Automation requires workpieces to be well organized. Current machine feeding technology is dominated by the use of human labor, mechanical feeders, or orientation preservation. All these techniques have problems and limitations that make an alternate solution desirable. Three vision algorithms are presented which enable a computer-controlled robot to acquire a single piece from a bin containing randomly placed identical pieces. The algorithms do not try to solve the problem by identifying the position and orientation of a piece in the bin. Rather, the algorithms recognize where there is a section of any piece which a particular type of gripper would be able to grasp with a reasonable chance of success. Thus the algorithms are holdsite driven and depend on the piece and gripper type. Vacuum cup and parallel-jaw gripper types are treated. The heuristically defined holdsites for vacuum cup grippers are patches of smooth surfaces; for parallel-jaw grippers, holdsites are opposing, linear, or curvilinear parallel edges. The three-dimensional position of holdsites cannot be computed from a single image of the bin. Thus once a hold-sight is identified, the gripper is made to travel along a line-of-sight path. The missing degree of freedom, distance from the holdsite to the camera, is obtained by a contact or noncontact proximity sensor in the gripper. Acquisition requires two other sensors: a grasping sensor to indicate success; and a mechanical overload sensor, to abort the attempt in case of error.