Using Partial Sensor Information to Orient Parts

Abstract

Parts orienting, the process of bringing parts in initially unknown orientations to a goal orientation, is an important aspect of automated assembly. The most common industrial orienting systems are vibratory bowl feeders, which use the shape and mass properties of parts to orient them. Bowl feeders rely on a sequence of mechanical operations and typically do not use sensors. In this paper, we describe the use of partial information sensors along with a sequence of pushing operations to eliminate uncertainty in the orientations of parts. We characterize the shorter execution lengths of sensor-based plans and show that sensor-based plans are more powerful than sensorless plans in that they can bring a larger class of parts to distinct orientations. We characterize the relation among part shape, orientability, and recognizability to identify conditions under which a single plan can orient and recognize multiple part shapes. Although part shape determines the results of the actions and the sensed information, we establish that differences in part shape do not always lead to differences in part behavior. We show that for any convex polygon, there exists an infinite set of nonsimilar convex polygons that behave identically under linear normal pushes. Furthermore, there exists an infinite set of nonsimilar convex polygons whose behavior cannot be distinguished even with diameter sensing after each push. We have implemented several planners and demonstrated generated plans in experiments.