Power diagram based simulation and rendering for dynamic 3D scenes
PowerSim: Differentiable Physics Simulation and Rendering with Power Diagrams
Computer Vision and Pattern Recognition
Summary
Simulating realistic motion and appearance changes in 3D scenes is difficult because geometry and physics must work together. The authors present PowerSim, which links a special 3D representation called PowerFoam with a physics simulation method called Material Point Method. This connection lets the simulated movements directly update the scene’s shape and look without conversions. PowerSim also allows interactive scene changes, material property recovery, combining different scenes, and realistic lighting effects that adjust with motion.
What this means in practice
- •For interactive graphics developers: Simulate and render deformable 3D objects that respond naturally to user interaction in games or VR environments.
- •For visual effects artists: Create complex dynamic scenes combining multiple captured objects with consistent physically based lighting and deformation.
Authors
Trong-Tung Nguyen, Anand Bhattad
Abstract
We introduce PowerSim, a method to bring physically grounded, differentiable dynamics to PowerFoam's power diagram based 3D representation. PowerSim directly couples a pre-trained PowerFoam scene to the Material Point Method (MPM) by exploiting a natural alignment between the two: the geometric and appearance properties of each primitive correspond closely to the quantities MPM already tracks as an object deforms. Consequently, simulated motion can drive the scene's geometry and appearance directly, without an auxiliary representation in between. Built on this framework, we enable a range of applications on real and synthetic scenes: (1) simulating a static scene under user interaction, (2) recovering spatially varying material fields, (3) compositing primitives from independently captured scenes into a single simulation-ready scene and (4) ray-tracing reflections that update consistently as the object deforms. Our results suggest that PowerSim excels over previous frameworks for physically grounded dynamics, while unlocking unique advantages-such as secondary ray lighting effects on dynamic scenes. Results are best viewed on our project website: https://power-sim.github.io/.