Abstract
There are various types of visual sensors. When we try to use only one of them as a sensor of an autonomous robot, it may be difficult or take a long time to obtain required information. In such a case, it is useful to use heterogeneous visual sensors whose features are different. Furthermore, we can improve the efficiency not only by fusing data gotten by each sensor but also by modifying the behavior of each using information from others.
We fabricated an active vision system that consists of a range finder and a monocular camera system as a sensor system which combines heterogeneous visual sensors. As an example, we applied it to a parts-picking problem.
