Flight Trajectory Planning for Fixed Wing Aircraft in Loss Of Thrust Emergencies

Transcription

1 Flight Trajectory Planning for Fixed Wing Aircraft in Loss Of Thrust Emergencies S A S WATA PA U L F R E D E R I C K H O L E * A L E X A N D R A Z Y T E K C A R L O S A. VA R E L A W O R L D W I D E C O M P U T I N G L A B O R A T O R Y D E P A R T M E N T O F C O M P U T E R S C I E N C E * D E P A R T M E N T O F M E C H A N I C A L, A E R O S P A C E, A N D N U C L E A R E N G I N E E R I N G R E N S S E L A E R P O L Y T E C H N I C I N S T I T U T E S e c o n d I n t e r n a t i o n a l D D D A S C o n f e r e n c e M a s s a c h u s e t t s I n s t i t u t e O f Te c h n o l o g y A u g u s t 7,

2 Motivation Loss of thrust emergencies may be induced by bird strikes, fuel exhaustion, etc. Image courtesy: dailymail.uk Pilots need to make quick decisions and choose a path to land the aircraft safely on the ground. This calls for an efficient system to help the pilots choose the best course of action by calculating available options and displaying the same to the pilots. 2

3 Air France Flight 447 June 1 st 2009, Flight 447 from Rio de Janeiro to Paris Thunderstorm caused airspeed sensors (pitot tubes) to ice and fail Autopilot system not able to deal with data failures---disengaged Pilots unable to react to erroneous data in a timely manner, eventually stalling the plane into the Atlantic Ocean rt.final.en.php 4/4a/Air_France_Flight_447_path.png 3 8/26/2017

4 Data Redundancy Primary cause of the AF447 accident: incorrect airspeed Airspeed could have been recomputed from ground speed and wind speed Take advantage of data redundancy between independently produced inputs airspeed wind ground speed wind speed ground speed = airspeed + wind speed 4 8/26/2017

5 Air France AF447 PILOTS Demo 5 8/26/2017

6 x Tuninter 1153 Flight Accident Flight from Bari, Italy to Djerba, Tunisia on August 6 th, 2005 ATR-72 ditched into the Mediterranean sea 16 of 39 people on board died Bari, Italy Actual route Planned route Palermo, Italy Djerba, Tunisia Final Accident Report for TS-LBB Mayday TV Series on Tuninter

7 Initial Cause of the Accident Incorrect fuel quantity indicator (FQI) installment FQI for ATR-72 was not working properly (LED failure) Technicians replaced the FQI with one designed for ATR-42 FQI showed 2,700 kg of fuel, but fuel actually weighed 550 kg Pilots did not realize data error eventually leading to fuel exhaustion Final Accident Report for TS-LBB FINAL%20REPORT%20ATR%2072.pdf 7

8 Complex Dependencies Between Data Streams Air France 447 Model v g : ground speed v w : wind speed v a : airspeed v g v w v a w h T fq fq : fuel quantity w : aircraft weight h : altitude T : temperature pw : engine power cf : aircraft configuration cf pw (angle of attack, flaps, landing gear, pitch, roll, yaw) Tuninter 1153 Model 8

9 Weight Error Detection and Correction by Error Signatures 9

10 Dynamic Data-Driven Avionics Systems To facilitate development of smarter (flight) data streaming systems, we investigate: 1. Programming technology to facilitate modeling spatiotemporal data streaming applications PILOTS (ProgrammIng Language for spatio-temporal data Streaming applications) 2. Error detection using error signatures and error correction based on data redundancy 3. Machine learning techniques to infer relationships from data Offline supervised training Online prediction and learning 10

11 US Airways Flight 1549 Need for real-time decision support: US Airways flight 1549 Caused by birds strike damaging both engines The pilots successfully ditched in the Hudson river Image courtesy: cnn.com Image courtesy: wikipedia.com 11

12 Dynamic Data Driven Avionics Our Goal: To create a dynamic data driven trajectory generation system to assist pilots in such situations by taking into consideration dynamic aspects such as partial power, wind, etc. Our model uses geometric criteria for generating trajectories using different bank angles and corresponding radii of turn and glide ratios for different drag configurations. Our aircraft model is parameterized on a Baseline Glide Ratio for clean aircraft configuration assuming best gliding airspeed in straight flight. We dynamically infer the baseline glide ratio to update our model to meet the current performance of the aircraft. 12

13 Why DDDAS? Fig: Dynamic Data Driven Flight Trajectory Generation Fig: Damaged Aircraft Model 13

14 Best gliding angle of attack (airspeed) In order to maximize the gliding distance of the aircraft, the L/D ratio must be maximized by flying at the optimal speed. 14 Image courtesy: faa.gov

15 Effect of bank angle on glide ratio Fig: Steady-speed engines-out glide ratio vs. airspeed observed in the A (Avrenli and Dempsey, 2015) 15

16 Loss of thrust aircraft model Bank angle has a significant impact on the radius of turn and glide ratio, hence we use different bank angles to generate possible trajectories. 16

17 2D Dubins Paths 2D Dubins Path: Four options with a straight line segment: LSR, RSR, LSL, RSL The shortest one is chosen RSL LSL LSR RSR 17

18 Improvements over Existing Work Fig: (Atkins, 2009) trajectory vs simulator for t+12 sec to LGA 22 18

19 Trajectories varying bank angle Fig: Effect of bank angle on trajectories. 19

20 Low altitude trajectories For a given runway and a given bank angle, we have two scenarios: Low Altitude scenario: In this case, the trajectory consists of a simple Dubins Airplane Path that brings the aircraft to the runway. 20

21 High altitude trajectories High Altitude scenario: When a simple Dubins Airplane Path brings the aircraft too high above the runway. We find an intermediate point to lose excess altitude by extending the final approach along the runway heading. In this case, the trajectory consists of: A simple Dubins Airplane Path to bring the aircraft over the runway An integral [0,1,2.] number of spiral turns to lose excess altitude An Extended Runway of length X [0, 2πR g/g 0 ) if integral number of turns cannot bring the aircraft down to the runway at proper altitude. A landing configuration glide ratio and airspeed is assumed for the extended runway segment. When the number of spiral turns is 0, the high altitude scenario simplifies to a Middle Altitude Scenario. 21

22 High altitude trajectories High Altitude Scenario where number of spirals > 0 22

23 Mid-altitude trajectories Middle Altitude Scenario (High where number of spirals = 0) 23

24 Algorithm (Flowchart) 24

25 US Airways 1549 We then ran simulations attempting to recreate the circumstances of the US Airways 1549 incident as closely as possible. We used bank angles of 20, 30 and 45. The simulations were evaluated using data from the Flight Data Recorder on US Airways 1549 The simulated Airbus A320 was producing around a glide ratio of 19:1 in a clean configuration during straight flight. The simulator was used to test the viability and accuracy of the generated trajectories for the US Airways 1549 case. 25

26 US Airways 1549 FDR Data 26

27 US Airways 1549 (t+4) LGA 22 LGA 13 LGA 31 Fig: Computed trajectory for (t+4) scenario 27

28 US Airways 1549 (t+8) LGA 22 LGA 13 LGA 31 Fig: Computed trajectory for (t+8) scenario 28

29 US Airways 1549 (t+12) LGA 22 LGA 13 LGA 31 Fig: Computed trajectory for (t+12) scenario 29

30 US Airways 1549 (t+16) LGA 22 LGA 13 Fig: Computed trajectory for (t+16) scenario 30

31 US Airways 1549 (t+20) LGA 22 LGA 13 Fig: Computed trajectory for (t+20) scenario 31

32 US Airways 1549 (t+24) LGA 22 LGA 13 Fig: Computed trajectory for (t+24) scenario 32

33 US Airways 1549 (t+28) LGA 22 LGA 13 Fig: Computed trajectory for (t+28) scenario 33

34 US Airways 1549 (t+32) LGA 22 LGA 13 Fig: Computed trajectory for (t+32) scenario 34

35 US Airways 1549 (t+36) LGA 13 Fig: Computed trajectory for (t+36) scenario 35

36 US Airways 1549 Results Summary We were able to generate trajectories up to 36 seconds after birds strike, but not for 40 seconds or thereafter. Runway LGA 13 at La Guardia was reachable for all times up to t+36 seconds. Runway LGA 22 was reachable for all times up to t+32 seconds. A left initial turn is recommended at t+28 and t+32. Runway LGA 31 was reachable up to t+12 seconds. Runway LGA 4 was not reachable. 36

37 US Airways 1549 Flight Simulation Using a Precision Flight Controls CAT III Flight Simulator running X-Plane software with an Airbus A320. We were able to simulate a landing at Teterboro 24 (see video) even though trajectory generation fails. 37

38 Generated vs Sim-Flown LGA 22 (t+8) Fig: Computed trajectory for LGA 22 (t+8) vs trajectory flown in Flight Simulator. 38

39 Trajectory Safety Metrics We introduce safety metrics to evaluate trajectories. Evaluation of trajectories makes it possible for a pilot to make a better informed choice in a shorter time. The Metrics are: Average altitude Average distance from runway Total distance Number of turns Time in air Average Bank angle over height measures the occurrence of steep turns near the ground Extended final runway segment distance Each metric is normalized relative to the minimum or maximum (whichever is desired) and then evaluated in our safety metric equation 39

40 US Airways 1549 Trajectory Ranking Fig: Safety Metrics applied to (t+4),(t+8),(t+12). 40

41 Further Experimentation and Results We ran simulations loosely based on the US Airways 1549 incident. We used bank angles of 20, 30 and 45. The simulations were done for the same initial (latitude, longitude, heading) of the incident, but at different altitudes, for an Airbus A320 with a clean configuration straight flight glide ratio of 17:1, with no wind conditions. We evaluated our generated trajectories on a Precision Flight Controls CAT III Flight Simulator running X-Plane software with an Airbus A320 and no wind conditions. We evaluated trajectories using our safety metrics. It took a total of seconds for generating and evaluating trajectories with three bank angles for an accident height of feet to LGA31. 41

42 Low altitude trajectories Fig: Trajectories to LGA4 from an altitude of 4000 feet Fig: Trajectories to LGA31 from an altitude of 4000 feet 42

43 High altitude trajectories Fig: Trajectories to LGA4 from an altitude of feet Fig: Trajectories to LGA31 from an altitude of feet 43

44 Low/high altitude trajectory ranking 44

45 Trajectory Ranking Summary From 4000ft, LGA 31 at 45 degrees of bank angle is ranked first, due to its extended runway segment, and consistent with avoiding steep bank angles close to the ground. From 10000ft, lower bank angles are favored with LGA 31 at 20 degrees bank angle being chosen, closely followed by LGA 4 at 20 and 30 degrees of bank angle. This is despite being ranking low on average distance to runway. The average bank angle / height metric proves really useful in ranking safer trajectories first. 45

46 Why DDDAS? We can distil the aircraft model into purely geometric constraints: Glide Ratios and Radii Of Turn for different bank angles and drag configurations. We use a Baseline Glide Ratio g 0 (the glide ratio for clean configuration with best gliding airspeed in a straight flight) to predict glide ratios for different bank angles and drag configurations. Allows us to account for dynamic factors such as partial power, wind and effects of surface damage and compute a trajectory corresponding to the current performance capabilities. For example, USA Airways 1549 data shows some power still available in Engine 1, providing some thrust, resulting in baseline glide ratios of up to 25:1. DDDAS therefore enables more accurate trajectory generation for the emergency at hand. 46

47 Using sensor data to refine model Fig: Trajectories flown in Flight Simulator to evaluate their quality. Fig: Trajectories to LGA4 from an altitude of 4551 feet. New trajectories computed using data from flight simulator 47

48 Dynamic Data-Driven Avionics Systems Using a data-driven feedback loop, DDDAS-based avionics continuously analyze spatio-temporal data streams from airplane sensors, identify potential failure modes, and correct erroneous data. Result is new layer of logical redundancy in addition to existing physical redundancy for safer flight systems. New mathematical concepts: Error signatures: Mathematical function patterns with constraints on specific data stream errors/anomalies. Mode likelihood vectors: Stochastic selection of DDDAS system operation mode based on well-behaved sets of error signatures. New DDDAS software: PILOTS programming language Enables declarative (high-level) definition of DDDAS data streaming application models (input-output relationships between data streams), error signatures, and error correction functions. PILOTS software detects specific (e.g., failure-induced) data errors based on signatures and corrects data before processing according to the application model. We have confirmed effectiveness of our approach using data from commercial flight accidents Air France AF447 accident in June 2009: Airspeed sensor failure of the AF447 flight successfully detected and corrected after 5 seconds from beginning of the failure. Overall error mode detection accuracy reaches 96.31%. Tuninter 1153 accident in August 2005: The underweight condition due to the installation of an incorrect fuel sensor successfully detected with 100% accuracy during the cruise phase of flight. US Airways 1549 incident in January 2009: Bird strike after takeoff requiring real-time trajectory generation assistance. We can successfully generate trajectories up to 36 seconds after birds strike. Sequential computation time: 120 ms per trajectory. 48

49 Multi-Aircraft Collaborative Flight Assistant System Previous flights data Updated weather Information from other planes Expert-Level Flight Assistant Aircraft sensors (Mathematical function patterns used to identify failure modes) Offline Training Corrected inputs PILOTS * System Learning Engine Prediction request Prediction response Avionics Application Measured error Failure Detection & Data Correction Cloud 49 External real-time data inputs Corrected outputs Identified failure Stochastic & Logic-based Flight Assistant (to be developed) *: ProgrammIng Language for spatio-temporal data Streaming applications 3D terrain data Left engine is damaged Failure & Recommended actions We should land at airport X immediately! Airplane pilots

50 Research Challenges (1/4) A quantitative spatial and temporal logic as a formalism: To enable reasoning about data streams that associate values to specific points or intervals of space and time. Reasoning about differential equations (location, velocity, acceleration) and derived data streams. To enable geometric reasoning capabilities, in particular, trigonometric formulae to calculate with aircraft speeds, headings, range, and endurance. v α a w,x r Speed (horizontal) Direction Aircraft Wind, crosswind Runway Ground speed and crosswind as functions of airspeed, wind, and runway heading 50 8/26/2017

51 Research Challenges (2/4) Extensions to logic programming to support stochastic reasoning. Language extensions to standard Horn clause-based knowledge bases to incorporate probabilities. Special language support for spatial and temporal data streams. Incremental reasoning algorithms to dynamically re-compute logical queries efficiently as new data gets injected into the application. If New pilot report: icing en route New winds aloft New surface winds at destination Imminent engine failure 51 then New route New altitude New airport Nearest airport Dynamic Data-Driven Flight Plan Adaptation Examples 8/26/2017

52 Research Challenges (3/4) Data streaming analytics in real-time using cloud computing More data are expected to be available through the Internet and in-flight through Next Generation Transportation system (ADS-B by 2020). Reason about spatial and temporal data in real-time Give pilots better information to make more accurate judgments during crucial emergency moments Offline and online components Analyzing key historical data and relatively static data (e.g., terrain, aircraft models) offline Combining it with dynamic data (e.g., failure conditions, weather) for real-time decision making 52 8/26/2017

53 Research Challenges (4/4) Domain-specific programming languages are needed for data scientists Easier data analyses, information generation, decision support. Separation of concerns Enables compiler (static) and middleware (dynamic) optimizations 53 8/26/2017

54 Questions? Download open-source PILOTS at: Consider textbook: Partial support from: Air Force Office of Scientific Research DDDAS Program Dr. Frederica Darema, Dr. Erik Blasch (AFOSR Grant No. FA , FA ), National Science Foundation EAGER/Dynamic Data Program (NSF Grant No. ECCS ). MIT Press, June