Robotic bimanual cleaning of deformable objects with online learning of part and tool models

Abstract

In this paper, we present an approach to perform automatic robotic cleaning of deformable objects with unknown stiffness characteristics. A bimanual robot setup is used, where one arm holds the part to be cleaned, while the other holds the cleaning tool. The robot maintains an approximate model of the deformation behavior of each part it interacts with and incrementally improves the model as it performs cleaning attempts, thereby gaining information. Simultaneously, the robot maintains a model of the cleaning tool performance which is independent of the particular part and can be learned over multiple episodes of interaction with different parts. During each attempt, the robot exploits its current knowledge of the part deformation behavior to select an optimal set of grasp locations that minimize the amount of deformation. Results indicate the system is able to incrementally learn the deformation model of parts with approximate linear geometry and that the improving model can be quickly used to select the correct grasp locations and tool parameters for rapid cleaning.