Scientists teach robots to recognize objects the way children learn to play with toys
Researchers at the University of Texas at Dallas have taught a robot called Ramp to recognize different objects by pushing them around.
It may sound simple, but it is a giant step for the world of robotics.
This discovery could one day make robots much smarter and more useful in homes or factories.
The old way of teaching robots to recognize objects was simple but limited. Typically, a robot would push or grab an object once and then try to remember it for the next time.
However, this new research leads to robots behaving more like children learning to play with toys.
Just as a child would push a toy car around many times to understand how it moves, the Ramp robot pushes each object 15 to 20 times.
This allows a lot of images and information to be collected so that the object can be better recognized later.
Why is that important? Let’s say you want a robot to bring you a bottle of water. The robot needs to know what a “bottle of water” looks like.
The new system helps the robot not only recognize one type of water bottle, but also different types.
It could be a large or small bottle, a known brand or an unfamiliar one – the robot would still know it was a water bottle.
Ramp is a special type of robot developed for these experiments.
It is about 1.20 m tall and moves on wheels. It has a long mechanical arm with a “hand” at the end that it uses to push or grab things. The robot uses a special camera that can also capture depth, allowing it to better understand what an object is made of.
The leader of the team, Dr. Yu Xiang says this is the first system to use long-term interaction to help a robot better understand objects.
Put simply, the more the robot interacts with different things, the more intelligent it becomes. By pushing and examining objects multiple times, the robot collects more data to make fewer mistakes later.
One of the students working on the project, Ninad Khargonkar, said that testing the robot in the real world was eye-opening.
Algorithms or fancy computer code is one thing, but seeing the robot actually do its job better in a real environment was an incredible experience.
What’s the future of this? Well, the researchers want to make robots even smarter. You want to teach them to plan and control more complicated tasks, like sorting things for recycling.
They hope these new skills will make robots more useful in a variety of environments, not just the lab.
While we’re still a long way from robots cooking dinner or cleaning the house, this research is a step towards smarter and more powerful robots.
It might not be long before you find a helpful robot to make your life at home or at work easier.
follow us on Twitter for more articles on this topic.
Source: University of Texas at Dallas.
