Hybrid Hand-Eye Coordination with a Single Stationary Camera

Abstract

Systems that use visual information for manipulator control usually fall into one of two groups: Classical open-loop methods which localise objects in video images and then guide the robot accordingly, or visual servo systems that close the control loop with a continuous feedback of visual information. We propose a hybrid hand-eye system that combines both methods in such a way that the required accuracy of the subsystems can be reduced, an approach which is inspired by the human "hand-eye system". 1 Introduction In the last years, using sensor information to control robots has become a very popular field of research, as it promises to lead to the design of autonomous robots, which, in contrast to their preprogrammed industrial counterparts, can deal with "the unexpected", for example obstacles or misplaced objects. This is especially important for "personal robots" that operate in the everyday world, that cannot and should not be adapted to the needs of machines. When it comes to se...