Self-supervised Learning of Latent Space Dynamics

TL;DR

This paper introduces a neural latent-space integrator for subspace simulation of deformable objects, significantly improving efficiency and stability, making real-time simulation feasible on portable devices.

Contribution

The paper presents a novel self-supervised learning framework that operates entirely in latent space for deformable object simulation, reducing computational costs and enhancing generalization.

Findings

Effective simulation of rods, shells, and solids

Enhanced inference stability and generalization

Suitable for portable device deployment

Abstract

Modeling the dynamic behavior of deformable objects is crucial for creating realistic digital worlds. While conventional simulations produce high-quality motions, their computational costs are often prohibitive. Subspace simulation techniques address this challenge by restricting deformations to a lower-dimensional space, improving performance while maintaining visually compelling results. However, even subspace methods struggle to meet the stringent performance demands of portable devices such as virtual reality headsets and mobile platforms. To overcome this limitation, we introduce a novel subspace simulation framework powered by a neural latent-space integrator. Our approach leverages self-supervised learning to enhance inference stability and generalization. By operating entirely within latent space, our method eliminates the need for full-space computations, resulting in a highly efficient method well-suited for deployment on portable devices. We demonstrate the effectiveness of our approach on challenging examples involving rods, shells, and solids, showcasing its versatility and potential for widespread adoption.