Robotic Manipulation for Flattening Wrinkled Cloth: A Human-Inspired Algorithm

This paper presents a Human-Inspired Algorithm for Robotic Cloth Flattening (HI-ARCF) by leveraging human strategies. Using a cloth with a single wrinkle, data from human participants were collected to examine the relationship between wrinkle attributes and pulling actions with a single finger. This enabled the development of a regression-based model that captured the human strategy for cloth flattening. This strategy was generalized into an algorithm to flatten cloth with single as well as multiple wrinkles. The algorithm was implemented using a robot with a finger-like end-effector and its efficacy evaluated using a flatness recovery metric. Based on this metric, results demonstrated successful flattening of cloth with single and multiple wrinkles, including cases outside the training data. These results highlight the success of a data-driven approach to developing human-inspired algorithms for robotic manipulation.