MOP 3: 3-TPA Ground Miss Distance

The 3-TPA ground miss distance was calculated using a MATLAB script which accepted the aircraft virtual navigation solution (downstream of the position and course slewing tool) as an input, and returned the terrain miss distance as an output. The script determined the minimum miss distance by using the Root Mean Square (RMS) error equation to calculate the distance between the aircraft and the DTED posts at every iteration point of the navigation solution for the duration of the escape maneuver. Eq. (4.2) shows the calculation.

RMSerror=

p

(XDT ED XAC)2+(YDT ED YAC)2+(ZDT ED ZAC)2 (4.2)

The minimum miss distance was the least of these values. This script was validated by taking sample data to plot the aircraft trajectory over a DTED map, and confirming that the miss distance was correct. The evaluation criterion was satisfied if the distance between the navigation solution and the DTED was more than the specified safety bubble radius at all points of the escape flight path. Though a distance found to be inside the bubble radius would not necessary indicate an impact with terrain, the fact that the aircraft model assumes a point mass makes further interpretation of the distance irrelevant. The specifics of the analysis can be further referenced in Appendix B.

The 3-TPA ground miss distance results are presented in Table 4.3, tabulated with one test run per container. The flight test results are also presented graphically in 3-dimensional

intersected the terrain. The data show that in all cases but one, the aircraft impacted the simulated terrain. The two primary causes for the ground impacts are the achieved flight path angle, and the open-loop path selection logic. Of note, the available reaction time is not included in Table 4.3 because the values are all essentially zero since no time was practically available before impact (even in the case with a 39 ft miss distance).

Table 4.3: 3-TPA Ground Miss Distance

Algorithm: 3-TPA Test Dates: 31 Aug-10 Sep 15 Pressure Altitude: 15,000 ft Airspeed: 310 kts Groundspeed

Ground Miss Distance (ft) Test

Point Flight1 Flight2 Flight3 Flight4 6 (FWD) 0 39 0 0 7 (Left) 0 0 0 0 8 (Right) 0 0 0 0

The algorithm assumed that the aircraft would be able to climb at a constant flight path of 15 , and the maneuver initiation point is based on this assumption. As discussed in the previous section, the aircraft was not able to maintain 15 of flight path angle for the duration of the maneuver, which resulted in less altitude gained than predicted, and the aircraft could not out-climb the terrain obstacle. This caused the aircraft to impact the terrain in three out of four test runs, as illustrated in Figure 4.4. The only case where test point 6 did not impact the ground is the second test run, which produced the highest flight path angle, due to the aircraft weight and center of gravity location. Even in this case, the average flight path angle was lower than planned, and the aircraft missed the ground by only 39 feet. The important point though, is that the aircraft must meet or exceed the algorithm’s expected performance or terrain clearance cannot be guaranteed.

ESCAPE Test Point: 6 (FWD Path) Average OAT: -8 C

Test/Virtual Altitude: 15,000/11,500 ft Test Dates: 31 Aug-10 Sep 15

Figure 4.4: Planned vs Achieved Flight Path Angle for Test Point 6, Forward Path

Another cause for ground impact during test points 7 and 8 was the fact that the algorithm stops scanning terrain once it implements a maneuver. In order to avoid nuisance activations, the algorithm did not command any maneuvers until it predicted that all projected escape paths would impact the terrain. Once every projected path impacted the terrain, the algorithm determined which path o↵ered the longest time of flight prior to impact, and commanded the aircraft to fly that maneuver with a time delay safety margin to avoid actual impact. The algorithm assumed that the escape path which o↵ered the longest time of flight prior to impact would o↵er the best chances to avoid the terrain by taking advantage of the time delay safety margin and the bubble safety margin. However, in some cases, such as the one depicted in Figure 4.5, the safety margins would have to increase to impractical values in order to protect the aircraft using this logic, since the left level turn

would have resulted in a collision even if it had been initiated 50 iterations earlier than the actual initiation point. (To be clear, this issue is somewhat artificial in nature since the aircraft was slewed to a location deep in the mountains, where realistically, it may not have been practical (or possible) to physically fly the aircraft without causing an activation that would have cleared the terrain. In any case, the presented problem shows a limitation in the algorithm’s performance that can be directly addressed.) The algorithm logic is illustrated in Figure 4.5, using data from test point 7.

ESCAPE Test Point: 7 (Left Level Turn) Average OAT: -8 C

Virtual Altitude: 12,000 ft Test Dates: 31 Aug-10 Sep 15

Figure 4.5: Ground Impact Predictions for Test Point 7

Figure 4.5 shows that as of the first iteration, the left level turn was predicted to intercept the terrain after 23 seconds time of flight, a byproduct of the slewed position. After 26 iterations, the right level turn was predicted to intercept the terrain after 12 seconds, and the left level turn was still predicted to intercept the terrain after 20 seconds. After 50 iterations, the forward climb was predicted to intercept the terrain after 17 seconds,

the right level turn was predicted to intercept the terrain after 10 seconds, and the left level turn was predicted to intercept the terrain after 22 seconds. At this point, the algorithm commanded the left level turn because it o↵ered the longest time of flight prior to impact. However, the left level turn was predicted to impact the terrain for more than the last 50 iterations, corresponding to more than 1.5 seconds. This caused the aircraft to choose an escape path which predicted that a collision would occur even if the maneuver had been initiated well prior to the expiration of the time delay safety margin.

With this flight test result, additional simulator test points were flown to evaluate the possibility of allowing the algorithm to continue to evaluate the terrain during the actual execution of one of the TPAs. This was done by simply restarting the algorithm immediately upon TPA execution. Through this limited analysis, it was found that the algorithm would alter its flight path and TPA choice during execution and avoid the terrain. Through this rudimentary experimentation, the algorithm proved to be even more robust. Unfortunately, it was not possible to test these points in flight due to flight safety restrictions for software changes to the ESCAPE paths. This does, however, show promise for future research and testing.

Although the e↵ects of wind on the flight test data were not specifically analyzed, a logical assumption would be that it also had an e↵ect on the ground miss distance. Theoretically, a no wind assumption would reduce the accuracy of the predicted turn radius. This would cause the downrange travel to be smaller than predicted with headwind, and longer than predicted with tailwind. Additionally, the turn radius would vary throughout the turn, depending on whether the aircraft was flying into or away from the wind.

Another factor which a↵ected the ground miss distance was the lag between the command vector and the aircraft response. The load factor and bank angle time history plots showed that on average, the aircraft’s load factor and bank angle started to increase approximately 0.25 seconds after the command vector was initiated, which is equivalent to

half of the time delay safety margin, or 130 ft of downrange travel at a groundspeed of 310 kts. For this reason, aircraft performance, inertia, and center of gravity have a direct a↵ect on the algorithm’s performance and must be accounted for within the time delay safety margin. Since this lag is typically very small, on the order of a tenth of a second, it need be only evaluated at a worst-case condition and applied throughout the individual aircraft’s flight envelope.