EDO-Net: Learning Elastic Properties of Deformable Objects from Graph Dynamics
This addresses the challenge of predicting deformable object behavior for robotics applications, but it is incremental as it builds on existing graph dynamics methods.
The paper tackles the problem of learning graph dynamics for deformable objects to generalize to unknown physical properties, proposing EDO-Net which learns a latent representation of elastic properties without ground-truth labels and achieves generalization and transfer capabilities in simulation and real-world evaluations.
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.