EDO-Net: Learning Elastic Properties of Deformable Objects from Graph Dynamics

TL;DR

EDO-Net is a model that learns the elastic properties of deformable objects from graph dynamics, enabling generalization to unknown properties and transfer to new tasks without ground-truth labels.

Contribution

The paper introduces EDO-Net, a novel approach that jointly learns a latent representation of physical properties and predicts future states of deformable objects without relying on labeled data.

Findings

Successfully generalizes to unknown physical properties.

Transfers learned representations to new downstream tasks.

Performs well in both simulation and real-world environments.

Abstract

We study the problem of learning graph dynamics of deformable objects that generalizes to unknown physical properties. Our key insight is to leverage a latent representation of elastic physical properties of cloth-like deformable objects that can be extracted, for example, from a pulling interaction. In this paper we propose EDO-Net (Elastic Deformable Object - Net), a model of graph dynamics trained on a large variety of samples with different elastic properties that does not rely on ground-truth labels of the properties. EDO-Net jointly learns an adaptation module, and a forward-dynamics module. The former is responsible for extracting a latent representation of the physical properties of the object, while the latter leverages the latent representation to predict future states of cloth-like objects represented as graphs. We evaluate EDO-Net both in simulation and real world, assessing its capabilities of: 1) generalizing to unknown physical properties, 2) transferring the learned representation to new downstream tasks.