Human behavior is a core element and key driving force in architectural design. Unlike traditional rule-driven agents, AI-Agents can learn and summarize human behavioral characteristics in diverse environments to support intelligent decision-making. 

This project is an achievement of the 2025 Digital Future Workshop. Taking the campus scenario as an example, it carries out innovative research on AI-Agent technology. First, virtual tasks are designed based on large language models (LLM), and human participants are organized to conduct role-playing to collect multi-factor-driven behavioral data of different roles. Then, machine learning algorithms are used to train the Agent-Me model to predict human behavior patterns on campus. Together with participants, we also explore the application of Agent-Me in the Rhino-Grasshopper platform, conduct real-time performance evaluation of the built environment from the agent’s perspective, and assist design decision-making and intelligent optimization.