Intuitive visual definition of part aligned coordinate systems and mo-tions for industrial robots

Abstract:
Apart from unconstrained free space motions, robot motions often are related to geometrical features like edges, surfaces and contours of the objects to manipulate. Unfortunately, robots do not know the pose of these features, and conventional methods to acquire them are cumbersome and time consuming. We show that if the robot is equipped with vision sensors, the feature coordinate systems can be taught by the instructor in a convenient and fast point and click manner. We further ease the extraction of relevant features by machine vision algorithms. For evaluation we experimentally compare our approach against teach pendant programming on a Universal Robot UR10 and obtain significant time savings for novice as well as experienced robot programmers.