Researcher from University of California has designed algorithms that allow robots to perform tasks without pre-programmed details.
The algorithm, known as BRETT, is a kind of reinforcement technique. It enable robot to learn through trial and errors. For instance, robot can effortlessly assemble plan toy.
Trevor Darre, co-researcher of the study, discusses the experiment in detail. He notifies that it is pretty difficult for robots to work in a real-life setting. The newly developed algorithm permit robot to perceive and adapt things from its surroundings.
Researchers carefully studied a branch of artificial intelligence to design the algorithm. The programs developed through branch, known as deep learning, form layers of artificial neurons. The neurons can be easily created through sound waves or image pixels.
It took robot around 12 minutes to properly assemble toy airplane. Afterwards, scientists gave lego bricks to robot for the same purpose. Astonishingly, robot successfully assembled the bricks with the help of same algorithm. In the lego bricks task, robots put together the bricks in only few minutes
Scientists are expected to present the end result of the experiment at the International Conference on Robotics and Automation (ICRA). The conference is scheduled to take place on 28th of May in Seattle.