Scene-Level Heterogeneous Physics Simulation with 3D Gaussian Splats

1The University of Hong Kong    2University of Cambridge
CVPR Findings 2026

TL;DR A Representation Abstraction Framework for making captured 3DGS assets physically interact with meshes, fluids, cloth, rigid bodies, and complex scenes.

Abstract

3D Gaussian Splatting (3DGS) has achieved state-of-the-art photorealistic rendering, but the representation gap prevents these assets from being physically interactive. Production-grade physics engines do not understand the 3DGS representation, while prior physics-for-3DGS methods are monolithic silos. These prior works are fundamentally limited, demonstrating only object-centric physics in isolated environments, such as on an ideal plane — they are incapable of interacting with complex static collision geometry or heterogeneous assets. We propose a novel framework that, for the first time, bridges this gap by enabling 3DGS assets to participate in scene-level, heterogeneous, multi-solver physical simulations. Our core contribution is a Representation Abstraction Framework that “translates” all diverse assets—including 3DGS, virtual meshes, and fluids—into a unified physical particle set. This abstraction is key to enabling complex behaviors, such as the non-rigid deformation of 3DGS assets, within a unified physics pipeline. This particle set, along with the static scene collision boundaries derived from scene capture, is processed within a solver-agnostic physics kernel. The physical results are then mapped back to drive each asset’s specific visual reconstruction. This architecture unlocks capabilities impossible with prior art. We demonstrate complex, two-way interactions between deformable 3DGS assets, standard CG assets (like fluids and meshes), and large-scale captured static environments, showcasing realistic coupled phenomena that were previously unattainable.

Representation Abstraction Framework

RAF method overview
AAbstract

Prepare static scene boundaries and unified simulation particles.

BSimulate

Couple MPM, SPH, PBD, rigid-body, and articulated-body solvers.

CRecouple

Map physical states back to 3DGS, mesh, and particle visual outputs.

Qualitative results

Rendered using Unreal Engine.

Fluid in a virtual bowl

SPH fluid interacts with a mesh bowl and a captured 3DGS garden scene.

Sauce on GS soft body

Viscous fluid couples with a captured 3DGS donut soft body.

Robot-arm manipulation

An articulated robot manipulates a rigid object in a captured scene.

Cloth on statue

PBD cloth follows complex captured statue geometry.

Rigid bodies in 3DGS basket

Virtual fruits collide with each other and an imported 3DGS container.

Additional qualitative examples

Rendered with a pure 3DGS renderer.

Heterogeneous materials

Material variation inside a captured environment.

Granular-soft interaction

Particle-like dynamics coupled with deformable visual assets.

Tool manipulation

A tool interacts with physically simulated objects.

Viscous coupling

High-viscosity fluid coupled with an elastic object.

BibTeX

@inproceedings{liu2026raf,
  title     = {Scene-Level Heterogeneous Physics Simulation with 3D Gaussian Splats},
  author    = {Liu, Xiaoyang and Wu, Shangzhe and Han, Kai},
  booktitle = {CVPR Findings},
  year      = {2026}
}