Airflow_Hazard_Vis by IjzLIf9j

VIEWS: 0 PAGES: 75

									 Improving Aviation Safety with
    Information Visualization:
Airflow Hazard Display for Pilots

         Cecilia R. Aragon
              IEOR 170
             UC Berkeley
             Spring 2006
              Acknowledgments

• This work was funded by the NASA Ames Full-
  Time Graduate Study Program (Ph.D. in
  Computer Science at UC Berkeley)
• Thanks to my advisor at UC Berkeley, Professor
  Marti Hearst, and Navy flight test engineer
  Kurtis Long
• Thanks to Advanced Rotorcraft Technology, Inc.
  for the use of their high-fidelity flight simulator
Spring 2006             IEOR 170                    2
                  Talk Outline

• Introduction
• Related Work
• Preliminary Usability Study
• Flight Simulation Usability Study
• Conclusions and Further Work



Spring 2006             IEOR 170      3
              Introduction




Spring 2006       IEOR 170   4
                             Motivation

• Invisible airflow hazards cause aircraft accidents
     –   Wind shear
     –   Microbursts
     –   Vortices (turbulence)
     –   Downdrafts
     –   Hot exhaust plumes


• Crash of Delta Flight 191 at DFW 1985 (microburst)
• NTSB database 1989-99
     – 21,380 aircraft accidents
     – 2,098 turbulence/wind related


Spring 2006                      IEOR 170              5
Spring 2006   IEOR 170   6
Spring 2006   IEOR 170   7
                      The Problem

• Invisible airflow hazards cause aircraft accidents
     – Air is invisible, so pilots can’t see hazards
     – If air flows past obstacles, flow will become more
       turbulent
• Helicopters are especially vulnerable
     – Rotorcraft aerodynamics
     – Must operate in confined spaces
     – Operationally stressful conditions (EMS, military
       operations, shipboard operations)


Spring 2006                  IEOR 170                       8
                       A Possible Solution

• If pilots could see hazards, could take appropriate action
• New lidar technology suggests a solution
     – Lidar (light detection and ranging) is essentially laser radar. A laser transmits
       light which is scattered by aerosols or air molecules and then collected by a
       sensor. Doppler lidar can detect the position and velocity of air particles.




• My research focuses on the human interface -- how to
  visualize the sensor data for pilots -- too much
  information could overload pilot during critical moments



Spring 2006                                IEOR 170                                        9
                 Research Approach

• User-centered (iterative) design process
• Simulated interface for head-up display (HUD)
  based on lidar sensors that scan area ahead of
  helicopter and acquire airflow velocity data
• Focused on helicopter-shipboard landings
• Importance of realism:
     – Used actual flight test data from shipboard testing,
       high-fidelity helicopter simulator, experienced military
       and civilian helicopter pilots


Spring 2006                   IEOR 170                        10
  Rationale for using Shipboard Landings

• Why focus on helicopter shipboard landings?
     – Problem is real: dangerous environment, want to improve
       safety
     – Ship superstructures always produce airwake

• Large quantities of flight test data due to
  demanding environment




Spring 2006                     IEOR 170                         11
              Related Work




Spring 2006       IEOR 170   12
                Related Work

• Flow visualization
• Aviation displays
• Navy “Dynamic Interface” flight tests




Spring 2006             IEOR 170          13
                     Flow visualization
• Detailed flow visualizations designed for scientists or engineers
  to analyze at length
• Much work has been done in this area [Laramee et al 04]
    – Streamlines, contour lines (instantaneous flow) [Buning 89], [Strid et
      al 89], [Helman, Hesselink 91]
    – Spot noise [van Wijk 93], line integral convolution [Cabral, Leedom
      93], flow volumes [Max, Becker, Crawfis 93], streaklines, timelines
      [Lane 96], moving textures [Max, Becker 95] (unsteady flow)
    – Automated detection of swirling flow [Haimes, Kenwright 95]
    – Terrain and turbulence visualization [LeClerc et al 02]
• But usually no user tests [Laidlaw et al 01], and
  not real-time



 Spring 2006                       IEOR 170                             14
                 Aviation displays

• Synthetic and enhanced vision and
  augmented-reality displays [Hughes et
  al 02], [Parrish 03], [Spitzer et al 01],
  [Kramer 99], [Wickens 97]
• Weather visualization, microburst
  detection [NASA AWIN, TPAWS],
  [Latorella 01], [Spirkovska 00],
  turbulence detection/prediction [Britt et
  al 02], [Kaplan 02]
• Wake vortex visualization [Holforty 03]



Spring 2006                  IEOR 170         15
     Navy Ship-Rotorcraft Compatibility
     Flight Testing (“Dynamic Interface”)
• Very hazardous environment [Wilkinson et al 98]
• Significant amounts of flight testing [Williams et
  al 99]
• Recognized need for pilot testing
• Goal: improve safety




Spring 2006             IEOR 170                   16
               Current state of the art
• Ship/helicopter flight
  tests, wind tunnel tests,
  CFD
• Develop operational
  envelopes
     – Limit allowable landing
       conditions significantly
     – Envelopes are conservative
       for safety reasons
• Pilots use intuition, but
  accidents still occur

Spring 2006                     IEOR 170   17
              Preliminary Usability Study




Spring 2006               IEOR 170          18
        Preliminary usability study: goals

• Assess efficacy of presenting airflow data
  in flight
• Obtain expert feedback on presentation of
  sample hazard indicators to refine design
  choices




Spring 2006            IEOR 170              19
   Usability study: low-fidelity prototype

• Rhino3D (3D CAD modeling program)
     – Easy access to ship models, ease of rapid prototyping
     – Chosen over 2D paper prototype, MS Flight
       Simulator, WildTangent, VRML-based tools, Java and
       Flash
• Series of animations simulating helicopter’s final
  approach to landing
• Different types of hazard indicators
• Get pilot feedback and suggestions (interactive
  prototyping)
Spring 2006                 IEOR 170                      20
      Low-fi usability study screen shots




Spring 2006           IEOR 170              21
      Low-fi usability study screen shots




Spring 2006           IEOR 170              22
       Low-fi usability study participants

•     Navy helicopter test pilot, 2000 hours of flight
      time, 17 years experience
•     Navy helicopter flight test engineer, 2000+
      hours of simulator time, 100 hours of flight time,
      17 years experience
•     Civilian helicopter flight instructor, 1740 hours
      of flight time, 3 years experience




Spring 2006                IEOR 170                   23
              Low-fi usability study results

• All participants said they would use system
• Feedback on hazard indicators:
     – Color: all preferred red/yellow only
     – Transparency: should be visible enough to get attention,
       but must be able to see visual cues behind it
     – Depth cueing: all preferred shadows below object, #1 said
       shadows alone sufficient. #2 wanted connecting line. No
       one wanted tick marks or numeric info.
     – Texture: #1, #2 didn’t want. #3 suggested striping
     – Shape: Rectilinear and cloud shapes favored. Keep it
       simple! Watch for conflicting HUD symbology.

Spring 2006                    IEOR 170                        24
    Low-fi usability study results (cont’d)

• Motion is distracting!
     1: absolutely no motion
     2: didn’t like motion
     3: slow rotation on surface of cloud OK, nothing fast




Spring 2006                  IEOR 170                        25
       Low-fi usability study conclusions

• They want it!
• Keep it simple
     – Color: red & yellow only (red = danger, yellow = caution)
     – Less complex shapes preferred
• Use accepted symbology/metaphors
     – Watch for conflicting HUD symbology
• Decision support system, not scientific visualization
  system
     – Show effects rather than causes
     – Don’t want distraction during high-workload task
     – Preference for static rather than dynamic indicators


Spring 2006                       IEOR 170                         26
Flight Simulation Usability Study
         Flight Simulation Usability Study
• Implement visual hazard display
  system in simulator based on
  results from low-fidelity prototype
• Advanced Rotorcraft
  Technology, Inc. in Mountain
  View, CA, USA
   – High-fidelity helicopter flight simulator
   – Accurate aerodynamic models
• Use existing ship and helicopter
  models, flight test data
• Simulated hazardous conditions,
  create scenarios, validated by
  Navy pilots and flight engineers

 Spring 2006                            IEOR 170   28
       Flight Simulation Usability Study:
                 Participants
• 16 helicopter pilots
     – from all 5 branches of the military (Army, Navy, Air Force,
       Coast Guard, Marines)
     – civilian test pilots (NASA)
     – wide range of experience
          • 200 to 7,300 helicopter flight hours (median 2,250 hours)
          • 2 to 46 years of experience (median 13 years)
          • age 25 to 65 (median age 36)
• No previous experience with airflow hazard visualization

Spring 2006                        IEOR 170                             29
              Simulation Experiment Design
• 4 x 4 x 2 within-subjects design (each pilot flew
  the same approaches)
• 4 shipboard approach
  scenarios

• 4 landing difficulty levels
  (US Navy Pilot Rating Scale - PRS 1-4)
• Each scenario was flown at all difficulty levels
  both with and without hazard indicators
• Orders of flight were varied to control for
  learning effects
Spring 2006               IEOR 170                    30
  Airflow Hazard Indicators in Simulator




Spring 2006        IEOR 170                31
              Simulation Experiment Design
Landing            Description               Purpose                 Hazard
difficulty                                                           indicator
LD 1               No problems;              Control                 None
                   minimal pilot effort
                   required

LD 2               Moderate pilot effort     Test negative effects   Yellow/None
                   required; most pilots     of hazard indicator
                   able to land safely

LD 3               Maximum pilot effort      Test benefit of         Yellow/None
                   required; repeated        hazard indicator
                   safe landings may
                   not be possible

LD 4               Controllability in        Test benefit of         Red/None
                   question; safe            hazard indicator
                   landings not              combined with pilot
                   probable                  SOP

Spring 2006                               IEOR 170                               32
Spring 2006   IEOR 170   33
                  Dependent Variables

• Objective data: sampled at 10 Hz from simulator
     –   aircraft velocity and position in x, y, z
     –   lateral and longitudinal cyclic position and velocity
     –   collective and pedal positions and velocities
     –   landing gear forces and velocities
     –   (A “crash” was defined as an impact with the ship
         deck with a vertical velocity of more than 12 fps)
• Subjective data: 21-probe Likert-scale
  questionnaire administered to pilots after flight


Spring 2006                     IEOR 170                         34
                 Hypotheses
1. Crash rate will be reduced by the presence of
  hazard indicator (LD 3).
2. Crashes will be eliminated by red hazard
  indicator if a standard operating procedure
  (SOP) is given to the pilots (LD 4).
3. Hazard indicator will not cause distraction or
  degradation in performance in situations where
  adequate performance is expected without
  indicator (LD 2).
4. Pilots will say they would use airflow hazard
  visualization system
Spring 2006            IEOR 170                     35
              Hypothesis 1 confirmed
• Presence of the hazard indicator reduces the
  frequency of crashes during simulated shipboard
  helicopter landings (t-test for paired samples,
  t=2.39, df=63, p=0.00985). 19% --> 6.3%
                                           Landing Difficulty 3:
                                     Crash Rate vs. Presence of Hazard
                                                 Indicator
                              0.25


                              0.20
                 Crash Rate




                              0.15


                              0.10


                              0.05


                              0.00
                                           Absent                      Present
                                                    Hazard Indicator



Spring 2006                                   IEOR 170                           36
              Hypothesis 2 confirmed
• Presence of the red hazard indicator combined
  with appropriate instructions to the pilot prevents
  crashes (t=4.39, df=63, p < 0.000022). 23%-->0%
                                               Landing Difficulty 4:
                                    Crash Rate vs. Presence of Hazard Indicator

                             0.35
                             0.30
                             0.25
                Crash Rate




                             0.20
                             0.15
                             0.10
                             0.05
                             0.00
                                             Absent                 Present
                                                    Hazard Indicator



Spring 2006                                        IEOR 170                       37
                                         Hypothesis 3

• No negative effect of hazard indicator. 8%-->8%

                                             Landing Difficulty 2:
                                  Crash Rate vs. Presence of Hazard Indicator

                           0.12

                           0.10
              Crash Rate




                           0.08

                           0.06

                           0.04

                           0.02

                           0.00
                                         Absent                      Present
                                                  Hazard Indicator



Spring 2006                                             IEOR 170                38
                                           Hypothesis 3 (cont’d)
• Pilots believe hazard indicators were not
  distracting (Probe 6 results).
                                         6. The airflow hazard visualization distracted me
                                                from the task of flying the aircraft.
                                    12

                                    10
              Number of responses




                                    8

                                    6

                                    4

                                    2

                                    0
                                          Strongly   Disagree    Neither Agree   Agree   Strongly
                                          Disagree               Nor Disagree             Agree
                                                                Pilot response

                                                           6% Agree, 94% Disagree
                                                            M edian 2, Std Dev 0.7


Spring 2006                                                        IEOR 170                         39
              Hypothesis 4 confirmed
• Pilots would use the system (Probe 21 results).
                                        21. I would use this display system if it were
                                                  available on my aircraft.
                                    9
                                    8

                                    7
              Number of responses




                                    6

                                    5
                                    4

                                    3
                                    2

                                    1
                                    0
                                         Strongly   Disagree    Neither Agree   Agree   Strongly
                                         Disagree               Nor Disagree             Agree
                                                               Pilot response

                                                        81% Agree, 13% Disagree
                                                         Median 4.5, Std Dev 1.0


Spring 2006                                                    IEOR 170                            40
                                                                  Pilot workload:
   Power spectrum analysis of control inputs

                                                 Power spectrum of lateral cyclic position (pilot 5, landing spot 9)


                                      40

                                      35
         Power spectrum coefficient




                                      30
                                                                                                                   Landing
                                      25                                                                           difficulty 1
                                      20
                                                                                                                   Landing diff 3
                                      15                                                                           (w/o indicator)

                                      10                                                                           Landing diff 3
                                                                                                                   (w/indicator)
                                       5

                                       0
                                           0.5                1              1.5               2
                                                                   Frequency (Hz)



Spring 2006                                                                     IEOR 170                                             41
         Go-Arounds (Aborted Landings)

• Does the presence of the hazard indicator
  increase the go-around rate?
              Landing       Hazard       Go-         Total     Go-Around   Standard
              Difficulty   Indicator   Arounds    Approaches     Rate        Error
               LD 1         No            3           64        0.0469     0.0266
                            No           17           64        0.266      0.0556
               LD 2
                           Yellow        12           64        0.188      0.0492
                            No           22           64        0.344      0.0598
               LD 3
                           Yellow        23           64        0.359      0.0605



• No significant differences found.


Spring 2006                                      IEOR 170                             42
      Analysis by Pilot Experience Level

• Does pilot experience level have any effect on
  the benefits produced by the hazard indicators?
• To find out, divide pilots into three groups:

        Pilot Experience    Helicopter Flight   Number of Pilots in
              Level              Hours              Group
        Less experienced        200 – 850              5
              Moderately
                              1500 – 3200               7
              experienced

       Highly experienced     4000 - 7300               4



Spring 2006                        IEOR 170                           43
  Analysis by Pilot Experience Level (cont’d)
• Same general trends -- but small sample size
• No significant difference between the groups

                   Crash Rate vs. Experience Level

                    0.25
                     0.2
                                                    LD 2/No
              Crash 0.15                            LD 2/Haz
              Rate 0.1                              LD 3/No
                    0.05                            LD 3/Haz
                                                    LD 4/No
                      0
                           Low     Mod        Hi    LD 4/Haz
                           Pilot Experience Level

Spring 2006                        IEOR 170                    44
              Analysis of Subjective Data
• 94% found hazard indicators helpful
                                         18. The presence of the hazard indicators gave me
                                          more confidence as to the state of the winds and
                                                        airwake on deck.

                                     9
               Number of responses




                                     8
                                     7
                                     6
                                     5
                                     4
                                     3
                                     2
                                     1
                                     0
                                          Strongly    Disagree     Neither Agree   Agree   Strongly Agree
                                          Disagree                 Nor Disagree
                                                                  Pilot response

                                                           94% Agree, 6% Disagree
                                                            Median 4, Std Dev 1.0


Spring 2006                                                      IEOR 170                                   45
    Analysis of Subjective Data (cont’d)
• Is motion (animation) helpful or distracting?
                                        14. It would be distracting if the hazard indicator
                                                     showed airflow motion.
                                    9
                                    8
              Number of responses




                                    7
                                    6
                                    5
                                    4
                                    3
                                    2
                                    1
                                    0
                                          Strongly   Disagree     Neither Agree   Agree   Strongly Agree
                                          Disagree                Nor Disagree
                                                                Pilot response

                                                       31% Agree, 63% Disagree
                                                         Median 2, Std Dev 1.1

Spring 2006                                                     IEOR 170                                   46
          Conclusions and Further Work




Spring 2006           IEOR 170           47
                 Conclusions
• Flight-deck visualization of airflow hazards yields
  a significant improvement in pilot ability to land
  safely under turbulent conditions in simulator
• Type of visualization to improve operational
  safety much simpler than that required for
  analysis
• Success of user-centered design
  procedure

Spring 2006             IEOR 170                   48
              Further Work

• Additional data analysis
• Further studies
• Steps toward system deployment
• Extensions to other areas




Spring 2006          IEOR 170      49
              Additional data analysis

• Power spectrum analysis of control input
  data
• Flight path deviations and landing
  dispersion
• Quantitative measures of landing quality



Spring 2006             IEOR 170             50
                   Further studies
• Quantitatively compare hazard indicators
  with other types
     – light/buzzer in cockpit
     – animated indicator
     – numeric information such as airflow velocity
• Adaptive displays
     – more detailed at beginning of approach, simpler at
       end
     – how adapt to pilot state? physiological sensors vs.
       pilot-selectable modes

Spring 2006                  IEOR 170                        51
   Steps toward system deployment

• Collaboration with lidar developers,
  integration with real-time data
• Integration with synthetic vision displays
• Augmented reality image registration




Spring 2006          IEOR 170                  52
              Extensions to other areas

• Other aviation domains
     –   aerial firefighting
     –   search and rescue
     –   offshore oil platforms
     –   unmanned aerial vehicles (UAVs)
     –   fixed-wing operations
• Space exploration
• Emergency response
• Automobiles or other motor vehicles

Spring 2006                  IEOR 170      53
              Extra Slides




Spring 2006       IEOR 170   54
Crash Statistics for All Landing Difficulties
  Landing       Hazard                     Total     Crash    Standard
                           Crashes
  Difficulty   Indicator                Approaches   Rate       Error
    LD 1          No         6                  64   0.0938   0.0367
                  No         5                  64   0.0781   0.0338
    LD 2
                Yellow       5                  64   0.0781   0.0338
                  No         12                 64   0.188    0.0492
    LD 3
                Yellow       4                  64   0.0625   0.0305
                  No         15                 64   0.234    0.0534
    LD 4
                 Red         0                  64   0        0



Spring 2006                          IEOR 170                          55
                              Control group (LD 1)
• No significant difference between crash rate at
  LD 1 (control) and LD 2 with hazard indicator
  and LD 3 with hazard indicator. 9% - 8% - 6%
                             Comparison of Crash Rates at Landing Difficulty 1 with
                               Rates at LD 2 and 3 with Hazard Indicator Present

                           0.14

                           0.12

                            0.1
              Crash Rate




                           0.08

                           0.06

                           0.04

                           0.02

                             0
                                  LD 1/No Haz Indicator   LD 2/Haz Indicator Present LD 3/Haz Indicator Present
                                     Landing Difficulty (LD) and Presence/Absence of Indicator


Spring 2006                                                 IEOR 170                                              56
                                                          Learning Effects?

• First half: 25 crashes/224; second half: 22/224.
• Not a significant difference --> no apparent bias.
                                                  Number of crashes as a function of approach order

                                              5
              Number of crashes (out of 16)




                                              4


                                              3


                                              2


                                              1


                                              0
                                                  1   2   3   4   5   6   7   8   9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

                                                                                   Approach - order flown

Spring 2006                                                                             IEOR 170                                               57
 Airflow Hazard Indicator (Aft Scenario)




Spring 2006        IEOR 170            58
Airflow Hazard Indicator (Bow Scenario)




Spring 2006      IEOR 170            59
                      Pilot Demographics
                                                                     Number of
                                Helicopter               Years of
     Pilot       Employer                    Age                     Shipboard
                                  Hours                 Experience
                                                                     Landings
       1       Coast Guard            800          30            3           40
       2       Coast Guard           1500          28          5.5          60
       3       Coast Guard            770          26          2.5         200
       4       Coast Guard            420          26            2          30
       5       Coast Guard            200          25            2          75
       6       Coast Guard           5600          43           22        1000
       7           NASA              3100          59           46         100
               Air Force/Air
       8                             3000          37           18          18
              National Guard
               Air Force/Air
       9                             1800          34            8           0
              National Guard
      10           NASA              2500          65           35         302
      11      Army, civilian         4300          56           34           6
               Air Force/Air
      12                             2000          33            7           0
              National Guard
      13       Army, NASA            7300          51           29         150
      14      Air Force, NASA        4000          60           36           0
      15      Navy, Marines          3200          41           18        1500
      16         Marines              850          33            8         600



Spring 2006                             IEOR 170                                  60
Spring 2006   IEOR 170   61
Spring 2006   IEOR 170   62
Spring 2006   IEOR 170   63
Spring 2006   IEOR 170   64
Spring 2006   IEOR 170   65
Spring 2006   IEOR 170   66
Spring 2006   IEOR 170   67
Spring 2006   IEOR 170   68
Spring 2006   IEOR 170   69
Spring 2006   IEOR 170   70
Spring 2006   IEOR 170   71
Spring 2006   IEOR 170   72
Spring 2006   IEOR 170   73
     Low-fi usability study: methodology
• 1 ½-hour interview in front of projection screen,
  videotaped
• Two experimenters, one operates computer, one asks
  questions
• Display series of hazard indicators in Rhino3D
• Variables:
     –   Shape
     –   Color
     –   Transparency
     –   Texture
     –   Depth cueing
     –   Motion
• Ask specific and open-ended questions throughout the
  interview

Spring 2006               IEOR 170                       74
  “The Holy Grail” – Quote from Pilot #1
• “The holy grail…”
     – increase safety and
     – increase operational capability
• Usually you either have:
     – increased safety but have operational
       restrictions…or
     – greater operational capability but have risks
       associated with employing that additional
       capability...
• “In this case you actually have a concept
  that could potentially give you both.”
Spring 2006               IEOR 170                     75

								
To top