Learning Deformable Object Manipulation from Expert Demonstrations

TL;DR

This paper introduces DMfD, a novel Learning from Demonstration method for deformable object manipulation that effectively balances exploration and expert guidance, achieving superior performance on simulated and real-world tasks.

Contribution

The paper presents DMfD, a new LfD approach that handles high-dimensional deformable manipulation tasks using states or images, with demonstrated success in simulation and real-world applications.

Findings

DMfD outperforms baselines by up to 12.9% on state tasks and 33.44% on image tasks.

DMfD achieves comparable or better robustness to randomness.

The method sets new benchmarks on challenging cloth folding environments.

Abstract

We present a novel Learning from Demonstration (LfD) method, Deformable Manipulation from Demonstrations (DMfD), to solve deformable manipulation tasks using states or images as inputs, given expert demonstrations. Our method uses demonstrations in three different ways, and balances the trade-off between exploring the environment online and using guidance from experts to explore high dimensional spaces effectively. We test DMfD on a set of representative manipulation tasks for a 1-dimensional rope and a 2-dimensional cloth from the SoftGym suite of tasks, each with state and image observations. Our method exceeds baseline performance by up to 12.9% for state-based tasks and up to 33.44% on image-based tasks, with comparable or better robustness to randomness. Additionally, we create two challenging environments for folding a 2D cloth using image-based observations, and set a performance benchmark for them. We deploy DMfD on a real robot with a minimal loss in normalized performance during real-world execution compared to simulation (~6%). Source code is on github.com/uscresl/dmfd