Virginia Center for Autonomous Systems

SPAAROs Take Flight

UAV on grassy field

VaCAS has a new fleet of unmanned aerial vehicles (UAVs) designed by an aerospace master’s student in his spare time. Justin Murtha wanted to test his reliability design process for UAVs and ended up creating a new fleet for Virginia Tech.

His autonomous aircraft are less expensive and more reliable than the retrofitted radio-controlled RC airplanes they replace. Equally important, they can easily carry a variety of payloads for expanded research and teaching missions.

Notorious unreliability

“One of the big advantages of UAVs is that they can perform the dirty, dangerous, or dull missions cheaper and with less risk to human life,” said Craig Woolsey, associate professor of aerospace and ocean engineering (AOE) and director of the Nonlinear Systems Laboratory. “Unfortunately, UAVs have yet to fully deliver on that promise,” he added. “They can be expensive and are notoriously unreliable.”

He cited a 2005 publication by the Department of Defense, which showed small UAV mishap rates at least ten times as high as the rate for manned aircraft. UAVs are doing valuable tasks in the field, however, so they want to help manufacturers make the aircraft more reliable.

Commercial aircraft companies can predict the reliability of their aircraft and set maintenance schedules with uncanny precision, Woolsey said. “It’s an expensive process to establish reliability, but worth it for vehicles that cost millions of dollars and carry human pilots and passengers.”

Effective small UAVs, on the other hand, may be made by “mom and pop” companies at a cost of only a few thousand dollars. “We don’t want to make the airworthiness process so expensive that it stunts the technology,” Woolsey said.

Designing for reliability

Funded on a contract from NAVAIR, Murtha investigated a cost-effective process to “design in” reliability for small UAVs. He developed a system to make informed design decisions about reliability based on imprecise and incomplete failure rate data. His research makes it easier for manufacturers to make good component selection decisions, such as when to buy an expensive servo and when to use redundant cheap ones.

Probability of Failure Vs. Cost

In developing a method to design-in reliability for UAVs, while still keeping the aircraft low-cost, Justin Murtha came up with a way for taking imprecise failure data and making better decisions about component selection and aircraft configuration. Imprecise failure data is collected from a variety of sources, including rigorous testing, numerical modeling, and expert opinion. This cost curve shows how the decrease in probability of failure relates to the cost of making improvements, such as including a redundant component or choosing a more expensive, more reliable one. Navy sponsors sought a method to find the “knee” in the cost curve, indicating a point of diminishing return on investment.

Graph showing a "knee" in the cost curve

Murtha likes to design and build planes. And the Virginia Tech UAVs were suffering. So, in his spare time, he designed and built a research and teaching UAV using his reliability-based decision-making process. He wanted it to serve as the department’s workhorse for years to come.

Students in the Nonlinear Systems Laboratory had previously modified off-the-shelf 110-inch-wingspan Sig® Rascal airframes, he explained in his thesis, but increased payload and fuel requirements forced the Rascals to be flown at 28 pounds, roughly twice the weight they were designed for.

He described how the r/c airframes were suffering from structural fatigue because payloads were difficult to install in cramped quarters. “The Rascal airframes were dangerously overweight, over-stuffed and obsolete,” he said.

Radio-controlled platforms typically do not have built-in redundancy, so the reliability of the entire system is dependent on any single component, he explained. He wanted his aircraft, called the small platform for autonomous aerial research operations (SPAARO) to survive the rigors of inexperienced students and multiple, continually changing payloads. This required some redundancies and additional safety/reliability features.

SPAARO features

A SPAARO airplane has a 12-foot wingspan and fully loaded weighs 55 pounds, including 10 pounds of payload. It sports a differential GPS system and redundant elevators, ailerons, and rudders. “If one rudder fails, we still have directional control,” said Woolsey. “The biggest decision, however, was selecting an electric starter for the motor,” he added. “The military and our own experience shows that 40 percent of all UAV losses are propulsion losses, so our engine accommodates some failures. We have saved the airplane twice by restarting it. We didn’t have to land without power.”

3 SPAAROs

Eddie Hale (left) and Mark Monda prep the SPAARO before a test flight. The payload bay is roomy and easy to use.

The commercial autopilot includes its own failsafes. For example, if the telemetry link fails, it is programmed to return the UAV to the ground station and circle above it. The autopilot also allows a human pilot to take over via manual control. A pilot can also completely override the autopilot using a secondary control link.

The SPAARO’s large payload capability and easy-to-install features make it a very flexible platform for many applications. “If you can fit your payload in this, we can fly it,” Woolsey commented.

Applications galore

SPAARO’s potential applications are boundless and the fleet is available for researchers in many fields, from agriculture to zoology. The team aims to keep three SPAARO aircraft available at all times. Woolsey anticipates a variety of projects within his research team alone. Although his team works primarily on algorithms and control theory, they like to demonstrate and test their findings on actual vehicles.

Ph.D. student Mark Monda is working on a machine vision application to help UAVs see and avoid other aircraft. He is developing an algorithm that uses data from a camera image and identifies the attitude of an object with known geometry. He plans to use the SPAARO fleet to test the algorithm and demonstrate a rough degree of situational awareness.

Another project involves helping NAVAIR specify the flying qualities to serve as criteria for certifying that UAVs are airworthy and mission-effective.

In one mission, SPAARO was equipped with a sophisticated camera that rotates and takes still-shots and movies to provide real-time mapping information. This is useful for updating satellite maps on geographical features that frequently change, such as the course of a river, debris, or new buildings. Similar systems could conduct surveillance missions, and direct troop and unmanned surface vehicle movement.

3 SPAAROs

The fleet of Virginia Tech SPAARO unmanned aerial vehicles awaits a mission on the runway at Fort Pickett.

SPAARO can also accommodate sensors that, for example, sample the air for pathogens or pollution or pollen. Collaborator David Schmale, assistant professor of plant pathology, physiology, and weed science (PPWS), has a long history of using small UAVs for “aerobiological sampling” and has recently worked with Woolsey to install and fly sampling devices on a SPAARO.

The SPAARO can be equipped with a miniature air data unit which includes a pitot-static probe to measure airspeed and vanes to measure aerodynamic angles. This device, together with the autopilot, provides a method for measuring the complete state of vehicle motion, enabling advanced flight tests and control experiments.

The first major mission for the SPAARO was for a northern Virginia firm that makes spectrum analyzers. The firm had developed a prototype system to localize very small aperture terminals (VSATs), which terrorists often use to communicate. Two prototypes flew simultaneously on two SPAAROs, and both the payloads and the UAVs were successfully demonstrated.

Aircraft for the classroom

The fleet will also be used for at least two aerospace courses. “We’ll be developing some simple aircraft performance and stability and control labs for students to do in the field,” Woolsey said. Virginia Tech teaches aircraft performance in the sophomore year and aircraft stability and control in the junior year. “It’s important to get a feel for stability and control and how rigid bodies can rotate under the influence of disturbances and control moments. Being able to illustrate this on a real aircraft will greatly enhance student learning.”

Aerospace programs once used piloted aircraft in their courses, but cost and safety issues have left less than a handful of schools with real airplanes for coursework. Many schools have moved to the micro, palm-sized aircraft that can be flown indoors. “We are making a commitment to the larger aircraft that can be flown outdoors and experience the vagaries of wind and weather,” he said.

The SPAARO fleet does present some challenges. “These aircraft have more capabilities than we are used to,” explained Monda, the Ph.D. student working with machine vision. “We are still learning how to take advantage of that and realize we no longer need some of our old workarounds. SPAARO provides us with capabilities we haven’t been able to do before.”

Another challenge is keeping a core team of certified UAV pilots. “We need to continually refresh our skills and make sure enough students can operate and repair the aircraft,” Woolsey said. “It’s a fun dilemma!”