Abstract
In assembly process visual features of a target object often occluded by not only surrounding objects but also the object itself or the robot hand. In this paper we propose planning functions of visual measurement to solve this problem. The planning system selects features which are appropriate for the measurement purpose using the object model and the vision sensor model. Under planning the system uses active measurement strategies such as (1) if the target object's features are not available, the scope of candidate feature is expanded to featuses of the objects which rigidly connect to the original object or the robot hand, (2) if needed the system selects the best hand pose to measure the ripped object.
These planning functions are implemented in a hand-eye system which uses light plane intersecting method. Measurement experiments showed effectiveness of the methods.
