Robocar speeds (sort of) toward the fininsh line.
"They made more progress in the last day than any other team," says David Parish (ME, 1983), the president of Omnitech Robotics, who supplied CU with the vehicle hardware for the competition. "I think things went very well."
While "RoboCar" may sound like something you buy in a toy store to go along with an action figure, it is anything but a plaything. The chassis of the vehicle was taken from a Kawasaki Bayou 220 All Terrain Vehicle. The driver's seat, not needing a human occupant, was devoted to housing a 66 MHz '486PC that controlled the throttle, brakes, and steering. The PC also processed information about its surroundings from a video camera, ultrasonic and infrared obstacle detectors, laser range-finder, compass, gyroscope, and odometer perched on the vehicle.
In a nutshell, "the camera and other instruments gather an image of the world, the computer analyzes this information, and then makes a steering response," says Associate Professor Michael Mozer from the Computer Science Department. Mozer directed an independent study, special topic course, Robotics and Autonomous Vehicle Navigation, that prepared the vehicle for the competition.
"The typical response time for computer vision code like this is 4 to 5 seconds," says Mozer, "The students had to tighten the loop for the competition so that it had enough smarts to perform reasonably but still run fast. They were able to cut down the response time to under a second."
The goal of the eleven engineering students who took the class was to program their vehicle to autonomously navigate an obstacle course set on a 10-foot wide track in the fastest time. Adding degrees of difficulty to this task was that the vehicle had to stay under five miles per hour at all times. Last minute computer problems also had to be dealt with.
"We had lost the computer's hard drive the night before and had to reload everything," says Tim Linn, an aerospace master's student who participated in the class. "We also had to do some last minute reprogramming, so we felt really good about the final outcome." The only teams to beat the CU-Boulder squad on May 22 were the University of Michigan and the Colorado School of Mines, who finished first using a different Omnitech robotic vehicle.
The contest was hosted by the U.S. Army Tank-Automotive Command according to Mozer, in part to test new technology for unmanned combat vehicles. Vehicles that do not need a human present to operate could save lives in war zones. However, Mozer says the technology could also lend itself to "building personal vehicles with more intelligence." For example, Mozer says, "if computers controlled all of the vehicles on highways, vehicles could use the highways more efficiently and travel closer to each other with more safety. Then, once the car gets off the highway, the driver would control the car again."
Mozer adds that his special topics course provides an excellent example of what the College's future Integrated Teaching Laboratory (ITL) is all about. "This project had a definite goal and deadline the students were working for. It wasn't abstract theory," says Mozer, "It was also rewarding to see students from diverse engineering disciplines, undergraduates as well as graduates, each contribute their own expertise to the project."
"It was a great learning experience and lots of fun," says Linn, who landed a summer internship at Omnitech because of the project. "Working with computer science majors and writing software were things I hadn't done much before. I would love to do it again next year if I don't graduate too soon!"