Photo courtesy of Bastian Robotics.
Older robots needed precisely known locations for both picking and placing. Machine vision, or simply “vision,” systems do away with that.
Vision technologies use cameras and software to let the robot “see” what it is doing. They used to be expensive and not very intelligent. Now they are cheap and smart. (You probably have one in your pocket that is smart enough to pick out faces). A vision-guided robot can pick and place parts from and to any location within their reach.
A camera directs this delta robot to pick up a candy on one conveyor and put it in a paper nest on another. Neither candy nor nest is precisely positioned on the moving conveyors.
Until recently virtually all vision systems were two dimensional (2D) and could only work between known, fixed, elevations. Now we have three-dimensional (3D) vision systems. Parts no longer need to be on a flat surface but can be jumbled in a bin, as shown in this video from Fanuc Robotics North America. The camera “sees” the bottle in this example and guides the robot to pick it up. This allows such applications as picking randomly oriented bottles from a bin and setting them upright onto a conveyor. What used to be a purpose-built bottle orienter is now a simple and standard robot, camera and suction cup. They are not fast (yet!) but get faster every day. Even at present speeds, they can free a person from manually placing bottles and let them do something more valuable.
Photo courtesy of Fanuc Robotics North America.