and housing properties towards the proportion of various urban environmental elements are relatively consistent. However, based on the results of the questionnaire survey, it is found that groupswith higher education levels have higher requirements for the quality of green spaces, and groups with different housing properties have different specific designing requirements for various elements. 3)There are significant differences in the demand for elderly friendly street elements among different age groups and pension (or income) groups. Secondly, the current situation of buildings, rest seats, greenery, and signage (lights) within the scope of this study has roughly meet the requirements of elderly friendly cities, but there is still an urgent need to improve the openness of the sky, walls (fences), pedestrian walkways, street lights, traffic flow, and lane planning. Finally, based on the significance of the percentage changes in environmental factors in each city, 118 streets can be divided into four types of elderly-oriented renovation: street form renovation (N=20), green landscape renovation (N=36), environmental facility renovation (N=29), and road and traffic renovation (N=33). Based on the above results, further research can be conducted in the future. In the construction of elderly-friendly cities, it is essential to tailor the detail-oriented renovations to the characteristics of the residents in the area where the street is located. For example, in areas with affordable housing, the street’s architectural facades could be enriched with colors, and park green space facilities such as fitness equipment, walking paths, children’s play areas, and rest seats could be added. For streets in high-density blocks, street-side green spaces can be cleverly integrated to meet the needs of various groups. Furthermore, the advancement of road space redistribution ensures the protection of pedestrian slow traffic spaces, especially in streets within resettlement housing blocks.
Key words:  elderly-friendly  urban design  large language model  street view image  individual attributes