UniPhy: Learning a Unified Constitutive Model for Inverse Physics Simulation

2025-05-22

AI Summary

This work addresses the limitation of conventional constitutive models—namely, their reliance on predefined material class assumptions and poor generalizability across diverse physical behaviors. We propose a unified latent-variable-driven neural constitutive model that eliminates explicit material categorization and task-specific architectures. Leveraging conditional neural networks parameterized by learned latent variables, the model establishes a shared constitutive representation spanning elastic solids, plasticine, granular media (e.g., sand), and Newtonian/non-Newtonian fluids. Integrated with differentiable physics simulation and observation-based latent space optimization, it enables prior-free inverse physical simulation: given only observed object motion trajectories, the framework jointly infers unknown material properties and synthesizes high-fidelity dynamics under novel mechanical conditions. Experiments demonstrate substantial improvements in cross-material motion reconstruction accuracy over state-of-the-art system identification and single-material neural constitutive approaches.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

We propose UniPhy, a common latent-conditioned neural constitutive model that can encode the physical properties of diverse materials. At inference UniPhy allows `inverse simulation' i.e. inferring material properties by optimizing the scene-specific latent to match the available observations via differentiable simulation. In contrast to existing methods that treat such inference as system identification, UniPhy does not rely on user-specified material type information. Compared to prior neural constitutive modeling approaches which learn instance specific networks, the shared training across materials improves both, robustness and accuracy of the estimates. We train UniPhy using simulated trajectories across diverse geometries and materials -- elastic, plasticine, sand, and fluids (Newtonian&non-Newtonian). At inference, given an object with unknown material properties, UniPhy can infer the material properties via latent optimization to match the motion observations, and can then allow re-simulating the object under diverse scenarios. We compare UniPhy against prior inverse simulation methods, and show that the inference from UniPhy enables more accurate replay and re-simulation under novel conditions.

Problem

• Learning a unified neural model for diverse materials' physical properties
• Enabling inverse simulation to infer unknown material properties
• Improving robustness and accuracy over instance-specific neural models

Innovation

• UniPhy uses latent-conditioned neural constitutive model
• Enables inverse simulation via differentiable latent optimization
• Shared training improves robustness and accuracy