Posts
Marc Ambasna-Jones, UK-based freelance writer and media consultant summarizes, "The Massachusetts Institute of Technology (MIT) has got to grips with robotic dexterity, says Prof. Alberto Rodriquez."
 |
| Massachusetts Institute of Technology Photo: Wikipedia, the free encyclopedia |
The penguins in Madagascar: Escape 2 Africa were very pragmatic if nothing else. To rebuild the crashed plane they needed human help. It’s good to know that for the moment at least, we humans still have a value, even in animated fiction. In the real world, things could be a little different.
As robotics evolves rapidly, features of human ability are being consistently copied and challenged. Perhaps we take grip for granted but at MIT in Cambridge, Massachusetts the problem of how to make robotic hands and instruments grip effectively seems to have been solved.
 |
| Photo: Alberto Rodriguez |
We caught up with MIT School of Engineering professor Alberto Rodriguez to find out more.
Q. Are dexterity and agility two of the biggest physical challenges for robotics?
A. From the ones that are physical, dexterity and agility are certainly prominent. But there are many abilities that today pose great challenges for robotics, including cognitive ones such as learning and reasoning. We humans are extremely good at all of them, in a way that might bring them closer to an artistic expression than engineering. We are agile and dexterous without an explicit understanding of what it means to be or how to become.
How complex is the process of creating dexterity in a robotic gripper and what was the biggest challenge?
Replicating dexterity, or hand manipulation in general, requires the combination of many disciplines. It involves the complexities of adaptive behavior, since a broad manipulation capability needs to deal with different task, objects, and environments, and involves the detailed and fine-grained particularities of interacting with the real world.
A manipulation system requires thinking, at the very least, about actuation, control, perception, and materials. Advances in one of them do not necessarily get us too far since they are interconnected. It is by thinking at the interconnection between them that we can solve the problem.
Manipulating an object in the hand, such as you would do when playing with a pen, requires a very complex gripper. Many robot systems, however, have very simple grippers, like the parallel-jaw grippers we see in industry. The key idea in our approach is to by-pass the strong requirements of gripper complexity by outsourcing them to the rest of the robot arm and the environment. We call it extrinsic dexterity, and it takes the form of a robot pushing against the environment to manipulate the grasped object.
How long did it take you to achieve the desired levels of force to make the grippers effective?
One of the interesting twists of this approach is that, by employing extrinsic dexterity rather than intrinsic, the need for gripper accuracy, speed and strength is outsourced to the robot arms, which these days happen to be very accurate, very fast, and very strong. In some cases all three at the same time. So current robots and grippers are already capable of
it.
The main challenge is in modeling the interaction between gripper, object, and environment, with a degree of accuracy so that a robot can decide the best way to interact/push with the environment to reconfigure a part.
Read more...
Source: IDG Connect
