Learning Dense Visual Correspondences in Simulation to Smooth and Fold Real Fabrics
This addresses the problem of efficiently learning multiple deformable manipulation tasks for robotics, though it is incremental as it builds on prior work in simulation-to-real transfer.
The paper tackles the challenge of robotic fabric manipulation by learning visual correspondences in simulation to enable imitation of various fabric smoothing and folding tasks on real robots, achieving an 80.3% average task success rate across 10 tasks on two robotic systems.
Robotic fabric manipulation is challenging due to the infinite dimensional configuration space, self-occlusion, and complex dynamics of fabrics. There has been significant prior work on learning policies for specific deformable manipulation tasks, but comparatively less focus on algorithms which can efficiently learn many different tasks. In this paper, we learn visual correspondences for deformable fabrics across different configurations in simulation and show that this representation can be used to design policies for a variety of tasks. Given a single demonstration of a new task from an initial fabric configuration, the learned correspondences can be used to compute geometrically equivalent actions in a new fabric configuration. This makes it possible to robustly imitate a broad set of multi-step fabric smoothing and folding tasks on multiple physical robotic systems. The resulting policies achieve 80.3% average task success rate across 10 fabric manipulation tasks on two different robotic systems, the da Vinci surgical robot and the ABB YuMi. Results also suggest robustness to fabrics of various colors, sizes, and shapes. See https://tinyurl.com/fabric-descriptors for supplementary material and videos.