NTSB Factual Report

Note: This case was reclassified from an accident to an incident as a result of applicable revisions to 49 Code of Federal Regulations Part 830.2, as amended at 75 FR 51955, Aug. 24, 2010. The case was previously identified under accident number DEN08FA130.

HISTORY OF FLIGHT

On July 28, 2008, at 1231 mountain daylight time, an experimental Raytheon Cobra, unmanned aircraft system (UAS), N601RN, sustained substantial damage when it collided with a stadium light pole while maneuvering for a pre-programmed landing near Colorado Springs, Colorado. The aircraft was registered to and operated by Raytheon Missile Systems Advanced Programs of Tucson, Arizona. There were no injuries to persons on the ground. Visual meteorological conditions prevailed for the Title 14 Code of Federal Regulations Part 91 test flight, and no flight plan was filed. The aircraft had departed from a paved road paralleling one of the athletic fields on the grounds of the United States Air Force Academy at 1208, and was returning to that location when the accident occurred.

The purpose of the flight was to test the system before demonstrating the Cobra to Air Force Academy staff. The staff at the Air Force Academy wanted to evaluate the aircraft's potential as part of an advanced airmanship curriculum course on UAS for the cadets. Prior to the flight, the Raytheon flight test team surveyed the road, the athletic fields, and the terrain north and south of the takeoff area. They then programmed the Global Positioning System (GPS) flight profile waypoints into the autopilot.

The aircraft underwent a preflight and systems tests. The aircraft's engine was started at 1206. It was also at this time that the aircraft's left wing-mounted camera was activated, which recorded the 25 minute, 5 second flight. Following automatic and manual engine run ups, the aircraft was taxied to a position on the road from which the command for an automatic takeoff was entered. The aircraft took off to the north and climbed to 8,000 feet above mean sea level (msl) or 1,000 feet agl. It then turned east to enter a counter clockwise, east-to-west rectangular flight pattern profile over the athletic fields north of the Air Force Academy's cadet area, as programmed. The aircraft then climber to 9,000 feel msl (2,000 feet agl) and continued the rectangular patterns.

The Cobra UAS performed several rectangular patterns over the athletic field with no problems noted. Part of the demonstration showed that the UAS could fly on its pre-programming in the event of a "lost-link" or loss of signal between the ground station and the aircraft.

After the aircraft had performed all of its maneuvers, it entered a north upwind leg for a right hand traffic pattern. The aircraft made several patterns at 8,000 and 7,600 feet before it was descended on the downwind leg for landing. The south-to-north landing pattern was set up so that a downwind to dogleg, dogleg to base, and base to final pattern would be flown. During the turn to final for landing, the aircraft undershot the final approach. The supplemental pilot and observer both called for a "wave off" (abort) and the internal pilot (aircraft operator in the truck manning the control console) selected the "abort" command, which sent the aircraft to the upwind waypoint to re-enter the traffic pattern. The aircraft established itself on crosswind and then flew a descending downwind, dogleg, base leg and turn to final. During the turn to final on the second landing pattern, the aircraft undershot the turn again. The supplemental pilot and observer both called for an abort and the internal pilot again selected the abort command and the aircraft again proceeded to the upwind waypoint. On the third traffic pattern, the aircraft overshot the final turn waypoint flying approximately 89 feet west of flight path centerline. The Raytheon supplemental pilot elected not to take over manual control of the UAS because both he and the observer thought the aircraft would correct back to the programmed flight path and miss the stadium lights. The aircraft's wing-mounted camera showed the UAS fly west of and above a first of three 80-foot tall stadium lights that aligned the east side of a soccer field. The camera then showed the UAS in a 45-degree bank descending turn, heading toward the middle set of stadium lights. The last image recorded by the aircraft's wing camera had the stadium lights within feet of the camera. The aircraft impacted the top of the lights. The aircraft's right wing fractured and remained in the top bank of lights. The remainder of the aircraft fell to the ground impacting terrain within feet of the base of the light pole. The right wing was subsequently dislodged by the wind and fell to the ground.

AIRCRAFT INFORMATION

The aircraft was serial number 001. It received an experimental airworthiness certificate from the FAA on September 29, 2006. The aircraft was being operated under a special program letter authorizing Raytheon to fly the Cobra over the U. S. Air Force Academy.

There was no Certificate of Authorization (COA) for the Academy demonstration. The flight was approved via the experimental certificate program letter. The flight was a civilian flight and not operated by the Department of Defense. FAA personnel were present at the demonstration.

The aircraft was one of nine aircraft in the Raytheon Cobra fleet. The aircraft was developed as a low cost, reliable test platform for numerous sensors, communications packages, and light weapons testing. The vehicle was also used to demonstrate UAS architectural concepts in the real world environment.

The aircraft was 9 feet long with a wingspan of approximately 10 feet. The aircraft's empty weight is 65 pounds. The maximum gross weight for the aircraft as cleared by the manufacturer is 110 pounds. The gross weight of the aircraft on that day was 79.9 at takeoff. The center of gravity was at 29.2 inches with a mean aerodynamic chord of 26.3 percent, placing it well within its operating envelope.

A Desert Aircraft DA-150 2-cycle, horizontally opposed, carbureted 16.5 horsepower engine powered the aircraft. The aircraft operated within an airspeed regime of 45 to 55 knots for climbs, cruise, descent, and landing. The stall speed for the aircraft was approximately 38 knots.

Guidance and navigation for the aircraft was provided by a Piccolo II autopilot using a Piccolo command center, a laptop software program where coordinates and actions inputted were uploaded by omni radio to the aircraft's autopilot and integrated with the onboard GPS receiver. A remote control box with manual power and steering controls served as a backup system to the command center.

The aircraft underwent a condition inspection on July 18, 2008. At that time, the aircraft had 25.6 hours total flying time. Prior to the accident, the aircraft had flown 61 times. It had logged a total flight time of 28.4 hours.

PILOT INFORMATION

The aircraft was automatic piloted on a pre-programmed schedule. A company-trained internal pilot monitored the aircraft's progress through the Piccolo command center, located under a pop-up shelter adjacent to a medium-sized moving truck that was used to transport the Cobra UAS and its support equipment. The internal pilot also monitored the aircraft's wing-mounted camera display. In accordance with the FAA-issued special program letter, the internal pilot held at least a private pilot certificate, a third class medical with no limitations. According to the internal pilot, he had approximately 60 hours operating Cobra UAS aircraft.

A supplemental pilot held the manual control box for the aircraft. He was positioned along the west side of the north-south road near the start point along with an observer. The supplemental pilot possessed no flight ratings, but met the criteria for his position based on the FAA special program letter. The observer held a private pilot certificate and met the criteria for his position based on the FAA special program letter

For this flight, a test director was positioned near the internal pilot. All flight crew members were within talking/shouting distance of each other; therefore, no radio communications were used.

METEOROLOGICAL CONDITIONS

At 1226, weather conditions at the Air Force Academy Airfield (AFF), 6 miles south, southeast of the accident scene were few clouds at 12, 000 feet, 10 statute miles visibility, wind conditions were 360 degrees at 12 knots, temperature was 86 degrees Fahrenheit (F), dew point 39 degrees F, and altimeter of 30.06 inches.

At 1255, weather conditions were clear skies, 10 statute miles visibility, wind conditions 020 degrees at 11 knots with gusts to 20 knots, temperature 87 degrees F, dew point 39 degrees F, and altimeter 30.06 inches.

The field elevation of the Air Force Academy Airfield is 6,572 feet. The elevation of the Air Force Academy athletic fields where the flight was conducted was 7,010 feet msl.

The density altitude based on the conditions recorded at AFF at 1226 was 10,427 feet msl.

WRECKAGE AND IMPACT INFORMATION

The aircraft wreckage came to rest approximately 15 feet east of the light pole that it impacted. The left wing was broken aft at the wing root and carry-through spar. The forward body of the aircraft was cracked and broken aft of the spinner to the payload bay. The nose gear was broken aft. The right main gear was broken off from the bottom of the payload bay. The right wing was broken downward but remained attached to the carry-through spar. Several areas along the leading edge showed aft crushing. The right aileron was separated at the outboard and middle hinges. The tail boom was broken from the aft portion of the payload and main body of the aircraft. The empennage was intact and showed no damage. The propeller spinner was bent aft and inward. The three-bladed composite propeller remained intact.

The aircraft batteries were disconnected and the wreckage was recovered to a location near the Colorado Springs Airport. The NTSB examined the aircraft at the removed location at 1830.

The aircraft's internal fuel tanks, avionics, communications, and navigation components showed no damage. The engine was examined and showed no external signs of anomalies. The propeller was turned. The engine crankshaft moved freely.

The aircraft's flight controls were examined for continuity. The surfaces are driven by electric motors, which in turn drove actuator arms that are connected to the control surfaces. All of the controls showed connectivity from the motors to the surfaces, and all surfaces displayed free movement and no binding.

The Piccolo autopilot was removed. The actual flight profile recorded by the Piccolo software was captured and downloaded. Both the autopilot and data files were retained for further examination.

TESTS AND RESEARCH

The Piccolo software data file from the accident aircraft was analyzed at the software's manufacturer on August 18, 2008. The data revealed no systems anomalies with the aircraft. The data showed that the software properly commanded the aircraft autopilot to maneuver the aircraft over the programmed waypoints at the proper altitudes and airspeeds. The data also showed that the aircraft experienced a 20-knot tailwind at pattern altitude. This was determined by comparing aircraft recorded airspeed against GPS recorded ground speed.

From examining the data file, the aircraft showed several instances of overshooting the programmed track as it transitioned through waypoints. The abnormal tracking was determined to be a result of software and gains parameters being inadequate to compensate for the winds at the aircraft's flight altitude and the programmed aggressive profile. This showed itself in several of the turns during the flight profile at altitude. Early in the flight profile these instances only happened occasionally. However as the flight continued, more instances of maximum banking was noted. The aircraft was left at altitude to fly several landing patterns to assess profile performance. The team decided to let the aircraft descend and make a landing approach. On the turn from base to final, the aircraft overcorrected resulting in the aircraft undershooting the first final turn for landing. The aircraft performed similarly during the second pattern resulting in the aircraft undershooting the turn to final. On the third approach, the aircraft made an abrupt rollout over the dogleg-to-base waypoint. It then rolled into a 45-degree bank, which was the autopilot-limited maximum bank angle. The airplane overshot the final turn waypoint and began a descending correction back to the final approach path, during which the aircraft struck the light pole.

The software manufacturer stated, "The large tail wind on downwind, combined with the excessive speed due to the steep approach is what causes the vehicle to switch early to base legs ... the approach is simply too aggressive, both too steep and too short ... The aircraft is constantly pulling max bank angle and then has to snap upright. It just can't do it fast enough to satisfy the aggressive tracker ..." This resulted in the aircraft being unable to fly the proper ground track.

The Raytheon team visited the Air Force Academy on July 7, 2008, to plan and coordinate the demonstration. While there, they examined the area where the flight would be conducted. Concerns were raised over some of the obstacles the aircraft would have to negotiate. The obstacles included trees along the road from which the aircraft would takeoff and land, rising terrain to the north of the takeoff/landing road and demonstration area, rising terrain and buildings south of the takeoff/landing road, and the stadium lights that were south and west of the takeoff/landing road. The team determined they would have to modify the flight profile for the Cobra to meet the constraints of the area over the Air Force Academy athletic fields.

Changes to the Cobra's flight profile to meet the constraints posed by the Academy's flight area included tighter patterns, a slightly steeper pattern a landing descent of 8 degrees versus the 3 degrees they normally flew, and the addition of a dog leg waypoint between the downwind and base legs of the landing pattern.

In preparation for the demonstration at the Air Force Academy, the Raytheon team conducted 24 flights of the Cobra at the Unmanned Vehicles International, Inc (UVI) flight test range near Whetstone, Arizona, to evaluate the Cobra's performance using the flight profile they planned for the Academy demonstration. The flights were conducted on July 17 and 23, 2008. According to maintenance records for the aircraft, 7 of the flights were conducted on July 17, and 16 flights were done on July 23. The aircraft flew total times of 2.0 and 2.7 hours respectively.

The elevation of airstrip at UVI is approximately 4,200 feet msl. Early morning temperatures on those days approximated 81 degrees F at 0755 on July 17, and 75 degrees F at 0755 on July 23. The aircraft performed well during all 24 flights.

Later analysis of the Piccolo data file for the UVI flights showed that in the last nine landing profiles, two of the approaches overshot final in a manner similar to the third pattern at the Air Force Academy. However, the magnitude and relative position of the overshoots was not fully understood until the data were analyzed after the accident occurred.

The Raytheon team planned to survey the flight test area on July 27 in preparation for the July 28 demonstration. On-going athletic activities at that time precluded the team from using that time to set up. The team returned on July 28 at 0645 to set up.

The team unpacked and set up their equipment. At 0840, initial control checks were done. The engine run was done at 0845. The engine ran fine and no adjustments were made. A pre-flight briefing was held, and the team was ready for engine start at 1206. A member of the team stated that the briefing was abbreviated due to the compressed schedule, but all the key points were covered.

During the third approach for landing when the aircraft overshot final, the supplemental pilot and the observer both thought the aircraft would not h

NTSB Probable Cause

The flight team's failure to program the UAS with flight-tested parameters that could tolerate the high density altitude and tailwind conditions encounted during the flight, which resulted in two undershoot approaches, and the existence of an undiscovered software anomaly that resulted in a momentary loss of control and overshoot of the final approach course, followed by a steep descent and subsequent collision with a light pole.