Physics grounded system synthesizes dynamic multi part object scenes
PhysMAS: Physics-Grounded Multi-Agent Synthesis of Compositional 4D Gaussians
Artificial Intelligence
Summary
Creating believable animations of moving objects made of many parts is hard, especially when those parts interact physically. The authors present PhysMAS, a system that uses multiple agents to understand and assign physical materials to different parts in a scene, ensuring realistic motion according to physics. Unlike previous approaches, PhysMAS avoids slow optimizations and handles scenes with multiple objects interacting naturally. This leads to more physically plausible and semantically accurate dynamic scenes faster.
What this means in practice
- •For game developers: Generate realistic animations for complex multi-part objects interacting physically within game scenes without costly per-scene optimization.$Commercial implications: Enables game studios to produce physically plausible dynamic environments efficiently, improving game realism and reducing animation costs.
- •For virtual reality designers: Create immersive VR scenes with multiple interacting objects exhibiting physically accurate motion from limited inputs and scene views.
Authors
Jiang Qin, Chunji Lv, Yangguang Wei, Yang Gao, Ming Liu, Lizhong Ding, Ye Yuan, Yinjie Lei, Changsheng Li
Abstract
Efficient, fully automatic, and physically plausible 4D Gaussian synthesis is an important goal for dynamic scene generation. Recent physics-based methods couple 3D Gaussians with the Material Point Method (MPM) to generate physically driven motion, but extending this paradigm to heterogeneous multi-part objects and interacting multi-object scenes remains challenging. Object-level physical assignment collapses distinct parts into a single material state, while one-shot predictions from large language models, vision-language models, or agents neither reliably bind different materials to identified parts nor verify that the resulting MPM configuration is executable. Score Distillation Sampling (SDS)-based parameter optimization, meanwhile, requires repeated per-scene score evaluations and gradient backpropagation, incurring lengthy optimization and potentially yielding suboptimal or unstable solutions. We therefore present PhysMAS, a physics-grounded multi-agent framework. From a motion prompt and four scene views, an Object-Part Scene Agent establishes persistent identities and calls a Material Reasoning Agent for part-wise profiles. It invokes solver-aware skills to bind these identities and profiles to per-particle MPM fields and execute all objects in a shared domain; the framework then screens candidate forward-simulation results. This supports heterogeneous multi-part and interacting multi-object scenes without per-scene diffusion-score backpropagation. Extensive experiments demonstrate that, compared with recent physics-based 4D Gaussian baselines that rely on SDS, PhysMAS achieves better semantic alignment and perceived physical plausibility while requiring less runtime.