The old saying goes that you can’t teach an old robot new tricks, but this one is certainly bucking the trend by being taught how to catch items – and one day it might even be able to catch people who are in danger.
The robotic arm is already at a disadvantage to other robots as it only has four fingers, however this hasn’t stopped it from learning how to pluck five different objects out of the air when they’re flung in its direction.
Scientists at the Learning Algorithms and Systems Laboratory at EPFL, Switzerland, said the technology could be adapted so it is able to save lives in dangerous situations.
Aude Billard, who is the head of the lab, said: “This controller can be used for many other different applications typically catching people who are in danger of falling, catching objects that may fall onto people.
“That also applies to any controller for a vehicle robot, here you’re seeing in arm robots but it could also be in a car. A car needs to be able to react very rapidly to oncoming traffic.”
The robot has a little way to go before it is saving lives, however, but it has so far been able to catch a ball, an empty bottle, a half-full bottle, a hammer and a tennis racket.
The scientists who taught the robot everything it knows treated the robot like it was a human – by using trial and error as a learning tool.
They do not give the robot directions but show the robot examples of possible trajectories that items may come in at, which involves manually guiding the arm to the projected target and repeating this exercise several times.
In the first instance objects were thrown several times in the robot’s direction and through a series of cameras located around it the robot creates a model for the objects’ kinetics based on their movement.
The scientists then step in and translate it into an equation which then allows the robot to position itself very quickly in the right direction.
In the milliseconds from the object leaving the researcher’s hand the robot refines and corrects the trajectory so it is able to capture the object.
Aude Billard, who worked on the project, said that future robots will need to be able to react in real-time.
Aude said: “Increasingly present in our daily lives and used to perform various tasks, robots will be able to either catch or dodge complex objects in full-motion.”
“Not only do we need machines able to react on the spot, but also to predict the moving object’s dynamics and generate a movement in the opposite direction.”
In total the arm is about 1.5m long and keeps an upright position and it has three joints and a hand with four fingers.
Images courtesy of Learning Algorithms and Systems Laboratory