New Robot Arm Can Mimic Human Movement and Catch Objects Thrown Its Way

The humanization of robots is becoming more and more advanced. Researchers from the Learning Algorithms and Systems Laboratory at the École polytechnique fédérale de Lausanne (EPFL) in Switzerland revealed their unique robot yesterday.

Researchers demonstrated that their new iCub humanoid robot possesses the ability to catch objects thrown its way, including water bottles, tennis rackets, balls, and hammers. Seungsu Kim, Ashwini Shukla, and Aude Billard had their study published on Monday in the IEEE Transactions on Robotics.

First, they taught the robot how to predict a moving object, and then they taught it where to grasp and how to move its arm. They key was the way the researchers transferred information from a human to the robot through programming-by-demonstration. The robot can simply observe the task from a human and also observe the object flying in the air multiple times so they can build a model of it and then use the model to catch the object.

The researchers used the dynamical systems approach to encode motion from several demonstrations, which enables the robot to develop a reactive arm motion in the event of sensor uncertainty. In order for the robot to catch an object in flight, the researchers were faced with three obstacles requiring solutions: accurate prediction of the trajectory of the objects; predicting the optimal catching configuration (intercept point); and fast planning of precise trajectories.

"Here you're seeing an arm robot, but in course [of time] it will be a car, and a car needs to be able to react very rapidly to oncoming traffic," said Aude Billard in this EPFL video via Live Science. "So this is also one of our examples of reacting and catching or avoiding things extremely rapidly."

The researchers chose the objects for the robot with purpose: a ball, an empty bottle, a half full bottle, a hammer, and a tennis racket. They did this because each object offers a varied range of situations in terms of flight, motion, and shape when thrown in the direction of the arm. By translating trajectories, speeds, and rotational movements into an equation, the robot can position itself quickly to react to and catch the object.

You can read more about the EPFL's new robotic arm in this news release.