Efficiently Learning Single-Arm Fling Motions to Smooth Garments

TL;DR

This paper introduces a data-efficient method for a robot to learn single-arm fling motions for garment smoothing, combining bandit algorithms and optimization to adapt to various garments with minimal training time.

Contribution

It presents a novel coarse-to-fine learning approach that significantly accelerates the learning of fling trajectories for garment smoothing, especially leveraging prior experience.

Findings

Achieves 60-94% garment coverage within 30 minutes.

Reduces learning time by up to 87% with prior experience.

Outperforms baseline methods in learning efficiency.

Abstract

Recent work has shown that 2-arm "fling" motions can be effective for garment smoothing. We consider single-arm fling motions. Unlike 2-arm fling motions, which require little robot trajectory parameter tuning, single-arm fling motions are very sensitive to trajectory parameters. We consider a single 6-DOF robot arm that learns fling trajectories to achieve high garment coverage. Given a garment grasp point, the robot explores different parameterized fling trajectories in physical experiments. To improve learning efficiency, we propose a coarse-to-fine learning method that first uses a multi-armed bandit (MAB) framework to efficiently find a candidate fling action, which it then refines via a continuous optimization method. Further, we propose novel training and execution-time stopping criteria based on fling outcome uncertainty; the training-time stopping criterion increases data efficiency while the execution-time stopping criteria leverage repeated fling actions to increase performance. Compared to baselines, the proposed method significantly accelerates learning. Moreover, with prior experience on similar garments collected through self-supervision, the MAB learning time for a new garment is reduced by up to 87%. We evaluate on 36 real garments: towels, T-shirts, long-sleeve shirts, dresses, sweat pants, and jeans. Results suggest that using prior experience, a robot requires under 30 minutes to learn a fling action for a novel garment that achieves 60-94% coverage.