When their invention "flew the coop" without any help, its creators were ecstatic--and they were winners in a national contest as well.
The contest was the 1996 Aerial Robotics Competition sponsored by the Association for Unmanned Vehicle Systems International in July, and "it" was a four-foot-long helicopter designed and built by students from MIT and Boston University along with two Draper Lab employees. Their team, whose faculty contact was Professor John Hansman of aeronautics and astronautics, bested about 20 others at locating barrels of mock hazardous waste over a test field.
A radio-controlled helicopter provided by Alex's Hobby Works in Watertown was the structural basis for the winning vehicle. Team members modified it so it could take off, fly in a search pattern, find and classify the targets (they were labeled as either radioactive or biohazard) and land again without any human direction or intervention during the "mission," although it had an override mechanism available in case of emergency. Alex Lob, the owner of the shop and the one who handled the radio-transmitter override, is the son of a retired physicist and MIT alumnus (Walter Lob, SB '41).
Designing and building the helicopter required a number of different skills, from computer programming to guidance-system expertise. "A project like this encompasses so many different disciplines," observed Eric Johnson, who received the master's degree in aero/astro with Professor Hansman and now works in Draper's simulation lab. Christian Trott, a graduate student in electrical engineering and computer science, built the power supply and integrated the electronic systems, while Mike Bosse of BU designed the image processing system.
Other team members were leader Paul DeBitetto, who also received the master's degree in aero/astro from MIT and who now works in Draper's guidance, navigation and control division on autonomous space and undersea vehicle research; Dmitry Brant, a junior in EECS, and Long Phan, a junior in mechanical engineering who worked on mechnical and electrical aspects of the helicopter, including prototyping the on-board electronics sensors.
During the contest, the helicopter had to fly over a simulated hazardous-waste dump measuring 60 by 120 feet, build a map of the barrels with one-meter accuracy and land in a small premarked area. "That's probably the hardest part," Mr. DeBitetto said. On its best run of the six official autonomous flights it performed in an hour, the helicopter found all five barrels and correctly classified two. None of the other teams even achieved autonomous flight, Mr. DeBitetto said. Those entries included a "tail-sitting" vehicle (like a vertical wingless airplane with a nose propeller) from the University of Texas at Arlington; in past years there have even been blimps and balloons, he added.
Mission instructions were stored in a laptop on the ground that acted as the guidance computer. This information was sent via RF modem to an on-board 486 computer, which performed navigation and control with additional in-flight data from a differential global-positioning system (GPS), a small camera, a compass and modified sonar ultrasonic altimeter, and a unit with gyros and accelerometers to determine the vehicle's orientation.
Part of the challenge was installing all this equipment and still enable the helicopter to fly. To minimize weight, the builders used pieces of balsa wood and copper-foil sheathing instead of lead, and they experimented with various shapes of rubber pieces to dampen vibrations that could affect the electronics.
MIT's participation in the contest had its origins in a 1995 IAP activity, although participants weren't able to build a flying model that first year. "We had a lot to learn," said Mr. DeBitetto. This year's effort was partly funded by an education/fellowship program at Draper that provided research support for some of the students.
"It was very rewarding to design something and see it start out as a bunch of separate pieces and then come together into a working system," Mr. Trott said.
A version of this article appeared in MIT Tech Talk on September 11, 1996.
