Flexible Neural Representation for Physics Prediction

TL;DR

This paper introduces a hierarchical particle-based neural network that effectively predicts complex physical dynamics, including collisions and deformations, across diverse 3D objects and large scenes.

Contribution

The paper presents a novel hierarchical relation network that models physical interactions with a flexible, differentiable graph-based approach for improved dynamics prediction.

Findings

Accurately predicts complex collisions and deformations.

Handles large scene configurations effectively.

Generates plausible long-term dynamics predictions.

Abstract

Humans have a remarkable capacity to understand the physical dynamics of objects in their environment, flexibly capturing complex structures and interactions at multiple levels of detail. Inspired by this ability, we propose a hierarchical particle-based object representation that covers a wide variety of types of three-dimensional objects, including both arbitrary rigid geometrical shapes and deformable materials. We then describe the Hierarchical Relation Network (HRN), an end-to-end differentiable neural network based on hierarchical graph convolution, that learns to predict physical dynamics in this representation. Compared to other neural network baselines, the HRN accurately handles complex collisions and nonrigid deformations, generating plausible dynamics predictions at long time scales in novel settings, and scaling to large scene configurations. These results demonstrate an architecture with the potential to form the basis of next-generation physics predictors for use in computer vision, robotics, and quantitative cognitive science.