June 5, 2013
Last year, inspiration struck Maruthi Akella as he made his way home after a long day at the office. Tuned into NPR on his car radio, he heard a story about a team of Columbia University researchers who trekked the nearly inaccessible terrain of the Himalayas above Bhutan for days in severe weather to collect data about melting glaciers.
There’s bound to be a smarter way to collect the data, thought Akella, associate professor in the Department of Aerospace Engineering and Engineering Mechanics at The University of Texas at Austin.
North of Bangladesh and approximately 900 miles northeast of India, Bhutan has an obvious interest in making sure the glaciers and water systems remain available and sustainable. As the baselines of the glaciers diminish over time, so too does the amount of water generated when they melt. The small country depends largely on revenue generated by hydroelectric energy sales to South Asia, and one-fifth of the world’s population depends on the glaciers, along with major Himalayan rivers and erratic monsoon seasons, for water. The first step to avoid catastrophe is to understand what’s happening.
Akella imagined teams of coordinated and autonomous unmanned aerial vehicles, commonly referred to as UAVs, making the measurements more efficiently and effectively.
The realization that NASA’s IceBridge program was collecting similar data in the polar regions soon followed, as did conversations with the space administration. Akella’s ideas could benefit the ongoing mission to close the gap in polar ice observations between ICESat (Ice, Cloud and Land Elevation Satellite), which was in orbit from 2003 to 2009, and ICESat-2, scheduled to launch in 2016.
“Unfortunately, this is a time when a lot of rapid changes are happening to the global climate,” Akella said. “So we really cannot afford to have a gap in measurements.”
As part of NASA’s existing IceBridge program, UAV-based methods for data collection require significant human investment and lots of advance planning. Prior to commencement of each airborne survey experiment, NASA personnel are required to map out the airplane flight paths based on fast-changing weather conditions, including wind gusts potentially as fast as 60 mph.
Akella began developing the onboard flight computer’s brains, called algorithms, for the airplane hardware and sensors already developed by NASA and other contractors. An algorithm is logic — a set of consistent and stable rules with ample safeguards to avoid system breakdowns — achieved through mathematical principles.
When fully developed, the algorithms, ported to a circuit board the size of a piece of chewing gum, will enable the airplanes to plan their own trajectories, avoid collisions among themselves, make their own measurements and share that information with their peers under constantly changing environmental conditions with minimum human intervention.
Building upon the vast body of coordinated UAV flight research at the university as part of the Controls Lab for Distributed and Uncertain Systems, Akella was positioned well to pursue his vision. His team of current and former graduate students had already developed coordinated target pursuit algorithms for teams of airplanes that were deployed by the U.S. Department of Defense.
The existing concepts were applied to solve the problem of taking continuous periodic measurements in the polar regions. However, the civilian operation presented new challenges. The mission would last an extended period of time and the airplanes would fly above fjords and gigantic ice cliffs in hostile winds occasionally blowing as fast as their own speeds. They would have to recognize when they were flying into headwinds, at the boundaries of their capabilities, and gracefully drift away to allow other airplanes with winds at their backs to move in and make the measurements.
“That’s where the technology leap is coming in as we are working on this IceBridge project,” Akella said.
The airplanes collect data from sensors on board that measure environmental conditions like wind velocity and temperature, as well as sensors they drop on the ice that measure the direction and amount of ice shift. Hierarchy is established autonomously among the airplanes, rather than depending upon humans to designate the signal routing structure among the airplanes, which might not be the most efficient process. As with military missions that monitor moving targets, the airplanes return to the areas where they expect to find the sensors and query the sensors’ unique IDs to locate and collect the measurements.
Yet, mathematically perfect solutions are impossible.
Constant communication among all the airplanes, which result in optimal path planning and coordination, would require unlimited battery power and uninterrupted two-way radio signaling. Neither is practical. Furthermore, even with a modest number of airplanes and sensors, computing globally optimal solutions that utilize all-to-all communication would impose computational requirements that scale up so rapidly in a polynomial fashion that they become unachievable with most available flight computers.
Sub-optimal solutions are satisfactory, though.
For example, transmitting, or speaking, is more expensive than receiving, or hearing, in terms of battery power for two-way radios. Therefore, each airplane periodically turns off its transmitter to conserve power, sharing information as efficiently as possible. When an airplane’s power is nearly exhausted, it returns to the base to recharge and the team reorganizes to cover more territory with fewer players.
“When the technology matures, I would suspect that even after ICESat-2 goes up, we will continue IceBridge-type missions because it’s cheaper to send an airplane to take a measurement than to reposition a satellite,” Akella said. “And you can get better quality and more timely measurements with an airplane because it’s closer.”
In August, Akella plans to perform flight tests for validation of the path planning and coordination algorithms using a team of three or four small quad-copters to assure that they are ready and right not only on computer simulations, but also in realistic flight conditions. Decker Lake’s flight range is the most likely site for this planned first test. Subsequent test flights might open for public viewing.
“Ultimately, we also want to help earth science researchers like those in Bhutan or someplace else, where they are walking for days, or even weeks, to make measurements,” Akella said.
