ABSTRACT
Many interesting objects, known as articulated objects, fall into this category: boxes, swinging office chairs, refrigerators, and pairs of scissors. This class of objects is of special importance since it includes most industrial robots and man-made factory tools. Furthermore, on the many occasions in daily life when a robot tries to interact intelligently and effectively with its environment, it must not only recognize and locate objects but must understand and describe the situation or status of the objects. Goldberg and Lowe extended Lowe’s system to deal with 3D articulated objects such as staplers. Take the box, for instance. Its representation not only represents a variety of views of the box, its different rotations and sizes, but also different states (e.g., wide open, half open, almost closed, etc.), all in one representation (except when completely closed), with different measurements and range, of course. These measurements and range for interpretation of the model object can be obtained from heuristic learning experience.
