Concepedia

Differentiable Cloth Simulation for Inverse Problems

Junbang Liang, Ming C. Lin, Vladlen Koltun

Neural Information Processing Systems · 2019 · 76 citations · 16 references

TL;DR

Cloth dynamics modeling faces high dimensionality that makes traditional gradient computation for collision response impractical, necessitating robust frameworks for dynamics, self‑collisions, and contacts. The study proposes a differentiable cloth simulator embedded in neural networks, computing gradients via QR decomposition of a reduced matrix to overcome high‑dimensionality challenges. The simulator uses QR decomposition of a reduced matrix to compute gradients efficiently, enabling its integration as a neural‑network layer and application to inverse problems such as parameter estimation and motion control. Experiments show the method accelerates backpropagation by two orders of magnitude and successfully solves inverse problems like parameter estimation and motion control.

Abstract

We propose a differentiable cloth simulator that can be embedded as a layer in deep neural networks. This approach provides an effective, robust framework for modeling cloth dynamics, self-collisions, and contacts. Due to the high dimensionality of the dynamical system in modeling cloth, traditional gradient computation for collision response can become impractical. To address this problem, we propose to compute the gradient directly using QR decomposition of a much smaller matrix. Experimental results indicate that our method can speed up backpropagation by two orders of magnitude. We demonstrate the presented approach on a number of inverse problems, including parameter estimation and motion control for cloth.