Learning to Simulate Unseen Physical Systems with Graph Neural Networks

TL;DR

This paper introduces GPE, a graph neural network-based physics engine that generalizes to unseen materials and improves simulation stability by embedding physical laws, enabling accurate multi-step physical system predictions.

Contribution

The paper presents a novel GPE model that incorporates physical priors and material parameters, enhancing generalization and simulation stability for unseen substances.

Findings

GPE generalizes to materials with unseen properties.

Embedding physical laws improves learning efficiency.

GPE performs accurate multi-step simulations.

Abstract

Simulation of the dynamics of physical systems is essential to the development of both science and engineering. Recently there is an increasing interest in learning to simulate the dynamics of physical systems using neural networks. However, existing approaches fail to generalize to physical substances not in the training set, such as liquids with different viscosities or elastomers with different elasticities. Here we present a machine learning method embedded with physical priors and material parameters, which we term as "Graph-based Physics Engine" (GPE), to efficiently model the physical dynamics of different substances in a wide variety of scenarios. We demonstrate that GPE can generalize to materials with different properties not seen in the training set and perform well from single-step predictions to multi-step roll-out simulations. In addition, introducing the law of momentum conservation in the model significantly improves the efficiency and stability of learning, allowing convergence to better models with fewer training steps.