Contact: Larry Bernard, Senior Science Editor
(607) 255-3651
ITHACA, N.Y. - Tommy and Lily can navigate their way around a room, locate a couch, determine the best way to move it and then move it across the room, all without speaking a word.
That they cooperate at all in such a task is a major feat, for Tommy and Lily are mobile robots in Cornell University's Robotics and Vision Laboratory who use advanced systems of infrared, ultrasonic sonar and computer vision to navigate and track objects. Both equipped with as many as 20 microprocessors controlling various aspects, Tommy and Lily are testimony that small teams of completely autonomous mobile robots could perform a variety of tasks in industry.
The largely student-built robots are prototypes in the lab, housed in Cornell's Computer Science Department. Designed by two computer scientists for undergraduate and graduate teaching, the lab swarms with students as they build examples of what they learn in class.
The researchers, Bruce R. Donald and Daniel P. Huttenlocher, both assistant professors, use the lab for undergraduate, graduate and postdoctoral training.
They believe it is the only computer science lab in the nation to be run by a pair of Presidential Young Investigators - awarded by the National Science Foundation to promising young researchers in science, math and engineering.
Huttenlocher also is a decorated instructor: Last month (September) he was named the New York Professor of the Year by the Council for Advancement and Support of Education, and last year he won two of Cornell's highest teaching awards.
Although the robots are used for teaching, they represent advances in the field and the professors' own research. "We want to make robotics as modular as can be," Donald said. "If you want a new sensor, just pop it in. That way, you can simply change the part that controls the sensor without modifying the surrounding program. We've been doing that in computer science for some time, but in robotics, it's a new thing." The Cornell robots, about 2 feet tall, move about by controlling motors attached to wheels. They are autonomous, and both have a ring of 12 ultrasonic sonar sensors. They have infrared sensors and contact, or "bump," sensors and are battery powered. Tommy can run for about three hours; Lily, five or six hours before recharging. Tommy, the second of four robots built at the lab, can do a variety of tasks with other robots. The first robot was a prototype for Tommy, then came Lily and a fourth is being built. "It's not too hard to get a single robot to do a manipulation task. But to have a robot team cooperate in manipulation is a very difficult problem," Donald said.
Additionally, the robots have a sophisticated landmark-based navigation system that Huttenlocher developed. He devised a two-dimensional system of "viewing" in which images are retained and tracked, so that the robot can "recognize" a landmark it has seen before. His algorithms for comparing and recognizing shapes of two-dimensional images allows Tommy to follow someone around a room or move toward something. He described the system at the International Conference on Computer Vision in Berlin (May 1993).
For Tommy to "see," the robot must be tethered to a workstation. But the newest robot, Penelope, will have an onboard vision system and a laser tracking system as well. They are described in Proceedings of the IEEE International Conference on Robotics and Automation (May 1992). Although many undergraduate and graduate students contributed, Tommy was built under Donald's direction largely with graduate students Jim Jennings, Russell Brown and Jonathan Rees, in work funded by the National Science Foundation, Air Force Office of Scientific Research, AT&T Bell Laboratories, Intel and Cornell's Mathematical Sciences Institute.