Engineers design five-fingered robot hand that learns tasks on its own

Advancements in robotics are growing at a rapid pace. These machines are able to perform a wide array of tasks from solving a Rubik’s cube to venturing out into space. However, hand dexterity has been somewhat of a challenge for engineers designing hand-like capabilities. Hand manipulation and movements like rolling, pivoting, bending and sensing friction have been capabilities that were only found in human hands.

Photo courtesy of University of Washington.

However, a University of Washington (UW) team of computer science and engineering researchers has built a robot hand that is able to perform dexterous manipulation as well as learn from its own experience without needing humans to direct it.

“Hand manipulation is one of the hardest problems that roboticists have to solve,” said lead author Vikash Kumar, a UW doctoral student in computer science and engineering. “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 team has been working on this project for years, custom building one of the most highly capable five-fingered robot hands. They developed an accurate simulation model that enables a computer to analyze movements in real time. In their latest demonstration, they apply the model to the hardware and real-world tasks like rotating an elongated object.

One of the unique features of the UW robotic hand is that is able to develop dexterity as it performs tasks. With each attempt, the robot hand gets progressively more adept at spinning the tube, thanks to machine learning algorithms that help it model both the basic physics involved and plan which actions it should take to achieve the desired result.

This autonomous learning approach developed by the UW Movement Control Laboratory contrasts with robotics demonstrations that require people to program each individual movement of the robot’s hand in order to complete a single task.

Design and control were challenges for the team when developing such a device. The team wanted to design the robot hand with similar reactions and abilities as a human hand. The robotic hand uses a Shadow Hand skeleton actuated with a custom pneumatic system and can move faster than a human hand.

The team developed algorithms that allowed a computer to model highly complex five-fingered behaviors and plan movements to achieve different outcomes in simulation.

In real-time, as the robot hand performs different tasks, the system collects data from various sensors and motion capture cameras and employs machine learning algorithms to continually refine and develop more realistic models.

Going forward, the team is hoping to explore global learning where the hand could figure out how to manipulate an unfamiliar object or a new scenario it hasn’t encountered before.

www.washington.edu