Parts orienting, the process of bringing parts in initially unknown orientations to a goal orientation, is an important aspect of automated assembly. Bowl feeders used in industry rely on a sequence of mechanical operations, without using sensors, to orient parts. In our work, we use partial information sensors along with mechanical operations to eliminate uncertainty in part orientation. We show that sensorbased orienting plans need Om operations, where mis the maximum number of states with the same sensor value. We characterize the relation between part shape, orientability, and recognizability to identify conditions under which a single plan can orient and recognize multiple part shapes. We describe implemented planners and experiments to demonstrate generated plans.
Srinivas Akella, Matthew T. Mason