UAV- Det nærmer seg løsninger

Klippet fra Flight Safety Foundations Newletter

 

Sense-and-Avoid Research Update

 

U.S. researchers in September observed the performance of two mature sense-and-avoid algorithms for unmanned aircraft systems (UASs) during a series of flight tests near Grand Forks, North Dakota. Flights comprised 120 encounters in which automatic maneuvers by a UA-surrogate airplane were expected to resolve virtual traffic conflicts with an intruder aircraft, participants said. A major purpose of the two weeks of flights was to compile data for later validation of sense-and-avoid computer simulations.

Complete results await final reports, but examples of successful conflict-avoidance maneuvers were replayed for AeroSafety World and other media representatives in a Web conference–based telephone briefing about the Limited Deployment–Cooperative Airspace Project (LD–CAP). The briefing was led by representatives of the U.S. National Aeronautics and Space Administration (NASA) Langley Research Center, MITRE Corp. and the University of North Dakota (UND) on behalf of all the research partners.¹ LD-CAP, now midway through its two-year research agenda, plans to share results with the aviation research community and regulators, they said.

“What we want to do is create the scientific data that the community needs to make decisions about how to mitigate [UASs’] lack of see-and-avoid with a sense-and-avoid solution [for routine flight in non-segregated civil airspace],” said Andy Lacher, MITRE’s UAS integration lead. “We’re using the [flight] data to validate our computer models and inform the community about the performance and the viability of a cooperative, autonomous, sense-and-avoid algorithm. … We’re conducting this research using a [NASA-owned] surrogate unmanned aircraft — an SR22, it’s a Cirrus aircraft. … The sensor we are focused on in this research is … automatic dependent surveillance–broadcast [ADS-B]. … We don’t necessarily believe [ADS-B] is the sole sensor that would be appropriate for sense-and-avoid, but it is a good sensor to evaluate because of its excellent accuracy. It’s a good sensor source to be used for determining whether you can have an automatic algorithm.

“We are focusing on conflict avoidance under visual flight rules in … airspace where aircraft [pilots] may not be receiving ATC [air traffic control] separation services. … We’re looking at conflict avoidance, keeping the aircraft well clear of each other [so] as to not present a collision hazard. We are not necessarily focused on collision avoidance, and [there] are some real differences between TCAS [traffic-alert and collision avoidance systems] and the [LD-CAP] activities. … TCAS is an alert to the pilot; it is not an automatic [aircraft/UA] maneuver.”

Each aircraft was equipped with a Garmin GDL 90 ADS-B datalink transceiver with a 978-MHz universal access transceiver [UAT] link. “There is every reason to believe that the algorithms would work with … a UAT [link] or the 1090-MHz extended squitter link,” he said. One algorithm was supplied by MITRE and the other was supplied by UND, each with a series of modifications. RTCA Special Committee 203 and other standards bodies will consider these and other data in producing a set of sense-and-avoid technical standards.

LD-CAP’s agenda covers the development and testing of algorithms that rely on ADS-B; identifying methods of commanding UAS sense-and-avoid maneuvers to avoid conflict with manned aircraft that lack ADS-B; education of the general aviation community about ADS-B benefits in this context; and reducing the size, weight and cost of ADS-B equipment.

The second item on this agenda will consider the feasibility of creating and transmitting alternative messages equivalent to ADS-B messages. Currently, aircraft transponder-reply targets appear on ATC displays in response to secondary-surveillance radar interrogations, and then “aircraft tracks are uplinked to ADS-B-equipped aircraft via TIS-B, traffic information surveillance–broadcast messages,” Lacher said. “But that only applies to aircraft that are transponder-equipped. We’re looking at a capability that will allow aircraft that are only being tracked with primary radar to have TIS-B messages. … That [testing] is planned for spring of 2013.”

The flight data animation replays showed algorithms commanding the autopilot of the SR22 surrogate UA to turn well clear to avoid conflicts with the intruder airplane. UA maneuvers not replayed at the briefing included climbs, descents and speed adjustments for successful avoidance of the intruder, Lacher said.

LD-CAP has addressed both the see/sense-and-avoid and lost-link issues surrounding UAS integration. “That’s a big deal,” said Mark Askelson, associate professor, Department of Atmospheric Sciences, University of North Dakota (UND). “[LD-CAP is] open to testing algorithms from anyone who is ready to do it.” In their work on these algorithms, he noted, UND’s “students have been able to develop a concept and see it to the point of being flight-tested.”

Flight testing essentially helps to assess the sense-and-avoid technology readiness level in winds and atmospheric conditions, said Frank Jones, LD-CAP deployment lead, NASA Langley. Before the flight tests, computer simulations already had analyzed more than 2 million encounters between a virtual UA and a virtual intruder aircraft, he said. During the flight tests, the latest data usually were reintroduced the same day to computer simulations, and the algorithms were modified and uploaded for the next day’s flight tests. MITRE algorithm-evaluator software also generated technical “report card” assessments.

“Essentially, what we have is a general purpose computer [aboard the SR22 UAS surrogate] that interfaces to the autopilot such that we can fly the aircraft [using] a research operator in the back seat of the aircraft,” Jones said. Playing the role of a pilot in a ground control station, the back seat pilot issued commands to the [UAS] autopilot for heading, speed, climb/descent and altitude changes. Rick Yasky, chief pilot, NASA Langley, occupied the front left seat as safety pilot of the SR22 with capability to override the UAS systems and fly the aircraft if required.

Another mitigation of the risk of in-flight collision during these experiments was operation of the Cessna 172 intruder only at real altitudes intentionally biased/offset by 2,000 feet from the altitude of the SR22 UAS surrogate. Data used by the algorithm, however, compensated for the offset. Therefore, the algorithm responded automatically to a “ghost aircraft” — that is, as if the C-172 were a physical intruder closing at the same altitude, overtaking or engaged in another of about 20 conflicting geometries.

What triggered the automatic maneuvers was the algorithm’s protective logic, in which the C-172 was continuously surrounded by protected volume of “no-fly zone” airspace — 500 ft above/below and 2 nm (3.7 km) laterally. The boundaries of this volume also were the procedural intervention point at which the safety pilot in the SR22 surrogate UAS was required to immediately take over the SR22 flight controls. The C-172 pilot similarly was required by procedure to avoid penetration of its protected volume.

In replays of cockpit video, the SR22 pilots monitored the ownship in the center and the relative positions of other aircraft, in part, on a navigation display. As the SR22 UAS surrogate overtakes the ghost C-172 in one encounter, “The [C-172] traffic with yellow highlighting means that the algorithm has triggered, sensing a potential conflict,” said NASA Langley’s Yasky. The pilots saw immediately on a separate display commands and triggers that the algorithm was generating and sending to the autopilot, and the SR22 turned as expected to avoid the conflict.

Testing the algorithms’ ability to automatically and simultaneously avoid conflicts between a UAS and multiple intruder aircraft also will be part of LD-CAP, possibly in June 2013. Lacher noted that the LD-CAP partners have coordinated their work with parallel research under way by others into complex UAS sense-and-avoid technology, sensors and algorithms.

LD-CAP work is one element of UAS integration, and is distinctive in looking at solutions that do not directly address the UAS pilot. “The research we’re focused on [applies] if the [UAS pilots are] unable to execute a maneuver,” MITRE’s Lacher said. “First, you want to provide information to [pilots] to give [them] situational awareness. They may make the decision to maneuver the aircraft before our automatic alert would trigger a maneuver. That would be the preferable mechanism. But …maybe they didn’t receive the [traffic-avoidance] alert for some reason, or maybe the link between the ground control station … and the aircraft is disrupted or interfered with in some way. You want the aircraft to still be able, in a self-contained manner, to maneuver by itself to avoid those collisions. So you might have multiple sensors, some of those sensors feeding data to the pilot on the ground, other sensors feeding data to algorithms … on board the aircraft that would [command an autopilot] maneuver.”