The use of tools has long been a hallmark of human intelligence, as well as a practical problem for solving a wide range of robotic applications. But the machines are still unwilling to exert the right amount of force to control tools that are not tightly attached to their hands.
To manipulate said tools more powerfully, researchers from the Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence (CSAIL) Laboratory, in collaboration with the Toyota Research Institute (TRI), have designed a system that can accommodate the tools and apply the appropriate amount of force for a particular task, such as scanning a liquid or writing a word By.
The system, called Series Elastic End Effectors, or SEED, uses soft bubble clutches and built-in cams to map how the two grippers deform over a hexagonal space (think of inflating and deflating an airbag) and applying force to the tool. With six degrees of freedom, the object can be moved left and right, up or down, back and forth, roll, pitch, and yaw. Closed-loop controller – a self-regulating system that maintains the desired state without it human interaction—Uses SEED and visuoactile feedback to adjust the position of the robot arm in order to apply the desired force.
This may be useful, for example, for someone who uses tools when there is uncertainty in the height of the table, because the preprogrammed path may miss the table altogether. “We were relying heavily on Mason, Raybert and Craig’s work on what we call the hybrid force position controller,” says Hyung Joo Suh, Ph.D. student in electrical engineering and Computer Science at MIT, a subsidiary of CSAIL, and lead author of a new research paper on SEED. “Here’s the idea, that if you really have three dimensions to move in while you’re writing on the board, you want to be able to control the position on some axes, while controlling the force on the other axis.”
Only hard-bodied robots and their counterparts can take us so far; Smoothness and compliance provide luxury and the ability to deform, sensing the interaction between the tool and the hand.
With SEED, every execution the robot senses is a fresh 3D image of the grippers, thus tracking in real time how the grippers change shape around an object. These images are used to reconstruct the position of the tool, and the robot uses a learned model to map the position of the tool to the measured force. The learned model is obtained using the robot’s previous experience, perturbing the torque sensor to see how hard the bubble clutch is. Now, once the robot senses force, it will compare that to the force the user is commanding, and might say to itself, “It turns out the force I’m feeling right now isn’t quite there. I need to push harder.” Then it moves in the direction to increase the force, and it’s all done on a 6D space.
During the “squeegee mission,” SEED was fitted with just the right amount of power to wipe some fluid on the plane, as baseline methods struggled to get the scan right. When asked to put the paper on the pen, the robot effectively wrote the phrase “MIT,” and was also able to apply the appropriate amount of force to drive the screw.
While SEED was aware of the fact that it needed to control force or torque for a particular task, if it was grasped tightly, the item would inevitably slip, so there is an upper limit to that hardness being exerted. Also, if you are a hard robot, you can simulate systems that are softer than natural mechanical hardness – but not the other way around.
Currently, the system assumes a very specific geometry of the tools: they have to be cylindrical, and there are still many limitations on how they generalize when they conform to new types of shapes. Upcoming work may involve generalizing the framework to various forms so that it can handle arbitrary tools in the wild.
“No one would be surprised that compliance can help with the use of tools, or that force sensing is a good idea; the question here is where compliance should go on a bot and how soft that should be,” says research co-author Ross Tedrick, Toyota Professor of Electrical Engineering, Computer Science, Aeronautics, Aerospace, and Mechanical Engineering at MIT and Principal Investigator at CSAIL. “Here we explore very soft stiffness regulation of six degrees of freedom directly on the hand/tool interface, and show that there are some great advantages to doing so.”
HJ Terry Suh et al, SEED: A series of flexible end-effectors in 6D for Visuoactile tool use. arXiv: 2111.01376v1 [cs.RO]And the arxiv.org/abs/2111.01376
Submitted by the MIT Laboratory for Computer Science and Artificial Intelligence
the quote: Soft robots that hold the right amount of strength (2022, September 23) Retrieved on September 23, 2022 from https://techxplore.com/news/2022-09-soft-robots-amount.html
This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.