Real-time modeling and parameter approximation of dexterous garment folding by robot

One of the challenging tasks for general household robot of doing household work is to manipulate deformable objects such as garments. The improper folding procedure by the robot will cause the failures such as generating wrinkles, cause the garment to misalign and the following folding process will possibly fail. Moreover, given different types and sizes of garment, the robot should able to understand its perception and planning strategy such as the start and end position of grasping points, the trajectory for each sequence and also the completed folding procedure. This paper present a novel generic approach of real-time modeling and parameter approximation applied in the dexterous robotic manipulation of unknown garment using particle-based polygonal model with online simulation. The proposed approach is a system that can judge which of the prepared polygonal models is appropriate. The prepared model is used to simulate a crudely spread-out garment placed on the flat platform that can be recognized by vision sensor. The parameters of the garment can be approximated from the proposed model and then the folding procedure can be planned based on these parameters. Later, the robot can follow the trajectory from simulation to achieve the acceptable result in the real garment folding.