Abstract
Robotic manipulation of cloth remains challenging due to the complex dynamics of cloth, lack of a low-dimensional state representation, and self-occlusions. In contrast to previous model-based approaches that learn a pixel-based dynamics model or a compressed latent vector dynamics, we propose to learn a particle-based dynamics model from a partial point cloud observation. To overcome the challenges of partial observability, we infer which visible points are connected on the underlying cloth mesh. We then learn a dynamics model over this visible connectivity graph. Compared to previous learning-based approaches, our model poses strong inductive bias with its particle based representation for learning the underlying cloth physics; it can generalize to cloths with novel shapes; it is invariant to visual features; and the predictions can be more easily visualized. We show that our method greatly outperforms previous state-of-the-art model-based and model-free reinforcement learning methods in simulation. Furthermore, we demonstrate zero-shot sim-to-real transfer where we deploy the model trained in simulation on a Franka arm and show that the model can successfully smooth cloths of different materials, geometries and colors from crumpled configurations.
@inproceedings{lin2021learning,
title={Learning Visible Connectivity Dynamics for Cloth Smoothing},
author={Lin, Xingyu and Wang, Yufei and Huang, Zixuan and Held, David},
booktitle={Conference on Robot Learning},
year={2021}}
Videos above show a Franka arm smoothing different types of clothes in the real world. Left: Smoothing sequences along with open-loop prediction of VCD on the best pick-and-place action, and scores of all sampled actions. Right: 24 smoothing sequences on three clothes (8 out of the 12 sequences evaluated for each cloth)
Simulator (Ground truth) VCD prediction Visualizing the inferred mesh edges
Square Cloth
Rectangular Cloth
T-shirt
