Dexterous Robotic Manipulation of Deformable Objects with Multi-Sensory Feedback - a Review
By Fouad F. Khalil, Pierre Payeur. 2010
84 citations recorded by the source corpus.
Designing autonomous robotic systems able to manipulate deformable objects without human intervention constitutes a challenging area of research. The complexity of interactions between a robot manipulator and a deformable object originates from a wide range of deformation characteristics that have an impact on varying degrees of freedom. Such sophisticated interaction can only take place with the assistance of intelligent multisensory systems that combine vision data with force and tactile measurements. Hence, several issues must be considered at the robotic and sensory levels to develop genuine dexterous robotic manipulators for deformable objects. This chapter presents a thorough examination of the modern concepts developed by the robotic community related to deformable objects grasping and manipulation. Since the convention widely adopted in the literature is often to extend algorithms originally proposed for rigid objects, a comprehensive coverage on the new trends on rigid objects manipulation is initially proposed. State-of-the-art techniques on robotic interaction with deformable objects are then examined and discussed. The chapter proposes a critical evaluation of the manipulation algorithms, the instrumentation systems adopted and the examination of end-effector technologies, including dexterous robotic hands. The motivation for this review is to provide an extensive appreciation of state-of-the-art solutions to help researchers and developers determine the best possible options when designing autonomous robotic systems to interact with deformable objects. Typically in a robotic setup, when robot manipulators are programmed to perform their tasks, they must have a complete knowledge about the exact structure of the manipulated object (shape, surface texture, rigidity) and about its location in the environment (pose). For some of these tasks, the manipulator becomes in contact with the object. Hence, interaction forces and moments are developed and consequently these interaction forces and moments, as well as the position of the end-effector, must be controlled, which leads to the concept of “force controlled manipulation” (Natale, 2003). There are different control strategies used in 28
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Sergey Levine
- Pieter Abbeel
- Thanard Kurutach
- Lerrel Pinto
- Xingyu Lin
- David Held
- Chelsea Finn
- Yufei Wang
Citations in the corpus
- SoftGym: Benchmarking Deep Reinforcement Learning for Deformable Object Manipulation (2020, background): “We refer to [26, 27] for a more detailed survey on prior methods for robot manipulation of deformable objects.”
- Robotic manipulation and sensing of deformable objects in domestic and industrial applications: a survey (2018, background): “Khalil and Payeur (2010) present a very comprehensive survey, where most of the surveyed papers focus on modeling and simulation, and approaches that have been applied to industrial tasks such as assembly and food handling.”
- Learning to Manipulate Deformable Objects without Demonstrations (2019, background): “For a more detailed survey, we refer the reader to [10], [26].”