Physical simulation · Learning

Mengyu Chu 楚梦渝

Tenure-track Assistant Professor

School of Intelligence Science and Technology, Peking University
Visual Computing and Learning Lab

From captures to simulation-ready worlds.

Portrait of Mengyu Chu
Peking University · Beijing

Bio

Research Interests

My research combines deep learning with physical simulation to improve the flexibility of simulation systems and the accuracy and generalizability of learning-based models.

  • Physics-enhanced deep learning
  • Physical simulation
  • Neural reconstruction
  • Visual computing
  1. Assistant Professor Peking University
  2. Lise Meitner Postdoctoral Research Fellow Max Planck Institute for Informatics
  3. Ph.D. in Computer Science Technical University of Munich
  4. M.Eng. in Computer Science and Technology Zhejiang University

News

Publication

3 papers appear in the SIGGRAPH 2026

LagrangianSplats, Robustness Parametric Surface CCD, and LoBoFit explore fluid reconstruction, floating-point-robust CCD, and flexible garment refitting.

Publications

19 works shown

Gaussian Fluids vortex-ring simulation SIGGRAPH 2025

ACM SIGGRAPH Conference Papers

Gaussian Fluids: A Grid-Free Fluid Solver based on Gaussian Spatial Representation

Jingrui Xing , Bin Wang , Mengyu Chu , Baoquan Chen

A grid-free fluid solver that uses Gaussian primitives as a continuous spatial representation.

  • Physical Simulation
  • Gaussian Splatting
  • Fluids
SIGGRAPH 2024

ACM SIGGRAPH Conference Papers

Physics-Informed Learning of Characteristic Trajectories for Smoke Reconstruction

Yiming Wang , Siyu Tang , Mengyu Chu

This work reconstructs temporally coherent smoke by learning characteristic trajectories while respecting the governing physics.

  • Reconstruction
  • Physics-informed Learning
  • Fluids
Data-driven smoke flow synthesis result SIGGRAPH 2017

ACM Transactions on Graphics

Data-driven synthesis of smoke flows with CNN-based feature descriptors

Mengyu Chu , Nils Thuerey

CNN-based feature descriptors enable example-driven synthesis of detailed smoke motion from coarse simulations.

  • Data-driven Simulation
  • Fluids

Preprints

JODA pipeline for composable articulated-object joint dynamics arXiv 2026

Preprint

JODA: Composable Joint Dynamics for Articulated Objects

Tianhong Gao , Cheng Yu , Yinghao Xu , Mengyu Chu

  • Physical Simulation
  • Articulated Objects

Funding

NSFC

重大项目课题

NSFC Major Program Project

面向空间智能的可微物理计算体系

Differentiable Physical Computing Architecture toward Spatial Intelligence

课题负责人 · Project Lead

NSFC

国家优秀青年科学基金项目(海外)

Excellent Young Scientists Fund (Overseas)

动态内容的高保真生成和智能交互创作

High-fidelity generation and intelligent interactive creation of dynamic content

负责人 · Project Lead

MOST

新一代人工智能重大科技专项课题

National Key R&D Program on New Generation Artificial Intelligence

城市动态现象物理建模与仿真

Physical Simulation for Urban Traffic Applications

课题负责人 · Project Lead

Service

Research conversations are welcome.

mchu@pku.edu.cn

Sciences II Building, Office 2214S, Peking University, No. 5 Yiheyuan Road, Haidian District, Beijing, China.