A new robot, developed by researchers at the Learning Algorithms and System Laboratory (LASA) at the Swiss technology institute École Polytechnique Fédérale de Lausanne (EPFL), is able to do something that no robot before it could: catch fast-moving objects out of midair.
“Programmed By Demonstration”
Comprised of a lightweight robot arm from the German manufacturer KUKA and an Allegro Hand from SimLab, the new robot is capable of adapting to changes in the path of a flying object and predict its path to catch it. The robot’s reaction time is less than five hundredths of a second—faster than you can blink.
To catch flying objects (in the video below, these include a ball, a half-full bottle, and a tennis racket) the robot must react and move quicker than traditional programmed calculations can be made. Fortunately, the KUKA robot arm can be “programmed by demonstration”—similar to how humans learn new tasks.
Instead of creating lines of code that give the robot specific instructions, LASA researchers physically guided it through the motion of catching the objects, allowing the system to “build” its own programming. After this “programming” is complete, cameras on the robot view the object in flight and apply the necessary information to movement, allowing the robot to move into position to catch the objects on its own.
Video courtesy of EPFL News.
Real-World (And Outer Space) Applications
Though it may seem like a novelty, the robot arm has a number of potential applications. A robot that can adjust to changing environments and determine its own movements in a fraction of a second could be used to catch someone before he falls, or pluck a falling object out of the air before it lands on someone or something below. A modified version of the robot could be used in automobiles to help drivers avoid collisions.
Currently, the EPFL’s research is being put to use in a project to develop technology for recovering and disposing of space junk. Similar robotic arms could be used to catch orbiting debris before it hits satellites or spacecraft.
LASA’s research was published in the 12 May 2014 edition of IEEE Transactions on Robotics.