本文针对已识别的研究空白，提出了一种新型集成架构，将生成式AI与面向数字孪生的仿真原型协同结合，用于城市交通仿真与决策支持。该方法采用基于 Gemini 2.5 Flash Lite 模型的对话式 AI 代理，实现仿真参数的自然语言配置，并在利用成熟仿真工具构建的数字孪生上执行（SimPy 用于离散事件仿真，NetworkX 和 OSMnx 用于空间网络表示，GeoPandas 用于地理空间数据处理）。

This paper addresses the identified gap by presenting a novel integrated architecture that synergistically links generative AI with digital-twin-oriented simulation prototypes for urban mobility simulation and decision support. Our approach employs a conversational AI agent built on the Gemini 2.5 Flash Lite model to enable natural language configuration of simulation parameters, executed on a digital twin constructed using established simulation tools (SimPy for discrete-event simulation, NetworkX and OSMnx for spatial network representation, and GeoPandas for geospatial data processing).