Type:

Year:

Editor:

Author:

Pingfei JAING; Ji Han

Series:

Institution:

Page(s):

Abstract:

Incorporating sustainability principles into early product design phases presents significant challenges due to the complex, interconnected nature of product lifecycles. While Large Language Models (LLMs) demonstrate considerable potential for design ideation, their current application in sustainable design relies predominantly on unstructured, single-shot prompting approaches that inadequately address holistic lifecycle considerations. This study presents a novel multi-agent LLM framework for earlystage sustainable concept generation. The proposed architecture employs specialized LLM-powered agents distributed across four product lifecycle phases: Material, Production, Use, and End-of-Life. By employing Chain-of-Thought (CoT) reasoning protocols at each stage, the system generates comprehensive product design specifications for sustainable design exploration. These specifications can then be processed by a textual summarization agent, which transforms technical requirements into descriptive prompts for visual concept generation via a text-to-image generation agent. The framework’s potential is demonstrated through a case study on a sustainable hair dryer concept development. Results demonstrate successful integration of lifecycle considerations into the design processes, producing concepts exhibiting both innovation and sustainability. This research establishes a foundation for leveraging AI in systematic sustainable design exploration, offering a structured alternative to conventional single-agent, single-shot prompting, enabling more rigorous and comprehensive sustainable design ideation.

Keywords:

Large language model, Sustainable design, Concept generation, Multi-agent system, Chain-of-thought reasoning