Zan Wang

Beijing Academy of Artificial Intelligence, Beijing University of Posts and Telecommunications

Papers

Total Citations

191

H-Index

About

Zan Wang is a leading researcher at the intersection of 3D scene understanding, generative modeling, and multi-robot systems. Their most impactful work introduces **SceneDiffuser**, a pioneering conditional generative model that unifies scene-conditioned generation, optimization, and planning within a single framework. Unlike prior approaches, SceneDiffuser is intrinsically scene-aware, physics-based, and goal-oriented, enabling realistic and functional 3D scene synthesis. This work has rapidly accumulated over 165 citations, establishing Wang as a key innovator in diffusion-based 3D modeling. In parallel, Wang has made significant contributions to robotics, developing methods for **multi-robot cooperative task allocation** with definite path-conflict-free handling, addressing a critical challenge in real-world deployment. Their latest research, **PhysPart**, tackles the complex problem of physically plausible part completion for interactable objects, bridging the gap between 3D generative models and practical applications in robot simulation and 3D printing. Through this body of work, Zan Wang is advancing the frontier of how machines understand, generate, and interact with complex 3D environments.