Between the red voice-catching contraptions in the Brickyard, Tim Josey, Adam Propst and Patrick Keistler stand around a helicopter. Passers-by stop and stare as the members of the Aerial Robotics Club try to fire up the engine with a hand-held starter. The machine sounds like a power drill as it brings the helicopter to a shuddering start.
The blades span about five feet across, and as they spin faster a wind flow is created and the helicopter begins to rise. Keistler, a graduate student in aerospace engineering, pilots the copter as it hovers over the ground.
“This is one of the first times we’ve flown it,” he said. “This helicopter is made to do tricks, like fly upside down, but right now I’m still working with it.”
Keistler keeps the helicopter in the air for about 10 minutes before bringing it down to land softly on the grass. Takeoffs and landings are two of the most difficult actions the club must conquer before heading off to their competitions next summer — especially because they must program the helicopter to do it autonomously.
The club is currently planning to compete in two contests, the International Ariel Robotics Competition with the helicopter in Georgia, and the Unmanned Vehicle Systems Student Competition with a robotic airplane in Maryland. Both competitions are part of the Association for Unmanned Vehicle Systems International’s effort to challenge engineering students’ minds, according to the AUVSI Web site. And they revolve around technologies that the U.S. military may need in crisis situations. This makes the challenges real, difficult and expensive.
Club members are already building, testing and applying for more money from Student Government and businesses — so they can build more and test more for the 2007 competition. Sitting at a computer in Broughton Labs, Josey, a sophomore in aerospace and mechanical engineering, acknowledges the financial commitment it takes to be involved. “It is really expensive, but we ask for a lot of donations,” he said. “We are working on sponsorship letters at the moment.” The team has spent the past few years making a name for themselves with robotic airplanes; their newest plane already wearing sponsorship stickers. But this is their first year seriously participating with the helicopter. “We’ve been [to the helicopter competition] before, the first year they had it there were two teams competing and neither had a serious entry,” Matt Hazard, a junior in aerospace engineering, said. “And we got second out of two, so it doesn’t really count.”
The airplane pros
This is the team’s mission for the 2007 UVS student competition, if they choose to accept it: create aircraft that can detect, identify and provide location and orientation of targets in a combat zone.
The competition focuses on technology that transfers into military need, in this case creating a mock combat zone where the planes must autonomously pick out targets and relay the information as soon as possible to the Marines, according to the UVS rules.
The competition consists of four parts: a journal paper, a static presentation, safety precautions and the flight.
The team won the journal paper entry for 2006, beating out schools such as the Massachusetts Institute of Technology and Cornell University — a large feat that was amply rewarded.
Prize money is a huge factor in these competitions.
“Last year we took sixth place and got $2400, combined with our journal paper, presentation and pretty terrible flight,” Hazard said. “So if you swept everything you could probably get $11,000.”
But $11,000 doesn’t go too far when the materials to build a plane that can fly by itself and identify targets can cost upwards of $6000.
The planes will continue to get more expensive as the competition evolves every year to compensate for students mastering new technologies.
In 2006 the UVS only gave the students GPS way-points to keep inside a certain area and then told them to identify the targets. The team fed the GPS way-points to a ground service station that ran with the autopilot, the plane then flying through those points, according to J.B. Scoggins, a sophomore in aerospace and computer science engineering.
The team’s method for working through these problems impressed the judges.
“The judges were kind of favoring us because they looked at our system and thought it was pretty good,” Josey said.
After taking off, the plane flew over the field with a Nikon D50 camera attached to the bottom, snapping pictures of the field. Having the camera permanently face down made it easier to process the pictures because they didn’t have distortions, according to Scoggins.
“One thing to note on is that with the software we have written the images that the cameras take are downloaded in real time,” Propst, a senior in aerospace engineering, said. “And they will pop up on the screen oriented and scaled just like you are making a map, so as you cover the area you get what you would see in one camera shot.”
The team could then figure out the targets’ orientations and sizes because the computer programs could calculate from the plane’s altitude and angle. And then they printed out the close-up pictures of the targets with the information on it for the judges.
“And we have 40 minutes to do all of this,” Scoggins said.
The 2006 plane had some trouble though — the battery died after only five minutes of flying time.
“We had a battery malfunction and all the systems shut off,” Scoggins said. “But we still found four targets in five minutes of flying time, which was pretty incredible I guess.”
The 2007 competition offers plenty of new obstacles for a team that feels as though their performance in 2006 did not live up to their potential.
“We talked the talk, but we didn’t walk the walk,” said.
The 2007 competition is leaning more towards autonomy, Scoggins speculated. The judges will now be giving points for autonomous landings, and they also gave the teams a list of the targets in the rules — signifying that they may be leaning toward autonomous target selection.
This means the computer does vision recognition and says “that’s not grass, that’s a target,” and then tells you what it is, according to Scoggins.
“They gave us a list of possible targets when in previous years it was ‘just go see what you can find,’ and so we’re looking for specific things like a red target with a green ‘A’ on it,” he said. “They are looking for a computer to actually find them, not a person sitting at a computer to find them, so stuff will be a lot harder this time around.”
A change also has to be made in their photographing technique, because now the camera must be able to rotate 60 degrees in all directions. This creates difficulties for the team since their plane isn’t built to handle that much rotation.
The team will have to modify the plane or move the mount down, shifting the camera closer to the ground — which isn’t a very good thing, according to Scoggins.
“Plus if you take a picture 60 degrees to the side it’s not a nice square picture,” Hazard said. “It can be really skewed because the picture could be 1000 feet wide at the top and only 200 feet wide on the bottom, and you see a lot more sky also so it’s a lot more work to fix the image.”
Their new airplane is a monster though, and the students have another 10 months before the next competition.
The airplane for this next competition has a wingspan of over 12 feet, according to Propst.
The body came from a donated kit, the students building and then modifying it for their specific needs. They split it dorsally, doubling the internal area so they could fit more electronics inside. They also replaced some of the bright red covering with plywood.
“It carries like 20-25 pounds of stuff over its own weight,” Hazard said. “And it flies about 35, 40 miles per hour.”
Scoggins said that the plane weighs about 45 pounds, making it much larger than their previous planes. This increase of size and weight will be an advantage for the team this year, allowing their plane to move slower and have an easier time identifying the targets.
A new era: the helicopter
This time there isn’t one reason provided for the military to need assistance, but three. It could the need for pictures of hostages held by terrorists. Or it could be to see inside a building ravaged by nuclear warfare. Or it could be to view the insides of an ancient building plagued by the Ebola virus.
Any of these scenarios given by the AUVSI explain why the military might need an autonomous aircraft to fly three kilometers, identify a building, fly through a window and take pictures of inside. “The IARC competition that we want to go to with the helicopter is you take off and you fly though a series of GPS way-points, the last one is like three kilometers away,” Scoggins said.
“And then, that GPS point is over a cluster of buildings, and one of the buildings has a sign on it. And there is supposed to be an open window somewhere on that building, so it’s supposed to find that and then take pictures of the inside. And this is all under 15 minutes, and it’s all autonomous, so you basically press go and just wait for it to say ‘Hey, I took pictures!'”
Every year $10,000 is added to the pot, and this is the seventh year — no one yet accomplishing the mighty feat.
This large pot attracts more than just students. The IARC is different from the UAV student competition because it is open to everyone, including graduate students and corporations. And there is more than $70,000 on the line — the team that wins will almost certainly receive development contracts.
“The technology is actually way beyond anything that the current military technology can do,” Hazard said. “So if you build one that can do it, then you’re going to get a contract.”
In fact, Georgia Tech students last year received a government contract of about $2 million just for getting close to accomplishing the task, according to Scoggins.
Tech’s approach found the window and shot in a pod with a rover to come out and look around. But foam surrounded the pod, and when the helicopter sent the pod out the foam should have split off once it landed — but instead when it shot off the foam fell off, the pod hitting a wall and splattering, Scoggins said.
In order to identify the building, the helicopter must employ a system that can identify objects and open windows.
“We’re planning to use the vision system that works like human eyes; so it has two cameras so it has depth perception,” Hazard said. “There are a lot of ways you can tell if there is actually glass there, like trying to look for your reflection, which is really complicated.”
Scoggins said that they will probably just look for a black square and hope for the best because the software needed is so complex. The software is so difficult in fact that he related it to a graduate project.
The software is so complicated because of what it must accomplish.
“The thing is that you’re guaranteed an open window of a certain size,” Scoggins said. “But what we’ve learned from past competitions is that there are half-open windows and doors and lights in random places. And there are also power lines next to the building and that is a death trap for any flying thing — just run into that and you’re screwed.”
Hazard has another idea that could simplify the software needed.
“I think we should just mount a paintball gun and shoot at the window and if it leaves a mark move on, and if not then you’re good.”