סמינר מחלקה של איתמר משעני - מניפולציה של כבלים באמצעות רובוט דו-זרועי על ידי חישה של כוחות ומומנטים

School of Mechanical Engineering Seminar
Wednesday, May 18, 2022 at 14.00
Wolfson Building of Mechanical Engineering, Room 206 [IM1] 

DUAL-ARM MANIPULATION OF WIRES USING SOLELY FORCE/TORQUE FEEDBACK

Itamar Mishani

MSc student of Avishai Sintov

School of Mechanical Engineering

Manipulation of wires has been a challenging task and main interest for many decades. There have been many attempts to use visual perception and image segmentation to perform wire manipulation. However, manipulate in a cluttered environment where there are many visual occlusions and uncertainties (due to poor lighting or shadows) is hard. Giving a robot the ability to manipulate wire with high confident is necessary and requires rapid reasoning of its shape in real-time. Furthermore, after having this ability, plan and control wire manipulations is required, however, there is no efficient ability to do so without visual perception. Recent work has shown that the shape of an elastic wire can be defined by a very simple representation. This representation can be interpreted as forces and torques at one end of the wire. To begin with, we experimentally analyzed the theoretical foundation. We deployed a dual-arm robotic system able to accurately manipulate an elastic wire. The system does not require complex visual perception and is able to reason about the shape of the wire by solely sensing forces and torques on one arm. Furthermore, we proposed a full framework in which the mechanical properties of the wire are rapidly approximated in real-time. Then, a simple control rule based on Force/Torque (F/T) feedback is used to manipulate the wire to some goal or track a planned path. However, the model used to develop the system relies on assumptions that may not be met in real-world wires and does not take gravity or the non-linearity of the F/T sensor into account. Therefore, the model cannot be applied to any wire with accurate shape estimation. Additionally, solving the non-linear non-convex inverse problem, i.e., from wire shape to F/T, is computationally expensive. Therefore, we investigated the learning of a model to estimate the shape of a wire solely from measurements of the F/T state. We propose to train a novel learning model which can both act as a descriptor of the wire where F/T states can be mapped to its shape and as a solver of the inverse problem where a desired goal shape can be mapped to an F/T state. Additionally, we trained a different model, with the same dataset, which gives the robot the ability to execute a planned path.

Join Zoom Meeting

 https://us02web.zoom.us/j/82108132163?pwd=Z2h4UzNzUS9mbXplT0lMU1pZenFEQT09