SWITZERLAND: A robot developed by researchers in Switzerland has been 'taught' how to grasp a variety of objects with complex shapes and trajectories in less than five hundredths of a second.

Record ID: 401928

  • Title: SWITZERLAND: A robot developed by researchers in Switzerland has been 'taught' how to grasp a variety of objects with complex shapes and trajectories in less than five hundredths of a second.
  • Date: 12th May 2014
  • Summary: LAUSANNE, SWITZERLAND (RECENT) (REUTERS) (SOUNDBITE) (English) SEUNGSU KIM, RESEARCHER, EPFL, SAYING: "First thing is to predict the whole trajectory. Second thing is finding best catching posture. And third thing is generating arm motion."
  • Embargoed: 27th May 2014 13:00
  • Keywords:
  • Location: Switzerland
  • Country: Switzerland
  • Topics: Technology,Space
  • Reuters ID: LVAB8JE65J11V17QH43T73OQU66L
  • Story Text: Swiss-based researchers have developed a robot arm capable of grasping moving objects with complex shapes and trajectories in less than five-hundredths of a second. The team from technology research institute EPFL (�ole polytechnique f��ale de Lausanne) says their research has potential application in space, developing technologies for the recovery and disposal of space debris orbiting the Earth.
    Originally designed by robotics manufacturers Simlab, the KUKA arm, around 1.5 metres in length, has seven joints and a sophisticated hand with four fingers. It was originally designed for use in machine assembly factories but was programmed at the Learning Algorithms and Systems Laboratory at EPFL (LASA) to test robotic solutions for capturing moving objects.
    With its palm open the robot is completely motionless, yet within a split second the arm can twist to catch a variety of flying objects thrown in its direction, such as a tennis racket, ball, and a partially filled water bottle.
    Researcher Seungsu Kim says the research involved three distinct phases.
    "First thing is to predict the whole trajectory. Second thing is finding best catching posture. And third thing is generating arm motion," said Kim.
    The researchers were inspired by the way humans learn by imitation and trial and error. Rather than giving specific directions to the robot, this technique called Programming by Demonstration, involves showing examples of possible trajectories to the robot. The arm was then guided manually to the projected target repeatedly.
    Research was conducted with a ball, hammer, tennis racket, an empty bottle, and a half filled bottle. In the case of the latter, because the liquid moves around unpredictably the object's centre of gravity is constantly moving, making grasping even more difficult.
    In the first learning phase objects were thrown several times in the robot's direction. Through a series of cameras located all around it, the robot created a model for the objects' kinetics based on their trajectories, speeds and rotational movement. Researchers then translated it into a set of equations which allowed the robot to position itself quickly in the right direction whenever an object was thrown. During the few milliseconds of the approach, the machine can refine and correct its trajectory.
    "We teach the robot how to reach towards an object, from many different directions, and from that same time we teach it the co-ordination it needs to have between the arm and the fingers to be able to successfully catch the object and we gather the data, we learn models, which help the robot predict the optimal motion which will be successful in catching this object for reaching towards it and also closing the fingers around at the correct time," said researcher Ashwini Shukla.
    Study author Professor Aude Billard says the LASA research is unique in giving a robot the ability to catch projectiles of various irregular shapes in less than five hundredths of a second.
    "This capacity to predict accurately very complex trajectory of what we call the grasping point, which is non linear, that is the major scientific advance. And similarly to be able to control in synchrony the arm of the robot and also the fingers is another major advance," said Billard.
    According to Billard, the robot's reflexes are as fast as the blink of a human eye. "So the speed we can send information and we can get back information to localize the robot and localize the position is crucial, both of which are running at the level of the millisecond, which is as fast as blinking would be," she said.
    The project was funded by European Union research commission projects AMARSI and First-MM. It was developed in conjunction with the Clean-mE project undertaken by the Swiss Space Center at EPFL, which aims to develop technologies for the recovery and disposal of space debris orbiting around Earth. Fitted on a satellite, the arm would be given the task of catching flying debris.
    According to Billard, "Assume now that the robot is mounted on a satellite and it's tracking the debris and as it's observing this junk rotating then it make inference as to what will be the next translation of velocity, so where this debris will be moving next and adapt its orientation to put the position of its arm so it can grab it and bring it back down to Earth."
    The research was published on May 12 by IEEE Transactions on Robotics, the leading robotic science journal.
  • Copyright Holder: REUTERS
  • Copyright Notice: (c) Copyright Thomson Reuters 2014. Open For Restrictions - http://about.reuters.com/fulllegal.asp
  • Usage Terms/Restrictions: None
Register