Don’t waste your time on trying to solve robotic manipulation! Robotic manipulation has been solved in 1968. It has been solved when the Unimate robot demonstrated how a robot could butter a toast, pour coffee, set a table, open a fridge and open a bottle of champagne. Look around on the internet, and you would not find a better video of a robot doing something with its hands. Indeed, robotic manipulation seems to have gotten much, much harder over the years. Looking closely at more recent videos, it seems it is not only difficult to make a robot grasp something without running into the object or table, topple the object due to accidental contact, make the object literally avoid the grasp by jetting around the table upon contact, or reliably holding it while the robot moves.
2016 robots are shaky and barely get a handle on the objects they wish to manipulate. So what’s the difference? Of course, motions in the 1968 video were pre-programmed and getting the video has probably taken many, many shots. But why is it so much harder to localize an object accurately and replay the correct motion then? We have shown that it is actually not too much different when equipped with the right sensors, finally allowing the robot to react to objects in unstructured environments, leading to videos that show shakey robots, but showing robots that work almost every time a video is taken.
Robotic Materials sensors provide the missing link between existing perception systems and (not so) accurate robotic manipulators. Proximity allows the robot to make up for inaccuracies of perception and manipulation, contact allows the robot to validate pose without exerting forces, and pressure allows the robot to confirm the object pose relative to its body. Don’t believe it? Get your evaluation kit now, try our solutions for Rethink Robotics, the – much less shaky – Kinova robots, or contact us for a custom solution to solve your manipulation problem.