Unfolding the Potential of Point-Based Correspondences for Cloth Manipulation

Abstract

Robotic cloth manipulation is an active area of research with numerous applications in domestic and industrial environments. However, prior work in this field has limitations that restrict their applicability in real-world scenarios. For instance, these approaches often require subgoals for long-horizon tasks and face challenges in handling unaligned configurations. By ``unaligned configurations``, we refer to situations where the initial orientation of the cloth surface differs from the goal orientation. To curb these issues, we propose utilizing point-based correspondences, which capture geometric relationships and deformations in cloth surfaces. Point-based correspondences refer to establishing correspondences between points on the cloth surface, allowing us to track and model the cloth's behavior accurately. In this work, we present two automated cloth manipulation solutions that incorporate the use of point-based correspondences. Our focus centers on fundamental cloth manipulation tasks, including folding, smoothing, and alignment. Through extensive experiments and evaluations, we demonstrate the effectiveness of our proposed approaches, which surpass state-of-the-art methods. Comparative experiments against existing techniques highlight the distinct advantages of using point-based correspondences for achieving efficient, robust, and long-horizon cloth manipulation.