Estimating Model Utility for Deformable Object Manipulation Using Multiarmed Bandit Methods

McConachie, Dale; Berenson, Dmitry

doi:10.1109/tase.2018.2822669

Cited by 28 publications

(19 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Model-based visual servoing approaches bypass planning entirely, and instead use a local controller to determine how to move the robot end-effector for a given task (Hirai and Wada, 2000; Smolen and Patriciu, 2009; Wada et al, 2001). Our recent work (Berenson, 2013; McConachie and Berenson, 2018) as well as that of Navarro-Alarcon and Liu (2018); Navarro-Alarcon et al (2014, 2016) bypasses the need for an explicit deformable object model, instead using approximations of the Jacobian to drive the deformable object to the attractor of the starting state. More recent work by Hu et al (2018) has enabled the use of Gaussian process regression while controlling a deformable object.…”

Section: Related Workmentioning

confidence: 99%

“…Let a robot controller be a function C ( q r , scriptP , scriptT ) , 1 which maps the system state ( q r , scriptP ) and alignment targets T to a desired robot motion q · r cmd . In this work, we restrict our discussion to tasks and controllers of the form introduced in our previous work (Berenson, 2013; McConachie and Berenson, 2018); these controllers are local, i.e., at each time t they choose an incremental movement q · r cmd which reduces the alignment error as much as possible at time t + 1 .…”

Section: Problem Statementmentioning

confidence: 99%

“…The role of the local controller is not to perform the whole task, but rather to refine the configuration of the deformable object locally. For our local controller we use a controller of the form introduced in Berenson (2013) and McConachie and Berenson (2018). These controllers locally minimize error ρ while avoiding robot collision and excessive stretching of the deformable object.…”

Section: Interleaving Planning and Controlmentioning

confidence: 99%

“…Finding the best robot motion Given a desired deformable object velocity P · d and relative importance weights W d , we want to find the robot motion that best achieves ( P · d , W d ) , that is, In general, f (… ) is not known. For our controllers we use a Jacobian-based approximation from McConachie and Berenson (2018: section V-C). Our method for ensuring the robot stays in C r valid is different, depending on which robot we are using.…”

Section: Table Of Multimedia Extensionmentioning

confidence: 99%

“…In addition, the dynamics of the deformable object are difficult to model (Essahbi et al, 2012); even with high-fidelity modeling and simulation, planning for an individual task can take hours (Bai et al, 2016). Local controllers on the other hand are able to very efficiently generate motion, however, they are only able to successfully complete a task when the initial configuration is in the “basin of attraction” of the goal (Berenson, 2013; McConachie and Berenson, 2018).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Manipulating deformable objects by interleaving prediction, planning, and control

McConachie

Dobson

Ruan

et al. 2020

The International Journal of Robotics Research

Self Cite

View full text Add to dashboard Cite

We present a framework for deformable object manipulation that interleaves planning and control, enabling complex manipulation tasks without relying on high-fidelity modeling or simulation. The key question we address is when should we use planning and when should we use control to achieve the task? Planners are designed to find paths through complex configuration spaces, but for highly underactuated systems, such as deformable objects, achieving a specific configuration is very difficult even with high-fidelity models. Conversely, controllers can be designed to achieve specific configurations, but they can be trapped in undesirable local minima owing to obstacles. Our approach consists of three components: (1) a global motion planner to generate gross motion of the deformable object; (2) a local controller for refinement of the configuration of the deformable object; and (3) a novel deadlock prediction algorithm to determine when to use planning versus control. By separating planning from control we are able to use different representations of the deformable object, reducing overall complexity and enabling efficient computation of motion. We provide a detailed proof of probabilistic completeness for our planner, which is valid despite the fact that our system is underactuated and we do not have a steering function. We then demonstrate that our framework is able to successfully perform several manipulation tasks with rope and cloth in simulation, which cannot be performed using either our controller or planner alone. These experiments suggest that our planner can generate paths efficiently, taking under a second on average to find a feasible path in three out of four scenarios. We also show that our framework is effective on a 16-degree-of-freedom physical robot, where reachability and dual-arm constraints make the planning more difficult.

show abstract