GausSim: Foreseeing Reality by Gaussian Simulator for Elastic Objects

ArXiv, 2024

Foresee Dynamics in Real World

Interactive Dynamics

The

Abstract

We introduce GausSim, a novel neural network-based simulator designed to capture the dynamic behaviors of real-world elastic objects represented through Gaussian kernels. We leverage continuum mechanics and treat each kernel as a Center of Mass System (CMS) that represents continuous piece of matter, accounting for realistic deformations without idealized assumptions. To improve computational efficiency and fidelity, we employ a hierarchical structure that further organizes kernels into CMSs with explicit formulations, enabling a coarse-to-fine simulation approach. This structure significantly reduces computational overhead while preserving detailed dynamics. In addition, GausSim incorporates explicit physics constraints, such as mass and momentum conservation, ensuring interpretable results and robust, physically plausible simulations. To validate our approach, we present a new dataset, READY, containing multi-view videos of real-world elastic deformations. Experimental results demonstrate that GausSim achieves superior performance compared to existing physics-driven baselines, offering a practical and accurate solution for simulating complex dynamic behaviors.

Experiments

Qualitative Results

Comparing with Ground Truth (Real Videos)

Comparing with Ground Truth (Synthetic Dynamics)

More Comparisons

The

GausSim

Our GausSim integrates with physics in the following ways:

  1. We formulate GausSim based on continuum mechanics.
  2. We propose a hierarchical structure with interpretable formulations for the simulation process.
  3. We constrain GausSim with mass and momentum conservation in an explicit manner.

Paper

Citation

@InProceedings{shao2024GausSim,
 author = {Shao, Yidi and Huang, Mu and Loy, Chen Change and Dai, Bo},
 title = {GausSim: Registering Elastic Objects into Digital World by Gaussian Simulator},
 booktitle = {CoRR},
 year = {2024}
}

Contact


Shao Yidi
Email: yidi001 at e.ntu.edu.sg