ROBOTICS AND AUTONOMOUS SYSTEMS
Military, Industry Racing to Create Sense-and-Avoid Systems
Certifying a sense-and-avoid system that allows unmanned aerial systems to avert oncoming traffic is a prerequisite for full integration of unmanned aerial systems into national airspace. However, that may not be feasible in the near future, according to the agency’s UAS roadmap released in 2013.
Airborne sense-and-avoid technologies are so complex that “significant progress” is not expected until the mid-term, and actual fielding of a system is a long-term objective, the roadmap said. Initial FAA certification of an airborne sense-and-avoid system will not take place until 2016 at the earliest.
William Semke, associate professor at the University of North Dakota’s mechanical engineering department, believes that it will take at least five years — and that would be an aggressive timeline.
“Things would have to go very well for that. … Seven to 10 [years] is a more realistic timeline without any major stumbling blocks,” he said.
The FAA has yet to establish what standards such a system will have to meet in order to be safe enough for use in the United States.
Most of the technologies needed to create sense-and-avoid systems are already mature, said Paul Schaeffer, the Air Force’s program manager for sense and avoid. “The biggest problem that we have right now is a lack of policy guidance. Technology itself is not … rocket science. It’s really the integration of capabilities that we already have into a system to perform a new function,” he said.
The service is working with the FAA through its aviation rulemaking committee to modify existing sense-and-avoid standards and with the Radio Technical Commission for Aeronautics to write the minimum performance standards for the systems, he added.
According to current regulations, an aircraft operating in civil airspace must have a “see-and-avoid” capability to help it evade other planes or wildlife. A human pilot fills that role in manned planes and helicopters. With unmanned aircraft, the military employs either chase planes or ground observers to notify operators when it is in danger of a collision.
The military has been able to fly its drones freely in countries with little air traffic such as Iraq and Afghanistan. However, they are not capable of navigating more congested airspaces populated by many unmanned and manned aircraft.
The Air Force began developing sense-and-avoid technologies in 2001 after the service recognized those limitations, Schaeffer said. In the 13 years since, the service has built a system it hopes to further mature before installation on selected UAS by the mid-2020s.
Doing so will give the military broader global access and the ability to support missions such as disaster relief and border control, he said.
In order to see cooperative and noncooperative aircraft as well as wildlife, many sense-and-avoid systems use a mix of sensors. The traffic collision avoidance system, or T-CAS, and automatic dependent surveillance-broadcast, or ADS-B, transmit information on an aircraft’s distance, altitude and velocity to others within receiving distance.
However, these systems only detect other aircraft with a corresponding transceiver or transponder. Because not all aircraft are required to carry such equipment, UAS would remain vulnerable to collisions with noncooperative aircraft. Incorporating cameras that detect both visible light and infrared, radar and light detection and ranging (LIDAR) would allow sense-and-avoid systems to see those threats.
The Air Force’s airborne sense-and-avoid system for UAS incorporates four types of sensors: T-CAS, ADS-B, electro-optical cameras and a purpose-built radar, Schaeffer said.
“We then have a series of algorithms that take the data from those sensors and convert it into a situational awareness picture for both the pilot and for the computer,” he said. The pilot can manually guide the aircraft out of the way, but if the operator is not able to respond in a timely manner, the system is automated to avoid collisions.
The system was created for use onboard the Global Hawk and is being retooled so that it can be installed on other large drones such as the Shadow, Predator and Reaper. The Navy’s Triton and unmanned carrier-launched surveillance and strike drones could also incorporate the system. The Air Force would like to scale it down even further for use on the Hunter and Raven, Schaeffer said.
To date, the system has flown 73 demonstration flights — most recently in January 2013 — using a Calspan airplane modified to behave like a Global Hawk. The next series of flights will take place in early 2015 and will test upgrades to the sensors and software, he said.
In September, the Air Force released a request for information for a common airborne sense-and-avoid system that can work with any sensor or UAS, including hardware, software and algorithms currently used and under development by the service, Schaeffer said.
“The RFI we put out is really for the brains of the system. It’s really [for] the sensor fusion algorithms and the collision avoidance algorithms and building that into a single box that is both sensor agnostic and platform agnostic,” he said. “I really want the taxpayer to only pay for this development once.”
The Army is developing its own sense-and-avoid technology using small ground radar, said Viva Kelley, the Army’s product director of unmanned systems airspace integration concepts, during a February briefing to reporters. That system is scheduled to undergo live flight testing this summer at Dugway Proving Ground, Utah.
The service plans to install ground radar for acceptance testing at the following installations: Fort Hood, Texas; Fort Riley, Kan.; Fort Campbell, Ky.; Fort Stewart, Ga.; and Fort Drum, N.Y. Installation at Fort Hood is scheduled to begin in the next few months, Kelley said.
“The money is there, we have a lot to do to get prepped for first flight,” she said.
Using a ground-based sense-and-avoid system has certain advantages, particularly for the Army.
For instance, the service’s UAS fleet primarily comprises small, light aircraft, such as the 13-pound Puma and 4-pound Raven. Putting a large airborne sense-and-avoid payload on such aircraft would sap their processing power and weigh them down, Kelley said.
Army drones fly low and in congested environments, but airborne systems have difficulty operating close to the ground because their radars and cameras see a lot of clutter. A ground-based radar can look up and direct aircraft more accurately, Schaeffer said.
Kelley said the Army Aviation Engineering Directorate has already approved the ground-based system’s design and requirements and will give final approval. The FAA will authorize it for operational use in U.S. airspace as early as next year.
The service has not chosen what company will supply radar to each of the test sites, but the testbed uses SRC Inc.’s LSTAR, Kelley said.
“We’ve been very careful to develop it the right way so that the cost is reduced on the backside,” she said. “These radar are relatively inexpensive … compared to other radar.”
“Ultimately the goal is to tie to the airborne sense-and-avoid system that the Air Force … comes up with, and then you can fly an unmanned system everywhere,” she added.
The military is not the only entity trying to develop a sense-and-avoid capability. The defense industry and academia are pursuing their own systems.
To build its sense-and-avoid technology, the University of North Dakota’s unmanned aircraft systems engineering lab combined commercial, off-the-shelf hardware with custom circuit boards and algorithms specially designed by its researchers, Semke said. The team of faculty and students used an ADS-B transceiver to collect and broadcast data on the velocity, position and direction of a UAS. Algorithms mine the data for evidence of potential collisions, and then override the autopilot system to direct the aircraft in a safe direction, he said.
The algorithm can also be used with other sensors, such as ground-based radar, to give a UAS the ability to avoid wildlife and noncooperative aircraft, Semke said.
Since the project started in 2007, NDU students and faculty have refined their algorithm so that their unmanned aircraft follow the same right-of-way rules as commercial planes and civil aircraft. They have also introduced terrain avoidance capability, he said.
“What we’ve been working on quite a bit recently is dramatically different aircraft speeds, which is a challenging problem” because there are different strategies for efficiently avoiding an aircraft based on how fast it is going, Semke said. The lab is also working out how best to respond to rotorcraft that can move sideways.
The size and weight of sense-and-avoid equipment has dramatically changed since the program started. Initially, the lab’s payload weighed 30 pounds, but that has shrunk to about four pounds and could be reduced to under a pound by using the latest equipment available on the market, Semke said.
ADS-B transceivers, which are the heaviest part of NDU’s system, “started out being about 15 pounds, and now they have some that are under a pound. Our computer system and our hardware would fit inside a deck of cards.”
Companies are also testing proprietary sense-and-avoid systems.
Insitu Pacific partnered with Queensland University of Technology in Australia to develop its own system for use onboard small UAS like the company’s ScanEagle.
“The sensor itself is a visual camera that detects differential movement in pixels which replicates the human eye’s perception of movement. Once a pixel has been identified as a threat it sends an aural warning to the operating crew,” said Dale McDowall, Insitu Pacific’s director of business development and strategy.
During a recent flight, the system provided real time warnings to the ground control station, allowing operators to move the ScanEagle out of the way of other aircraft. Although autonomous avoidance is not a part of initial tests, the company could develop and incorporate such algorithms in the future, McDowall said in an email.
Greg Cox, corporate vice president of Sierra Nevada Corp., said the company is closely following the Air Force’s common airborne sense-and-avoid system RFI. Sierra Nevada created and demonstrated an airborne radar that is small enough to fit on a 375-pound Shadow and can operate in any conditions in which UAV can fly.
General Atomics in November demonstrated its system, which includes a BAE Systems ADS-B transponder, General Atomics air-to-air radar and Honeywell’s T-CAS system. During the test, a Predator B used all three sensors to detect and evade two intruder aircraft.
In order to convince the FAA to allow UAS into domestic airspace, makers of these systems will have to prove their technologies are as reliable as a human pilot, Semke said.
Creating artificial intelligence that can mimic human decision-making processes is a major challenge, as is exercising sense-and-avoid systems through test flights and simulations that put the technology through its paces, he said.
“The human brain is incredible when you’re flying an aircraft. The decisions you make based upon the knowledge of all your training ... it’s just hard to recreate that with artificial intelligence,” he said. “The artificial intelligence community has made great strides, but … they’re not recreating the decision making of a human yet.”