A new robot can take and hold objects of any shape with 99% accuracy.
Take a moment and think about all the things you took today in hands. Perhaps it was the keys, a mug, a toothbrush, fork or spoon.
Now think about how much mental energy you spend on that right to take these items. Most likely, this is not for you easily. However, for robots it is a serious problem. But now, thanks to new work by researchers from the University of California, Berkeley, robots can acquire human dexterity.
People instinctively know the best way to grasp the subject, so he did not fall, but robots do not have this ability. Therefore, the researchers used the concept of deep learning, to help two manipulators successfully take objects of any shape with 99% accuracy. Deep learning is a type of automated learning, in which the computer receives a large amount of data and learning to make decisions, processing new incoming information.
In this study, the researchers created a database of contact points for 10,000 3D models, which in total contained about 6.7 million points. Then this information was used to create the neural network system in which a computer makes decisions in the same way as our brain processes information. The system was connected to two manipulators and a set of sensors. The researchers described the installation DexNet 2.0.
The sensor looks at every object in front of him, and a neural network selects the best setting for capture of the object. The system not only fulfilled every engagement, but makes it three times faster than the previous version.
© 2017, paradox. All rights reserved.
