CSL’s Gao to develop technique for flying UAVs in urban environments
Amazon wants to start using drones to deliver packages by 2017, but if you live in a high-rise apartment, you might be waiting a bit longer.
That’s because UAVs use GPS for localization (understanding current position) and navigation (knowing how to get from point A to point B). In urban areas, however, high-rise buildings may block the line of sight to GPS satellites, causing drop-outs or making the signal completely unavailable.
CSL Assistant Professor Grace Gao and her team are working to overcome those limitations by using vision-based positioning to fill in the gaps when GPS is unavailable.
“We expect civilian drones to be used in any number of applications, from construction monitoring to emergency response,” said Gao, a member of Illinois’ aerospace engineering faculty and affiliate faculty of electrical and computer engineering. “For these applications, it is imperative to have accurate, always-available, and safe positioning and navigation to help prevent drones from colliding with tall buildings.”
Current vision-based positioning systems lack the ability to aggregate data from multiple sources. Drones also have limited computational capabilities and are incapable of handling large data sets. In addition, data from sources like Google Street View require human effort to obtain and are quickly outdated.
Gao’s team seeks to integrate on-board GPS with camera vision. They will develop a universal platform and interface to make all image and visual video data accessible to UAVs in real-time. To accomplish that goal, the team will develop algorithms that leverage GPS Direct Positioning (DP), which estimates navigation from the raw signal, instead of GPS scalar tracking, the traditional method.
DP provides increased robustness when it comes to overcoming GPS signal noise, attenuation, multipath issues, and obstruction. However, it is also computationally intensive, requiring the team to develop algorithms that reduce computation load and increase robustness by tight-coupling of DP and vision data. The algorithms will partition the tasks into on-board and off-board components, sending some computational work to the cloud or a data center via a wireless network.
“Building surfaces, especially if made of glass, may reflect and diffract GPS signal rays, resulting in significant undetected errors in positioning,” Gao said. “It could cause collisions and even injury to people. Our goal is to make drone use both feasible and safe in urban environments.”
To test their technology, Prof. Gao and her students conducted flight tests in a GPS-challenged environment over Green Street using the UAV built by her research group. Although Gao’s group has received the FAA Certificate of Authorization (COA) for flying UAVs, their goal is to fly UAVs over busy streets safely. See their solution – flying UAVs in a moving cage.