Robot Learning for Deformable Object Manipulation Tasks
Doktorsavhandling, 2024

Deformable Object Manipulation (DOM) is a challenging problem in robotics. Until recently, there has been limited research on the subject, with most robotic manipulation methods being developed with rigid objects in mind. Part of the challenge in DOM is that non-rigid objects require algorithms capable of generalizing to changes in shape as well as different mechanical properties. Machine Learning (ML) has been shown successful in fields, such as computer vision and natural language processing, where generalization is important thus encouraging the application of ML to robotic manipulation.

This thesis tackles DOM problems using ML techniques for tasks with Deformable Linear Objects (DLOs), e.g. ropes and cables, found in a variety of industrial applications. DLOs encapsulate a lot of the general challenges in DOM, making them good case studies on the effectiveness of ML for other types of deformable objects. Typically, ML algorithms require large amounts of data that are better satisfied in simulation. Therefore, the ReForm simulation sandbox is introduced, which includes six DLO manipulation tasks. ReForm aims to facilitate comparison and reproducibility of robot learning research on tasks where the goal is to control the shape of a DLO. Such shape control tasks are categorized as: explicit, if a precise shape is to be achieved; or implicit, if its deformation is dictated by a more abstract goal.

Two representative DLO manipulation tasks are addressed: (i) shape-servoing (explicit) and (ii) cable-routing (implicit). For shape-servoing, special emphasis is given to Reinforcement Learning (RL) methods. Initial work tackles shape-servoing of an elastoplastic DLO towards a unique goal, using online RL with ReForm. Subsequent work moves towards a multi-goal task in a real-world experimental setup, using offline RL methods to learn directly from real data. In the cable-routing works, the aim is to lay the groundwork for solving this type of task through motion primitives, with limited use of ML. First, a vision-based approach is presented, which is able to route a cable through randomly placed fixtures. Then, a force-based approach is introduced for a similar problem, in which the state and stiffness of a DLO can be estimated through contact with fixtures.

Deformable Object Manipulation

Deformable Linear Objects

Robotics

Reinforcement Learning

Machine Learning

Robot Learning

EB, Hörsalsvägen 11, Chalmers
Opponent: Prof. Gianluca Palli, University of Bologna, Italy

Författare

Rita Laezza

Chalmers, Elektroteknik, System- och reglerteknik

ReForm: A Robot Learning Sandbox for Deformable Linear Object Manipulation

Proceedings - IEEE International Conference on Robotics and Automation,;Vol. 2021-May(2021)p. 4717-4723

Paper i proceeding

Learning Shape Control of Elastoplastic Deformable Linear Objects

Proceedings - IEEE International Conference on Robotics and Automation,;Vol. 2021-May(2021)p. 4438-4444

Paper i proceeding

Planning and Control for Cable-routing with Dual-arm Robot

Proceedings - IEEE International Conference on Robotics and Automation,;Vol. 2022-May(2022)p. 1046-1052

Paper i proceeding

Feel the Tension: Manipulation of Deformable Linear Objects in Environments with Fixtures using Force Information

IEEE International Conference on Intelligent Robots and Systems,;Vol. 2022-October(2022)p. 11216-11222

Paper i proceeding

While doing my PhD, people would often ask me: What do you work on? and I would reply: Robotics. This would usually lead to a follow-up question: What do you do within Robotics? to which I would reply: My research focuses on deformable object manipulation tasks. I eventually learned to directly follow that statement with an explanation of what Deformable Object Manipulation (DOM) means. I found it best to clarify the concept using a question: Wouldn't it be nice if we had robots that could fold our laundry for us? to which most people would reply with a resounding: Yes! I would then immediately disappoint them, since in fact my work has been limited to tasks involving cables, ropes and wires (i.e. deformable linear objects). However, I think the goal of DOM would become clear to them. Now, this thesis does not in any way try to solve the task of folding laundry, but it does tackle tasks that share a lot of the same DOM challenges. Moreover, the same Robot Learning (RL) approaches explored in this thesis can be applied to other DOM tasks, due to their generality. RL is a promising research area that enables robots to acquire new skills and adapt to their environments through experience, much like humans. Instead of being programmed with specific instructions for each task, robots can use machine learning to optimize their behavior. Therefore, research on RL opens new possibilities for robotics to solve tasks found in different industries, healthcare, and even everyday life.

Infrastruktur

C3SE (Chalmers Centre for Computational Science and Engineering)

Ämneskategorier

Robotteknik och automation

ISBN

978-91-8103-030-3

Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 5488

Utgivare

Chalmers

EB, Hörsalsvägen 11, Chalmers

Online

Opponent: Prof. Gianluca Palli, University of Bologna, Italy

Mer information

Senast uppdaterat

2024-03-21