An industrial robot is strong and fast and great at repeating itself. Give it a task, such as moving a hot welding tip to a certain point, and it will move that tip to precisely that point again and again and again.
Unfortunately, that's only half the engineering required for manufacturing with robots -- the easier half, as it turns out.
The objects coming down the assembly line also must be made to arrive at an exact position relative to the robot, and that's no mean feat. Even if the assembly line can be calibrated to stop objects like car bodies at precisely the right place in front of the robot, the objects themselves all have to be exactly the same size and shape -- down to the micron sometimes -- or they won't receive the touch from the robotic finger at the precise point specified in the blueprints.
This nightmare of aligning robots with their work is one of the reasons some frustrated manufacturers are thinking of yanking robots off their assembly lines, according to Steven B. Skaar, Notre Dame professor of aerospace and mechanical engineering. But Skaar says it's way too early to give up on robots.
Together with several graduate research assistants, he's working on expanding the senses of robots with computers and video cameras, giving the machines what they've always desperately lacked: hand-eye coordination.
Skaar's foremost robot-control innovation substitutes video imaging technology for "teach-repeat," the laborious process of walking a robotic arm through the various stoops and twists it will have to make to perform a task.
In a laboratory nearly as big as a loading dock inside the Fitzpatrick Hall of Engineering, the professor and his students have wired an industrial robot the size of a park statue to a personal computer. The computer is linked to three video cameras mounted near the lab's ceiling. The cameras are trained on a wooden work platform clamped to an old metal office desk.
One camera feeds a view of the object on the work platform to the computer's monitor. Using a mouse, an operator then points to and clicks on a spot on the screen. This specifies where the operator wants the robot arm to go. A laser pointer mounted on the first camera then casts a red dot on the object, which will serve as a target for the robot. The other two cameras track the robot arm's motion, giving the computer the visual feedback it needs to steer the arm toward the target.
Skaar says this approach is more like the kind of hand-eye coordination we employ unconciously when we, say, reach for a doorknob. Although some robots alread come equipped with video cameras or proximity sensors, these systems rely on more of a calibrated, measure-and-move approach, he says. Known as camera-space manipulation, Skaar's approach has already shown it can be used to drill holes or carve computer-stored patterns in wood. Last year the robot performed the impressive feat of fitting a car wheel onto a standard five-bolt brake plate. In a test akin to what a person faces when changing a tire, the researchers tilted the brake plate this way and that and rotated the lug bolts to different positions. In hundreds of trials, the robot was able to adjust to the changes and slip the wheel cleanly over the lugs.
To see what else the robot has done, click over to http://www.nd.edu/NDInfo/Research/sskaar/Home.html.
Skaar, who co-edited the 1995 book Teleoperation and Robotics in Space, says camera-space manipulation would be ideal for space exploration. An engineer on Earth, seeing what a robot is seeing through its video camera, could point to and click on a spot on the screen and get the robot to pick up a Martian rock or tighten a bolt on a solar collector.
That kind of flexibility has drawn the interest of NASA's Jet Propulsion Laboratory, which designs unmanned space probes, and of a NASA technologist investigating ways to build a moon base using lunar concrete concocted from moon dust and rocks.
Skaar's research team also is developing a robotic wheelchair that draws on visual clues to help find its way to programmed destinations.
His current graduate assistants are Toshihisa Doi (working on the wheelchair project); and Michael Seelinger, Matthew Robinson and Zacarias Dieck (camera-space manipulation).