RAISECity: A Multimodal Agent Framework for Reality-Aligned 3D World Generation at City-Scale
Shengyuan Wang, Zhiheng Zheng, Yu Shang, Lixuan He, Yangcheng Yu, Fan Hangyu, Jie Feng, Qingmin Liao, Yong Li
2025-11-27
Summary
This paper introduces RAISECity, a new system for automatically creating detailed and realistic 3D models of entire cities.
What's the problem?
Currently, making large-scale 3D city models is really hard. Existing methods struggle to create models that look good, accurately reflect the real world, and can be scaled up to cover large areas. They often have errors that build up as the model gets bigger, leading to inaccuracies and a less realistic final product.
What's the solution?
The researchers developed RAISECity, which works like a team of agents. These agents use different AI tools to gather information about real cities – things like images and maps – and then build the 3D model step-by-step. The system constantly checks its work and improves it, minimizing errors and making sure the final model is accurate and visually appealing. It’s designed to process data efficiently and refine the model through multiple iterations.
Why it matters?
This work is important because realistic 3D city models are crucial for developing things like self-driving cars, robots that navigate the real world, and immersive virtual reality experiences. RAISECity creates models that are high quality, look like the real world, can cover large areas, and work well with existing graphics software, making it a strong foundation for these kinds of applications.
Abstract
City-scale 3D generation is of great importance for the development of embodied intelligence and world models. Existing methods, however, face significant challenges regarding quality, fidelity, and scalability in 3D world generation. Thus, we propose RAISECity, a Reality-Aligned Intelligent Synthesis Engine that creates detailed, City-scale 3D worlds. We introduce an agentic framework that leverages diverse multimodal foundation tools to acquire real-world knowledge, maintain robust intermediate representations, and construct complex 3D scenes. This agentic design, featuring dynamic data processing, iterative self-reflection and refinement, and the invocation of advanced multimodal tools, minimizes cumulative errors and enhances overall performance. Extensive quantitative experiments and qualitative analyses validate the superior performance of RAISECity in real-world alignment, shape precision, texture fidelity, and aesthetics level, achieving over a 90% win-rate against existing baselines for overall perceptual quality. This combination of 3D quality, reality alignment, scalability, and seamless compatibility with computer graphics pipelines makes RAISECity a promising foundation for applications in immersive media, embodied intelligence, and world models.