SEATTLE (Diya TV) — A team of computer scientists from the University of Washington have built a robot hand that can not only perform dexterous manipulation, but also has the ability to learn from its experiences without the assistance of human direction.
The latest results from the trials have been detailed in a paper, which will be presented during this month’s International Conference on Robotics and Automation.
“Hand manipulation is one of the hardest problems that roboticists have to solve,” said Vikash Kumar, a UW doctoral student in computer science and engineering and lead author of the paper. “A lot of robots today have pretty capable arms but the hand is as simple as a suction cup or maybe a claw or a gripper.”
The research team spent years custom-building what has become one of the world’s most highly capable five-fingered robot hands, which possesses a model that enables a computer to analyze movements in real time. In the latest demonstration, the group of researchers applied the model to real-world tasks like rotating an elongated object.
With each attempt, the robot hand becomes more and more adept at spinning the tube, thanks to algorithms that help it model basic physics involved in the action, and the plan required to achieve desired results. See the robot hand in action in the below video.
The movement was developed at the university’s Movement Control Laboratory, and completely contrasts with robotics demonstrations that require people to program a robot’s individual movements in order to complete a task.
“Usually people look at a motion and try to determine what exactly needs to happen — the pinky needs to move that way, so we’ll put some rules in and try it and if something doesn’t work, oh the middle finger moved too much and the pen tilted, so we’ll try another rule,” said senior author and lab director Emo Todorov.
“It’s almost like making an animated film — it looks real but there was an army of animators tweaking it,” Todorov said. “What we are using is a universal approach that enables the robot to learn from its own movements and requires no tweaking from us.”
This research, development and production of the robot hand of course came at a price — $300,000 is what the university has spent thus far on the dexterous robot hand. It uses a shadow hand skeleton actuated with a custom pneumatic system and can move faster than a human hand. The overall result is far too expensive for commercial or industrial use, but has allowed researchers to push core technologies and test innovative strategies.
“There are a lot of chaotic things going on and collisions happening when you touch an object with different fingers, which is difficult for control algorithms to deal with,” said co-author Sergey Levine, who worked on the project as a postdoctoral fellow at University of California, Berkeley. “The approach we took was quite different from a traditional controls approach.”