TAKING HUMAN-ROBOT-COLLABORATION TO A NEW LEVEL
Too dirty, too dusty, too dull? Too hot or too cold? Not for cobots! They can work tirelessly in almost any work environment from production halls and canteen kitchens to clean rooms, freeing up their human colleagues from monotonous, ergonomically disadvantageous tasks. Scientists at Technische Universität Bergakademie Freiberg (Freiberg University of Mining and Technology) in Germany are currently working on two exciting projects to find new ways of how cobots can make the lives of their human colleagues easier: One is testing how cobots can learn from humans only by watching them performing a task. The second one is exploring in which ways robots can serve in the dangerous work environment of underground mining.
THE ROBOT THAT PLAYS WITH LEGO: WHY THE UR5 ROBOT BUILDS A TOY ROCKET AT TU FREIBERG
David Vogt, robot researcher at TU Freiberg and father of two boys, was playing Legos with his kids when the question came to his mind: Can you teach a robot to play Lego? He was curious to find out and started testing, with the university-owned UR5 robot serving as a playmate. The UR5 cobot is equipped with a 3-Finger Gripper and twelve cameras. The scientists on the other hand wear special sensors on their arms and hands so that the UR5 can monitor all of their movements and learn exactly like a child: Simply by watching.
Vogt wanted the UR5 to build the same Lego rocket he made with his son, which requires a high level of dexterity. No problem for the UR5: After the robot had watched Vogt building the Lego rocket just one time, he was already able to rebuild it. Besides, the cobot doesn’t just put Lego block on Lego block – he also tries to analyze what his human colleagues are doing and how he can help them in the best way. For example; when the robot arm sees that the scientist wants to add the next building block, he hands him the suitable one. Vogt and his colleagues are very happy with their success - so far, this is one of the few robots worldwide that have learned only by watching humans. Developing this technology further means that eventually, cobots don’t have to be programmed before starting a new job but can simply learn by seeing – which will make their use, e.g. in industrial production, much easier and cheaper in the future.