Type and Fabric Agnostic Methods for Robotic Unfolding of Folded Garments

This paper introduces two type-agnostic and fabric-agnostic methods for robotic unfolding of garments. In the first method, a new path extraction procedure for unfolding the upper layer of a garment’s fold, which is located on a working table, with one robotic arm is proposed. The path is extracted using learning by demonstration methods and it is customized on each fold by integrating features extracted by a depth sensor, such as the fold’s curve height. Moreover, the unfolding path can be re-adjusted through the classification of visual data by Support Vector Machines. In the second introduced method, the garment is lifted in the air by two points so that it unfolds due to gravity while the criterion proposed for the selection of these grasp points is the maximization of the free of deformations hanging’s garment area. The proposed methods are evaluated experimentally in simulation and real environment while comparisons with similar approaches prove their advantages.