Computers in the Sky

Designers use simulation tools to develop and test autonomous controls for unmanned aerial vehicles.

By Carl A. Bargainer, Ronald D. Knuppel, and David A. Ogden     image of PDF button


Authors (from left) Carl A. Bargainer, Ronald D. Knuppel, and David A. Ogden, members of the Unmanned Systems and Sensors Section in SwRI's Aerospace Electronics and Training Systems Division, have been involved in the design of UAVs since 1994. They are specialists in the design, simulation, and testing of robust, low-cost, control systems for UAVs.


Unmanned aerial vehicles are machines that fly with no one aboard. They vary in size from a private jet down to something that could fit in the palm of your hand. Some are flown using remote control and some are fully autonomous, using an on-board computer to take off, fly, and land.

Manned, heavier-than-air flight is relatively new -- less than 100 years old. The first guided, unmanned aerial vehicle (UAV) was a flying bomb, the Sperry Aerial Torpedo, which was part of an experimental program for the U.S. Navy during World War I. During World War II, Germany launched thousands of V-1 "Buzz Bombs" at England. Since that time, the primary use of UAVs has been to serve as aerial targets or drones to be shot down by military pilots and anti-aircraft gunners during training missions. Most recently, the U.S. military has flown UAVs in the form of cruise missiles and as reconnaissance planes used in Iraq and Kosovo. The UAV reconnaissance flights have helped avoid casualties among pilots.

UAVs also fill an important and growing role in the civilian aviation industry. Many jobs being performed by manned aircraft are dangerous, monotonous, or very expensive. It has been estimated that over the past five years, on average, eight deaths have occurred annually in the geophysical survey industry, where pilots fly their instrumented aircraft over long routes, close to the ground, and over severe terrain. The most common type of geophysical survey accident is categorized as "controlled flight into terrain" in which no mechanical failure causes the crash.


The Outrider's central processor is linked to various onboard avionics and flight sensors related to flying, navigating, and landing the aircraft, as well as operating its payload systems.


Regardless of the mission, precise guidance is imperative for a UAV. Engineers in the Avionics and Training Systems Division of Southwest Research Institute (SwRI) have developed airborne central processor (ACP) computers that perform flight control, guidance, and flight management functions on UAVs. SwRI has developed and flown these computers on two UAVs -- the Outrider tactical UAV, a 500-pound, twin-wing reconnaissance aircraft developed by Alliant Techsystems for the U.S. Army, and a 600-pound parafoil aircraft developed for the U.S. Marine Corps, NASA, and other clients.

SwRI performed complete turn-key development of the central processors, including hardware design and fabrication, flight control law design, flight control performance and stability analysis, software design and coding, software and hardware-in-the-loop simulation, software and hardware acceptance testing, and flight testing. The ACP hardware design uses many commercial off-the-shelf industrial-grade components to produce a lightweight, rugged, and low-cost avionics computer for UAV applications.


An Institute-developed test stand simulates flight, navigational, and powerplant data to test the airborne central processor (gold box at center), which is used in unmanned aerial vehicles.


Flight Control Design

The autopilot in a UAV has a challenging job. It must replace the functions a human pilot would normally perform, such as detecting and controlling aircraft attitude, speed, and position while observing the performance limits of the aircraft. The ground operator must be able to re-direct the aircraft at any time. Automated takeoff and landing require accurate position control and seamless sequencing through various maneuvers. The ACP must also detect and respond autonomously to emergency conditions such as loss of radio uplink from the ground operator, loss of global positioning system (GPS) navigation data, and engine failure.

For tactical unmanned aerial vehicles, the need for human piloting, whether from outside or inside the craft, is eliminated. The vehicle essentially flies itself, while an air vehicle operator directs the flight path using a mouse on a digital map display. Automated takeoff and landing are essential elements in achieving this goal. Training pilots and maintaining their proficiency is costly to the military because it requires UAV flight time and risks UAV assets.

The ground operator directs the UAV flight path with a waypoint table. Simple waypoints specify locations in three-dimensional space for the UAV to fly through. Location is specified by latitude, longitude, and altitude. More complex waypoints define line and arc paths to be followed through three-dimensional space. The ground operator preloads a mission in the waypoint table and updates it as required during flight to redirect the UAV. There are special emergency waypoints that tell the UAV where to go when radio uplink from the ground operator fails.

As with any manned aircraft, the most demanding control requirement for a tactical UAV is position accuracy at touchdown during landing. In a manned aircraft, this complicated maneuver requires much skill, training, and practice. In the UAV automated landing function, a vertical velocity controller was used with GPS and inertial sensors to achieve vertical errors less than 4 feet with atmospheric turbulence. Touchdown dispersions are about 80 feet longitudinally and less than 3 feet laterally. During normal flight operations at altitude, a pitch attitude controller is used instead of vertical velocity because position control errors can be much larger, but excursions in pitch attitude and throttle must be minimized.

The following maneuvers are performed during automated landing, each having its own unique control objectives and control laws. On final approach, the flight path control holds the vehicle on the glide slope line, controlling the rate of descent. During the flare maneuver just before landing, which increases the pitch of the UAV, the commanded flight path is altered from a straight slope to a vertical arc to reduce the descent rate. About one second before touchdown, the decrab control is activated to remove the "crab" angle that is used to keep the vehicle on the approach path in a crosswind. The decrab maneuver aligns the vehicle to the runway and minimizes tire skid. At touchdown, the engine is killed and path control is switched to ground steering. About one second after touchdown, braking is initiated with a closed-loop brake controller. All mode transitions must be managed without significant controller startup transients.

The autopilot software places a limit on minimum airspeed to maintain stall margin and on maximum airspeed as necessary to achieve altitude control targets with available engine performance. A dynamic lift limit controller is used to prevent stalls in maneuvers or in turbulence. An angle of attack sensor was not available on the aircraft, so instantaneous coefficient of lift is computed, based on dynamic pressure and normal acceleration.

The ACP performs continuous air vehicle state estimation. The state estimator blends two types of inertial data, automated landing data from GPS and pressure altitude data, to produce a single robust state estimate with less than 10 milliseconds of latency. This state estimate includes roll, pitch and yaw rates, attitudes, accelerations, velocities, and position, and is used for all flight control feedback.

Flight management functions performed by the ACP include autonomous operation of the radio datalink such as selecting transmitter power and antennas, plus directional antenna pointing. The ACP also manages the state of other complex subsystems such as the inertial measurement unit and GPS.

Hardware and Software Design

The airborne central processor hardware is based on commercial temperature-range components rather than full military-specification components. This facilitated a faster development cycle, and allowed avionics computers to be produced for approximately one-third the cost of full military-specification equipment, even with ruggedization costs included. Previous experience with full military avionics designs for the U.S. Air Force provided the insight necessary to successfully ruggedize commercial hardware for this application.

Techniques for ruggedizing commercial components included thermal management, conformal coating, component selection, supervision of vendors, environmental stress screening, and thorough acceptance testing of the complete ACP computer. On future programs it may be cost effective to test incoming parts supplied by commercial vendors.

The operational flight program (OFP) software which runs in the ACP is written in MIL-STD-1815 Ada. The Ada kernel provides real-time task scheduling and multi-tasking features, because the ACP operating system is DOS. Common Ada software modules and architecture are used in ACPs for both tactical UAVs and parafoils.

Simulation and Testing

Aircraft crashes are the number one cause for cancellation of UAV development programs. Therefore, a design approach was required that would verify the performance of the autopilot with an extremely high level of confidence before the aircraft ever left the ground. This was accomplished by creating two levels of simulation: software-in-the-loop (SIL) simulation and hardware-in-the-loop (HIL) simulation.

For SIL simulation, both the plant and the controller are modeled by software. The plant includes the aircraft aerodynamics, mass properties, propeller, engine, tires, brakes, servos, sensors, and other aircraft subsystems. These components are modeled separately in software and integrated to create a simulated operating environment for the ACP. Effects of external conditions such as atmospheric temperature, barometric pressure, winds, turbulence, and runway surfaces, are also modeled. For HIL simulation, only the plant is modeled by software, but the controller is the actual ACP hardware running the OFP software.

SIL simulation is used to test control loop algorithms before hardware is built or production OFP software is written. It can also be run at non-real-time speeds, either to closely study the control response with high resolution and slow speed, or to accelerate testing with faster than real-time speeds. It is also used to verify controller stability margins, to adjust control gains, and to evaluate controller performance.

HIL simulation is used for integration testing of ACP software and hardware, formal acceptance testing of the OFP software, and integration testing of the ACP with other system components such as the ground station. HIL simulation has also been used for pilot and operator training. Formal OFP acceptance testing verifies simulation results for predefined flight scenarios against requirements and SIL simulation results.

The HIL simulator is augmented with additional test equipment to permit automated hardware acceptance testing of ACPs. Hardware acceptance testing verifies that the ACP hardware functions correctly and performs within tolerances, without the OFP software.

In the tactical unmanned aerial vehicle program, SwRI software and hardware demonstrated a high level of quality and reliability in aircraft operations. In over 250 sorties, no crashes were caused by flight control failures. During the Army Systems Capability Demonstration in November 1999, the ACP logged 41-plus flight hours in 19 sorties with no flight control anomalies during automated takeoff, automated landing, or in-flight operation.


A flexible, parachute-like fabric wing called a parafoil allows compact storage and ease of portability. The aircraft's forward motion inflates the wing for stable, low-speed flight. The gray compartment at the front of the vehicle contains the flight control processor and sensors that allow the unmanned vehicle to fly to a predetermined location, drop a payload package, and then return for an automated landing.


Regulatory Environment

Commercial application of UAVs within the United States is restricted because of their regulatory status with the Federal Aviation Administration. At low altitudes away from controlled air spaces such as airports, "see-and-be-seen" flight rules apply. These are the air spaces in which UAVs need to operate for many potential applications. For UAVs, "see-and-be-seen" has been interpreted to mean that the UAV must be kept within sight of an observer either on the ground or in another aircraft. This is not an issue for military applications or flight operations on military ranges because they do not fall under FAA jurisdiction.

Future Directions

Technology has become available to make practical the widespread use of UAVs. The shrinking size and reduced cost of electronics allows new applications never before possible. These include inspecting electrical transmission lines, oil and gas pipelines, and airborne telecommunications relays in countries that lack other required infrastructure, and geophysical surveys for oil and mineral companies. This technology can also be used for supply drops in remote areas, mine detection in war zones, gas monitoring over active volcanoes, search and rescue, and temperature monitoring over the ocean. Even applications in commercial air transportation are being considered -- although that might stretch the definition of "unmanned."

Comments about this article? Contact Bargainer at (210) 522-5656 or cbargainer@swri.org, Knuppel at (210) 522-3461 or rknuppel@swri.org, and Ogden at (210) 522-6928 or dogden@swri.org.

Bibliography

Cradle of Aviation Museum: www.cradleofaviation.org

"Safety in Airborne Geophysical Surveying," The Leading Edge, pps. 635-637, May 1999

Published in the Spring 2000 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Maria Stothoff.

Technics Spring 2000 Technology Today
SwRI Publications SwRI Home