Procreation is one capability robots don't have--yet--but their numbers are on the rise. In fact the North American robotics industry had its best year in 1997, according to the Robotic Industries Association (RIA) trade group. At $1.1 billion, for the first time shipment revenues bested the billion-dollar benchmark, which represented 12,459 robots shipped, 28% greater than the previous year. Over the last five years surveyed, the number of new robots sent to industry each year has jumped to nearly 2 3/4 times the deliveries in 1992 (4,561 units).
Donald Vincent, RIA executive vice president, notes the growth in robotics materials handling applications, lessening the industry's dependence on the automobile industry, heretofore its leading customer. "Materials handling, such as parts transfer, machine tending, packaging, and palletizing, cuts across many industries," he explains. "Manufacturers of consumer goods, electronics, food and beverages, and other non-automotive products are now taking advantage of robots to become stronger global competitors." For those companies investigating robotics, Vincent adds the RIA has two new resources: its 1998 Robotics Industry Directory providing information on more than 125 suppliers of robots and automation products; and A Guide to Robot Solutions videotape.
Vincent says PC-based control and programming is a growing area of robotics technology. "And users want help in completely simulating robotic operations before installation, so that no surprises occur. More and more end users are turning to robotics systems integrators." Their expertise allows optimizing systems by using equipment from many sources. (Refer to cutting edge story page 91.) "Interest is also widespread in tying-in any robotic systems with an overall computer-aided manufacturing operation."
Be thou my vision. Machine perception is one area key to robotics applications growth. Vincent notes some small parts assembly can be done with today's technology. "But there is a lot of work to do in the next few years in developing, for example, the ability to permit two robot arms to work together in assembly. Technology integration and vision sensors will be critical in, say, allowing a workpiece to be properly positioned for a robot to work on it."
One such development is underway at the University of Notre Dame (Notre Dame, IN). Aerospace and mechanical engineering professor Steven Skaar is investigating robot "hand-eye coordination" using both uncalibrated video and a laser. Here a test facility features a trio of video cameras directed toward a work platform and linked to a PC. The computer controls a six-axis robot arm next to the platform.
One camera views an object on the platform. Using a mouse, the operator designates a spot on the PC-displayed image as a target for the arm and keys in the final end-effector orientation at the target. The position of the target spot is one of three 2D "camera space targets" for the robot arm's tool tip. The only requirement is the remaining two widely separated cameras must each view the target within their own 2D frames. To aid in target location, a laser pointer projects an array of light pinpoints onto the target surface.
The cameras track arm motion, providing highly redundant visual information to refine locally the arm joint-pose sequence that will eventually satisfy each individual camera's objectives. Controlling the maneuver in the 2D image plane of each camera results in the 3D positioning of the tool tip at the right point. Ring shapes are painted at various junctures on the arm end effector for easy detection via computer vision during approach. Their cues ensure correct information is factored into the joint-pose sequence to give the desired end-effector orientation as well as position.
By developing the ability to reach for target objects in the camera reference frames, calibrating robots and cameras can be eliminated. "The need to use expensive jigs and fixtures to position each workpiece on a production line is also eliminated," says Skaar. Much like reaching for a doorknob, "Our system works the way humans see --with no reference to a calibrated frame of reference and all real-time processing. You can place the surface of interest in front of the robot in any orientation."
The Notre Dame system has carved patterns in wood and fit a tire rim onto a five-lug wheel. "Holes are drilled to 1/4-mm precision with this large robot," adds Skaar, "and it's like a drill press in terms of straightness and accuracy of the hole. Accuracy is limited by only camera and pixel resolution." A longer lens allows greater accuracy. "Since nothing is calibrated, we can move everything and not lose precision," he adds.
"In the future, for fully autonomous tasks, the human operator would be replaced in deciding where in the selection camera field key junctures (positions) on the surface are located," concludes Skaar. "Depending on workpiece geometry, the same array of pinpoints for automatic location might be used to do image analysis on many types of surfaces and geometries."
But no matter what advanced robotic systems are developed, Donald Vincent emphasizes, "The first reaction is to look at return on investment." Not withstanding system capabilities, if the payback period is, perhaps typical of the auto industry, longer than one to two years, it may not be justified in the eyes of management. "It's always risky trying to justify a new productivity tool for the first time. But department- and plant-level managers are buying into the technology that's seen in other industries--after their applications are analyzed to be sure robotics will work." Thus, having progressed from the auto industry into packaging and palletizing, robotic systems are poised to become more widespread beyond.
For more information on the Robotic Industries Association resources:
University of Notre Dame robot guidance: Circle 558
See a video clip about