Software Architecture For Deformable Linear Object Manipulation: A Shape Manipulation Case Study

Abstract

Deformable linear object manipulation is challenging due to their high dimensional configuration space and their underactuated nature when manipulated by a robotic gripper. Due to the complexity of the task, robotic manipulation relies on sensors and computationally demanding models, which end up in multiple different software components interacting with each other. Research in deformable object manipulation usually focuses on modeling, planning or control, without focusing on a software architecture. This paper presents a novel software architecture for deformable linear object manipulation. The software architecture includes components for deformable linear object manipulation, namely perception-, observation-, robot control-, planning-, communication- and decision component. On top of these components, a layered software architecture consisting of a decision layer, a skill layer and a functional layer is presented The proposed concept aims to be a blueprint for a unified software architecture satisfying the requirements of robotic systems to achieve deformable linear object manipulation. The validation of the software architecture is done in a case study of an autonomous shape manipulation task, where one robot and a stereo camera shape a deformable linear object to a predefined desired shape. This use case is inspired by an automated cable routing process, which today in the industry is still mainly handled manually and therefore offers a vast potential for automation.