The soft-sensitive finger robotic gripper built at MIT can handle cables with unparalleled dexterity.
Manipulating small , flexible items like chains, wires, or cables can be difficult for humans. But if these problems are difficult for humans, then for robots they are almost impossible. As a cable slides between the fingers, its form changes constantly, and the fingers of the robot must constantly feel and modify the location and motion of the cable.
A sequence of gradual and incremental deformations, as well as mechanical fixtures, have been used by traditional methods to get the work done. Recently, the challenge has been tackled from a different perspective by a group of researchers from the Computer Science and Artificial Intelligence Laboratory (CSAIL) of MIT and the Department of Mechanical Engineering of MIT, in a way that mimics us more closely. To successfully handle freely moving wires, the team ‘s latest device uses a pair of soft robotic grippers with high-resolution tactile sensors (and no added mechanical constraints).
For both industrial and household tasks, one might imagine using a device like this to one day allow robots to assist us with things like tying knots, wire forming, or even surgical suturing.
The first move the team took was to create a new two-fingered gripper. The opposing fingers are lightweight and easy to move, enabling nimble, real-time force and position changes. Vision-based “GelSight” sensors, made from soft rubber with integrated cameras, are on the tips of the fingers. The gripper is mounted on a robot arm that, as part of the control system, can rotate.
The second move for the team was to create a system for perception and control to allow cable manipulation. To estimate the location of the cable between the fingers and to measure the frictional forces as the cable slides, they used the GelSight sensors for perception. Two controls work in parallel: one modulates the pressure of the grip, while the other changes the position of the gripper to maintain the cable within the gripper.
The gripper will reliably follow a USB cable when placed on the arm, starting from a random grip position. Then, the robot can pass the cable “hand over hand” (as a person would) in conjunction with a second gripper in order to locate the end of the cable. It could also adjust to various materials and thicknesses of cables.
The robot conducted an action that humans regularly do while plugging earbuds into a mobile phone as a further demo of its prowess. The robot was able to slip the cable between its fingertips, stop when it felt the plug touch its fingertips, change the posture of the plug, and eventually insert the plug into the jack, beginning with a free-floating earbud cable.
In our everyday lives, manipulating soft objects is so popular, such as cable manipulation, fabric folding, and string knotting, “says Yu She, postdoc at MIT and lead author on a new system paper.” “We would like to have robots help humans do this kind of work in several instances , especially when the tasks are repetitive, boring, or dangerous.”
String me along
For two factors, cable follow-up is difficult. First, the “grasp force” (to allow smooth sliding) and the “grasp pose” (to prevent the cable from falling from the fingers of the gripper) need to be managed.
During continuous manipulation, this information is difficult to capture from traditional vision systems, since it is generally hidden, costly to decode, and often unreliable.
What is more, with only vision sensors, this detail can not be explicitly observed, hence the use of tactile sensors by the team. The joints of the gripper are flexible too, shielding them from possible effects.
It is also possible to generalize the algorithms to various cables with different physical properties, such as content, stiffness, and diameter, as well as to those at different speeds.
The team found that, because of the convex surface of the GelSight sensor, it was hard to pull the cable back when it hit the edge of the finger. Therefore, to increase the overall performance, they hope to enhance the finger-sensor shape.
They expect to research more complex cable manipulation tasks in the future, such as cable routing and cable insertion by barriers, and ultimately want to investigate autonomous cable manipulation tasks in the automotive industry.
No CommentsLeave a comment Cancel