Video: 5-fingered robot learns to get a grip on its own

Have you ever thought, a robot hand can perform incredibly dexterous manoeuvres and can learn on its own? Yes, you heard it right!

A team led by Indian – origin computer engineer from the University of Washington has developed a five – fingered robotic hands that can not only perform dexterous manoeuvres but can learn from its own experience without depending on the human.

Vikash Kumar, a doctoral student in computer science and engineering, spent years in building a highly capable five-fingered robot hands in the world. Post this achievement; his team developed a simulation model that enables a computer to analyse movements in real time. In the latest demonstration, they apply the model to the hardware and real- world tasks such as rotating an elongated object.

With each attempt, the hand gets progressively more adept at spinning the tube, because of the machine learning algorithms.

Vikash Kumar said, “ hand manipulation is one of the hardest problems in the roboticists to solve.” “Today a lot of robots have a pretty capable arms, but a hand is as simple as a suction cup or maybe a claw,” he added.

The device cost approximately 300,000 USD and uses a “shadow hand” skeleton with a custom pneumatic system, that can move faster than a human hand.

The research team developed algorithms which allowed a computer to model highly complex behaviour and plan movements to achieve different outcomes like typing on a keyboard or dropping a catch.
Kumar and team will present their work at the IEEE International conference on robotics and automation on Tuesday, may 17.