성과공개

3차

구성기술1

연세대학교
Yonsei University

학술발표

아방가르드 실험적 주거성능분석을 위한 대화형 인공지능 기반의 디지털 휴먼 개발에 관한 연구
Developing a Conversational Artificial Intelligence-Equipped Digital Human Model for Analysing Performances of Housing Experimented by the Avant-Garde Architects

대한건축학회 2023년도 추계학술발표대회논문집

43

2023-10-26

497-500

양홍주

홍승완

The Avant-Garde architect's experimental housing plans are pioneering, creative precedents to resolve the given tasks in contemporary architecture, but the resident-focused performance analysis and value-rediscovery are challenging because most of them are unbuilt and imaginary. To resolve the shortcoming, this study aims to develop a new simulation model that generates and customises the digital human schedules and behaviour rules through natural language processing-based conversational artificial intelligence (ChatGPT), in order to analyse the performance of the Avant-Garde housing related to resident behaviour. As a method, this study adopted the three experimental housings focused on residents’ lifestyles, and through ChatGPT, generated sociocultural behaviours of digital humans in two different historical periods. Visibility, density, and circulation were also measured to analyse housing performances. As a result, the simulation analysis indicates (1) the house performance factors, including home workability, privacy, and efficiency for household management, are different between 21st and 19th-century residents. This result demonstrates that the simulation model reflects the distinctive daily behaviours of residents in each era. (2) the simulation model also generates behavioural types, use durations, and places where to visit in order to compare and analyse heterogeneous design features of the housing. The findings of this study address that AI-powered behavioural generation of digital humans can be a means of evidence-based interpretation to discover hidden values in the Avant-Garde housing.