To test the algorithm, the researchers fitted the jamming gripper and a Microsoft Kinect 3D camera onto an industrial robot arm. The robot was tested in attempts to pick up 23 different objects, including toys, tools, and dishes. In these tests, the robot's success rate averaged 90 percent to 100 percent.
In most cases, the robot arm could successfully grasp new objects that it had not reviewed during its training. When the team ran the same tests with a simple directive to pick up an object at its center, the robot arm scored only had a 30 percent to 50 percent success rate. The exception was in picking up flat objects. With these, both the learned grasps and simple center grasps tied at an 89 percent success rate. The algorithm was also tested with the standard parallel jaws that most modern robots use, which produced similar results.
The team, which includes graduate students Yun Jiang and John Amend, presented their results May 16 in a paper at the International Conference on Robotics and Automation in St. Paul, Minn.
Ann, this might mark me out as a bit wierd, but I think about this a lot. Whenever I put the silverware away I thnk to myself, how would I program a robot to do this?
What really strikes me about this, and some other situations I have seen, is that people are programming robots to do things using a fairly simple vision system along with memory (a database) and an algorithm. This contrasts with robotics approaches that use all kinds of complex sensors. In many cases they are trying to automate something we do with our simple sensors naturally. Interesting.
Picking up an object is only part of the problem. The picture shows a gripper spilling a glass of water. After the object is grasped, some purpose must be accomplished. If the water were wine and needed to go from a pitcher into a glass, it would be inportant not to spill it onto the floor or table, and that the robot's 'fingers' not get into the wine. While this is an interesting line of research, I can't see it replacing purpose-built grippers yet.
naperlou, not everyone thinks about how a robot would do things they themselves are doing. But that does sound like how engineers think. Thanks for the observation about the lack of sensors here--I think that's a good point, and it's interesting to know this isn't the only research team taking that approach.
Glenn, thanks for that observation about the photo. I should have pointed out in the caption that this universal gripper, without the algorithm, can pick up objects but that this shows how it does so in a non-optimal manner, forming a "before" picture.
Ann R Thryft; Yes, optimal vs. non-optimal is the clarification. For some applications the optimal gripper is vacuum cup(s). The human hand is a very versatile end effector. Duplicating it is not easy. There could be applications where this gripper would be optimal, but I don't think the water glass is one of them.
Right now this looks like a technology development seeking a solution. As the robots develop, solutions will appear. I've seen this notion of robots learning how to do things by trial and error. That's impressive.
The point here is that, with a less expensive universal gripper, such as Cornell's, plus the algorithm the team invented, a robotic assembly line can quickly adapt to optimally picking up all kinds of new objects with different sizes and shapes that it's never encountered before. The alternative, which we've heard a lot about in DN articles and comments, is lengthy and expensive programming in 4D, presumably with highly specialized grippers. This would be a big benefit in assembly lines, especially those of EMS, which are continually changing products.
I have had a bit of experience with assembly lines. I can't think of any application for this gripper. Printed circuit board assembly needs very fast small part placement with vision compensation, or fast very fine placement of large parts with many leads, using vision compensation. I have only seen vacuum nozzles used. I can't see this gripper being used in a high-speed vision application. In automotive speed, accuracy, and payload are important. I don't think this gripper has any of these 3. Even where I have seen off-line programming using 3-D modeling, an actual human had to step through the program to touch-up positions and movements. Robots, aka Flexible Automation, vs. 'hard automation', was the answer to changing products. The gripper or 'end effector' is always customized to the application. The part must be both 'picked' and 'placed'.
To belabor the point: I don't think this gripper could pick up a 1mm x 2mm chip, take a vision shot, and then place it into a solder screened location, and do it again 1/10 second later. I also doubt that it could pick up a 50 lb bag of flour and place it to a pallet.
The gripper and the algorithm are interesting research, without a current practical application.
I really liked the article. I don't know if I completely understand the inner workings of the pressure adaping inside of the big blue ball, but the statistics of success for picking up parts is pretty cool.
Transfers the control of a large number of motion axes from one numerical control kernel to another within a CNC system, using multiple NCKs, and enables implement control schemes for virtually any type of machine tool.
Industrial trade shows, like Design News' upcoming Pacific Design & Manufacturing, deserve proper planning in order to truly get the most out of them as marketing tools. Here's how to plan effectively.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.