The research was presented to this month’s Autralasian Conference on Robotics and Automation in Melbourne by Richard Moore of The Vision Centre and the University of Queensland.
He says small and easy-to-maneuver unmanned aerial vehicles (UAVs) or pilot-less aircraft can be used for a range of purposes.
These include crop dusting, tracking whales, monitoring bushfires, as well as border surveillance and defence, according to Moore, who is completing a PhD under the supervision of Professor Mandyam Srinivasan.
“Any scenario where the aircraft has to fly low to the ground … it’s much more beneficial to be able to use a much smaller aircraft and it’s also less dangerous than having a human pilot on board,” he says.
But navigation using a compass to tell direction is not precise enough, according to Moore, and gyroscopes and accelerometers, which can determine the aircraft’s pitch and roll, are affected by the vehicle’s acceleration.
“This means that UAVs can’t perform significant maneuvers without losing their sense of direction for a while,” says Moore.
To develop more effective and safe navigation Moore and colleagues have developed a system that imitates the visual navigation of insects.
According to Moore, slow motion films show that when insects fly, they try as much as possible to keep their heads locked in the upright position to get a stabilized image of the horizon.
They are then able to navigate by paying attention to distant things such as the horizon, the sun, clouds and different levels of brightness in the sky, which are pretty static compared with their local environment.
The new UAV navigation system relies on panoramic signals captured by fisheye-lensed cameras, which are then processed using special software.
The system constantly compares and matches the panorama it sees to keep track of where the aircraft is going.
Moore says the navigation system would mainly be for use in rural environments or over the ocean and would be less effective in a built-up environment.
In a trial just south of Brisbane, Moore says the team showed that a small UAV was able to use the system to perform four minutes of complex maneuvers without losing its sense of direction, and then successfully land on a runway.
A laptop was used to communicate with the aircraft while it was on the ground, but once airborne it was able to perform all the required computations on-board and fly completely autonomously.
Unlike other vision-based navigation systems, Moore says his team’s combines both information on the pitch and roll of the aircraft, as well as information on its direction, to get the most complete and accurate picture of the state of the aircraft.
“I don’t know of anyone else who has performed these sorts of maneuvers using vision,” he says.
Vision-based navigation systems relying on cameras can be easily miniaturized, setting the scene for smaller and smaller UAVs, according to Moore.
“There are groups around the world that are actually developing these systems to be the size of insects,” he says.
Moore says funding for UAV development has come from NASA, the U.S. Air Force and industry, including Boeing.