This paper presents current approaches for robotic garment folding-oriented 3D deformable object perception and manipulation. A major portion of these approaches are based on 3D perception algorithms that match garments to a model, and are thus model-based. They require a full view of an extended garment, in order to then apply a preprogrammed folding sequence. Other approaches are based on 3D manipulation algorithms that are focused on modifying the pose of the garment, also oriented at matching it with a model. We present our own garment-agnostic algorithm, which requires no model to unfold clothes, and works using a single view from an RGB-D sensor. The unfolding algorithm also has been validated through experiments using a garment dataset of RGB-D sensor data, and additional validation with a humanoid robot platform. Finally, conclusions regarding the current state of the art and on the future trends of these research lines are discussed.
Future Trends in Perception and Manipulation for Unfolding and Folding Garments