final technical report (according to nasa handbook nhb ... · syed firasat ali principal...
TRANSCRIPT
Final Technical Report
(According to NASA Handbook NHB 5800.IC)
A LOW COST SIMULATION SYSTEM
TO DEMONSTRATE _5-] J
PILOT INDUCED OSCILLATION PHENOMENON
Syed Firasat Ali
Principal Investigator
Aerospace Science Engineering Department
Tuskegee University, Tuskegee, AL 36088
NASA Dryden Flight Research Center
Research Grant No. NAG 2-4006
April 1, 1994 to January 15, 1997
https://ntrs.nasa.gov/search.jsp?R=19970026582 2019-06-13T21:55:00+00:00Z
ABSTRACT
A flight simulation system with graphics and software on Silicon Graphics computer
workstations has been installed in the Flight Vehicle Design Laboratory at Tuskegee
University. The system has F-15E flight simulation software from NASA Dryden which uses
the graphics of SGI flight simulation demos. On the system, thus installed, a study of pilot
induced oscillations is planned for future work. Preliminary research is conducted by
obtaining two sets of straight level flights with pilot in the loop. In one set of flights no
additional delay is used between the stick input and the appearance of airplane response on
the computer monitor. In another set of flights, a 500 ms additional delay is used. The
flight data is analyzed to find cross correlations between deflections of control surfaces and
response of the airplane. The pilot dynamics features depicted from cross correlations of
straight level flights are discussed in this report. The correlations presented here will serve as
reference material for the corresponding correlations in a future study of pitch attitude
tracking tasks involving pilot induced oscillations.
ACKNOWLEDGMENTS
The research project covered by this report was sponsored by the NASA Dryden Flight Research Center under grant number
NAG2-4006. Mr. Larry Shilling's support and encouragement and Mr. Ken Norlin's expertise on simulation software were
vital to the completion of this work.
Dr. Amnon Katz from the University of Alabama and Dr. Vascar G. Harris from Tuskegee University, who are both pilots
and aerospace engineering professors, contributed greatly by providing their invaluable support and insights on such
issues as experimental design, statistical data processing, and most importantly, relating the data to a pilot's performance.
Jason Williams, an aerospace engineering student at Tuskegee University, enthusiastically participated in all aspects of
the research project including alterations in the software, system trouble shooting, and developing C++ codes to aid in
calculating and graphing of the cross correlations. Together with Yusef Johnson, a fellow aerospace engineering student,
Jason coordinated the conduct of the test flights performed by the volunteer pilots and even participated as one of the
simulator pilots in the study. It should also be noted that Yusef demonstrated keen interest in obtaining and understanding
the cross correlations.
Despite their professional commitments, Dr. Larry Koons, Mr. Nathaniel Glover, and Glen Morean devoted their valuable
time as volunteer pilots to conduct all of the required test flights.
Special appreciation is expressed to Dr. C. L. Chert, Abraham George, and Ernest Kashiri who are faculty and staff members
of Tuskegee University for providing substantial advice on software related issues.
Ronald Jones and Dennis Ezell, Tuskegee University aerospace engineering students, deserve special recognition for their
efforts in helping to initiate flight simulation research at Tuskegee University. It was during their tenure as Co-op students
at the NASA Dryden Flight Research Center that they developed this desire and their continued interest in that regard was
well received by Mr. Chuck Brown and Mr. Larry Schilling of NASA Dryden as well as by Dr. Shaik Jeelani of Tuskegee
University. In part, their desire has materialized into the research project covered by this report. Dennis has continued to
show exemplary devotion to the flight simulation program and has given noteworthy help in editing this report.
I extend my sincere gratitude to all those whose guidance, support, help, and good wishes have resulted in the completion
of a phase of research work that is reported here.
Syed Firasat Ali,
Principal Investigator
ii
TABLE OF CONTENTS
Abstract
Acknowledgments
List of Figures and Tables
ii
iv
Introduction
The Simulation System
Flight Data Processing
The Pilots and the Test Flights
Flight Test Results
Conclusions 17
Nomenclature
References
Appendix A: The Graphs of the Raw Data for the Five Selected Flights.
Appendix B: Copy of the paper; "Pilot Input and Airplane Response Cross Correlations
On a Flight Simulator With and Without Transport Delay."
18
19
A1
B1
iii
LIST OF FIGURES AND TABLES
Figures:
-- Figure 1 : Correlation Between Elevator Deflection and Time Derivative of Altitude.
Figure 2: Correlation Between Elevator Deflection and Airplane Altitude. 10
Figure 3: Correlation Between Elevator Deflection and Airplane Heading.
Figure 4: Correlation Between Aileron Deflection and Airplane Heading.
Figure 5: Correlation Between Rudder Deflection and Airplane Heading.
Figure 6: Correlation Between Deflections of Elevator and Ailerons.
11
12
13
14
-- Figure 7: Correlation Between Deflections of Elevator and Rudder. 15
Figure 8: Correlation Between Deflections of Rudder and Ailerons. 16
Tables:
-- Table I: Comparison of Pilots' Performance in Their Best Flights. 8
iv
INTRODUCTION
McRuer and Graham (1981) have presented a survey of eighty years of flight control, and Ashkenaz
(1984) has presented a survey of twenty five years of handling qualities research. From both survey
presentations, it appears that prior to 1984, the cross correlations between pilot inputs and airplane
response on a real time simulator had not been studied, but transfer functions representing human
pilots had been proposed (see, for example, Washizu and Miyajima, 1965, and Smith, 1965), and the
pilot induced oscillation (PIO) theory and resulting prediction technique had been developed (see
Smith, 1977). The prediction techniques were tested on eighteen aircraft/flight control system landing
configurations by Bjorkman (1986). Bjorkman indicated the need of more data and simulator studies
to gain physical insights into PIO mechanization. From more recent works on PIO including the
paper by Hess and Kalteis (1990), the need of more data is further desired. Cardullo et al, and
Middendorf et al are some of the recent studies on the effect of simulator transport delay on pilot's
performance. In the present work a low cost flight simulator system with graphics and software on
computer workstations has been installed. On the simulation system, a future study of PIO is
proposed. As prepatory work for a future PIO study, the flight data has been obtained for two sets of
flights with pilot in the loop. For conducting their flights, the pilots utilized the heads-up display on
the computer monitor and operated a joystick to maintain a given altitude and a given heading. One
set of flights is obtained without any additional delay between the stick input and the appearance of
airplane response on the computer monitor. In another set of flights, a 500 ms additional delay is
used. The flight data has been processed to find cross correlations between deflections of control
surfaces and response of the airplane. It is anticipated that these cross correlations for straight level
flights would serve as reference material to compare the respective cross correlations for future
simulator studies of PIO. A portion of the work of this report was presented at an AIAA Southeastern
Student Conference By Williams and Johnson (1997), and their paper is included as Appendix B in
this report. While the test flights for the present study were being conducted, the raw data on some of
them were displayed on the internet at http://silicon.tusk.edu/-jpwill.
THE SIMULATION SYSTEM
The simulation system is comprised of two SGI workstations and a BG System Joystick called a
FlyBox. The two workstations are SGI IRIS with 2 CPUs and SGI Indigo 2 with one CPU; they are
connected through ethernet. The simulation software, which represents the dynamics of flight of an
F-15E jet fighter has been provided by NASA Dryden Flight Research Center. For the graphics of the
system, the graphics of the SGI flight simulation demos has been integrated with the software of the
airplane dynamics. Norlin (1995) has provided a description of the design of software, and its
simulationcapabilities.According to Norlin, the F-15E real-timesimulationis coded with aFORTRAN77 shell and C supportroutinesand it operateson a UNIX-basedmultiprocessorcomputer.TheFORTRANcodeincludestheaircraftmodels,equationsof motion,integrationtablelook-ups,initializationand displaygenerationroutines.The C supportroutinesare usedfor thegraphicaluserinterface,memorymapping,priority boosting,andinterrupthandlers.In addition,C isusedfor thegraphicsanddistributedsystemfunctions.Theintegrationschemeis optimizedfor real-timeoperation.In thismethod,thederivativesareonly calculatedoncefor eachframe,a weightedaverageof themidpointandpreviousframederivativeisusedto predictthederivativeat theendoftheframe.Theequationsof motionassumeflat earthandsix degreesof freedom.The airplaneisassumedsymmetricalaboutits ownx-z plane.The simulationmodelcalculationsandintegrationoftheequationsareperformedin thereal-timeloop. Forpilot in the loop or hardwarein the loop,themainreal-timetaskis interruptdrivenat thehighestallowablesystempriority. The useof 3 CPUsensuresthattimingconstraintsareadequatelymet.For their simulators,NASA Drydenpreferto usethe name engineeringsimulatorsinsteadof production training simulators.The engineeringsimulatorsprovidehighfidelity simulationthatis responsiveto theneedsof theresearchers.TheF-15Esimulatorusedin thepresentwork is anengineeringsimulator.It is calledthe F-15 ACTIVEwhichimpliesthatit incorporatesAdvancedControlTechnologyfor IntegratedVehicles.Thecodehasthe flexibility to programa desiredtransportdelaybetweenthe stick input and the airplaneresponseon themonitor. Theuseof UNIX "makefiles" to compilethe simulationallowsminorupdateswithin about 5 minutes.Thejoystickhasthreedifferentmovementsto deflect the threecontrolsurfaces:forward-rearwardmovementfor elevatordeflection,right-leftmovementfor aileron,andclockwise-counterclockwiserotationaboutits ownaxisto simulatepedalmovementfor rudder
deflection.It maybenotedthattheexperiencedpilots wouldprefer theuseof rudderpedalsinstead
of theexistingstickmovementsto simulatethepedalmovements,therefore,the absenceof pedalsmay causea lack of coordinationbetweenthe rudder and aileronsin their test flights duringsimulation.In a simulatedflight, the digital dataon forty two different flight parameterscanberecordedat every25mstimeintervals.Theflightparametersincludealtitude(H), timerateof changeof altitude(HDOT),andheading(PSI)aswellasstickdeflectionfor elevator(DEP),stickdeflection
for aileron(DAP),andpedalmovementfor rudder(DRP).A pull on thestickor a positivevalueofDEPdeflectstheelevatorupwardwhichraisesthenoseof theairplaneandincreasesits altitude(see,for example,Roskam1995,p. 241).In thepresentreporttheDEP,DAP,andDRPdataareessentiallyusedto find crosscorrelations.It isunderstoodthatthedeflectionsof elevator,aileron,andrudderare
proportionalto the valuesof DEP,DAP, and DRP,respectively.Therefore,for computing thedimensionlesscrosscorrelationcoefficients,thevaluesof DEP,DAP,andDRParetreatedasrelativevaluesof deflectionsof elevator,ailerons,andrudderalthoughthedimensionalvaluesof DEP,DAP,andDRParenot thedimensionalvaluesof deflectionsof therespectivecontrolsurfaces.
FLIGHT DATA PROCESSING
In the present study, all flights conducted on the simulator were straight level flights of three minute
duration. For any flight conducted, the simulator can provide digital data for 42 flight parameters at
every 25 ms time interval. For the present work, however, the digital data was obtained for only
altitude (ft.), heading (deg.), time derivative of altitude (ft./sec.), and the deflections of elevator,
rudder, and aileron. For the altitude, H represents its instantaneous value and h represents the
fluctuating part of its instantaneous value. Therefore,
h(t) = H(t) - avg. H
where the avg. H is the average altitude of the three minute flight and t represents time.
For the heading, ud represents its instantaneous value and _ represents the fluctuating part of its
instantaneous value. Therefore,
_(t) = W(t) - avg.q _
Similar relations are used between instantaneous values, fluctuating values and average values of the
control surface deflections. A is used for instantaneous value of control surface deflection, and 8 for
the fluctuating part of the instantaneous value. For distinction between the three control surface
deflections, the subscripts a and r represent aileron and rudder, respectively. No subscript is used on A
or 8 to indicate elevator. The rms values of the above signals are denoted by, 8 ', h', and x_'. The
coefficient of cross correlations between 8 and h for an arbitrary time gap x is denoted as CORR (x)
and it is defined as:
t=T
1jCORR(x) - T 8' h' 8(0 h(t + '_) dtt=0
The simulator provides data at every 25 ms interval, thus a 3 minute flight has 7201 points for each
signal. The coefficient of cross correlation is therefore calculated as follows:
3
CORR(j) =
( -')
(7200-j) _5'h' h i+ji=l
To cover a time gap, x of 20 seconds, CORR(j) is calculated for j -- 1, 2 ..... 801. For 100 second time
gap j ranges from 1 to 4001. It is suggested that care be exercised in interpreting the long time gap
correlations. For the correlation with zero time gap, the averaging is based on 7200 pts, with 20
second time gap the averaging is based on 6400 points, and with 100 second time gap the averaging
is based on 3200 points. The cross correlation coefficients between any other two signals are defined
and calculated in the same manner. In the above formulas 6 and h may be replaced by any other two
signals between which the cross correlation is desired. In the present work, the cross correlations are
considered between 8 and h, 8 and dh/dt, 8r and _, 8a and % 8r and 8, 8a and 8, 8r and 6a.
THE PILOTS AND THE TEST FLIGHTS
Four volunteer pilots conducted the required flights on the F-15E real-time simulation. Three of
them were experienced licensed pilots and one had no prior experience of flying. In this report, the
volunteer pilots are named P1, P2, P3, and P4. The ages, flying hours, and the license status of the
pilots were P1 : 69 yrs, 3200 hrs, SEL, MEL, INST; P2:48 yrs, 100 hrs, SEL, MEL,/NST; I:'3:'23 yrs,
830 hrs, SEL, MEL, INST; and P4:26 yrs, no flying experience.
In every test flight, a pilot was required to maintain straight level flight for three minutes duration
heading due North, with preadjusted throttle for flying at 450 knots indicated at 10,000 ft. altitude
with autotrim OFF. During the flight, the pilot inputs were limited to the joystick movements only.
The volunteer pilots were provided with the facility to train themselves on the simulator by
conducting several three minute flights until they reached their own asymptotic performance in terms
of (h' + 10_g') a weighted sum of the rms values of fluctuations in altitude and heading in ft. and
deg. respectively. Every pilot conducted two sets of straight level flights. One set of flights was
obtained without any additional delay between the stick input and the appearance of airplane
response on the computer monitor. In another set of flights, a 500 ms additional delay was used.
Including the flights with the asymptotic performance, the pilot P1 conducted 15 flights without
delay and 15 flights with delay; P2 conducted 16 flights without delay and 21 flights with delay; P3
conducted 22 flights without delay and 17 flights with delay; and P4 conducted 40 flights without
delay and 25 flights with delay. From amongst the asymptotic performance flights, for each of the
two kinds of flights, and for each pilot, based on the value of (h' + 10_g'), the best flight, the average
flight, and the worst flight were picked up for further analysis. For all twenty four of these flights,
4
WilliamsandJohnson(1997)havereportedaveragevaluesandrmsvaluesof altitudeand headingand_i-hcorrelationversustimegapgraphs.
A comparisonof theaverageandrmsvaluesof altitudeandheadingfor thefourpilotsindicatedthatthepilotsP3 and P4 had remarkablysmalldeviations,whereaspilots PI and P2 had ratherlargedeviationsfromthedesiredvalues.It is believedthatthecrosscorrelationsbetweenanytwo randomly
changingbutrelatedvariableswouldbebetterrepresentativeof theirmutualinterdependenceif theirdeviationsfrom thedesiredquantitiesarerathersmall.Thereforethecorrelationsfor the flights of
pilotsP1andP2arenotreportedheredueto their largedeviationsin _ andh. Theymayinsteadbeseenin WilliamsandJohnson'spaperprovidedin AppendixB. In themainreport,thediscussionsoncorrelationsdwell uponfive differentflightsandthosearethebestflightsof pilotsP3andP4without
delay,thebestflights of pilotsP3andP4withdelay,andtheaverageflight of pilot P3with delay.Theaverageflight of pilot P3withdelayis includedin thediscussionbecausethebestflight of P3showedanunusualcorrelationfeaturewhencomparedwith theotherflightsof pilotsP3and P4.The
graphsof therawdatafor theseselectedfive flightsareprovidedin AppendixA.
FLIGHT TEST RESULTS
For comparison of pilot's performance in flights without transport delay and those with 500 ms
delay, the rms values of altitude and elevator deflections for the best flights of the four pilots are
shown in Table I. For both kinds of flights, pilots P3 and P4 have achieved better performance
compared with pilots P1 and P2. For the flights of pilots P3 and P4, the rms values of altitude and
elevator deflection for flights with delay are approximately twice their respective values for flights
without delay. It is obvious that the 500 ms transport delay has resulted in deteriorated pilot's
performance. With the existing related publications in view, this is not surprising. For example,
Middendorf et al (1991) have noted that 300 ms transport delay resulted in significant deterioration
of pilot's performance.
In a study of pilot dynamics and pilot's choices of feedback alternatives for controlling altitude, the
frequency response functions of pitch angle and altitude have been obtained and used by Goto and
Matsuo (1988). In the present work, the cross correlations between the control surface deflections and
airplane reponses are obtained to understand some aspects of the pilot dynamics. Let us start with the
consideration of cross correlations between elevator deflection, 5 and airplane altitude, h. An upward
instantaneous deflection of elevator from its average deflection is positive _5 and a downward
deflection is negative 5, although the convention is to take downward elevator deflection as positive,
for example Etkin and Reid (1996, p. 33). A higher instantaneous altitude compared with the average
altitude of the three minute flight is positive h, and a lower than average is negative h. It is understood
thata positivechangein 5 resultsin a positivechangein h anda negativechangein _5resultsin anegativechangein h. Allowingfor a timelag betweenelevatordeflectionand altituderesponse,thetwo signals_5andh shouldhaveapositivecorrelation.
Figure1presents5-hcorrelationsfor pilotsP3andP4for a rangeof timegapsbetweenzeroand20seconds.Figure 1 hasthe graphsfor the five selectedflights, their selectionis indicatedin the
previoussection.All theflightsof Figure1 havea commonfeature- that they do not havelargedeviationsfrom therequiredaltitudeof 10,000ft; themaximumdeviationin averagealtitudeis 6 ft.in the best flight of pilot P4. All the flights showan appreciablemagnitudeof negative5-hcorrelationfor zerotimegap.Thisis expectedbecauseupon noticingpositiveh, thepilot movesthestickto deflecttheelevatordownwardandviceversa.
Letusconsiderthebestflight withouttransportdelayfor pilot P3asa typicaloneamongstthefive
flights of Figure1. In this flight the largestmagnitudeof negativecorrelationis reachedwhenthetimegapis 400ms;to talk aboutit, let uscall it thefirst timegapon thegraph.Thefirst timegap,perhaps,includesthreestagesof airplanealtitudecontrol: the pilot's reactiontime,the time gapbetweenstick movementand the resultingelevatordeflection,and the time gap betweenelevatordeflectionandaffectedaltitudechange.Thisconsiderationappearsreasonablewhenit is seenwithRoskam's(1995,p.765) assumptionsof 100 ms reactiontimefor testpilots and 120 to 200 msreactiontimefor otherpilots.Thecorrelationstaysnegativeup to 2700ms.Let thesecondtimegapon thecorrelationgraphbe thetimegapbetweenthelargestnegativemagnitudeof correlationandzero correlation,here it is 2700 400 = 2300 ms.The secondtime gap,perhaps,representslimitationson theratesof climbanddescent;higherrateswouldleadto lowervalueof the second
timegap.With transportdelay,thebestflight of pilot P3hasanexceptionalfeaturethatits' first timegapis zero;thisexceptionalfeatureisnot exploredfurther.Therawdatain AppendixA, however,showsnosignificantdifferencesin thebestandaverageflightsof pilot P3withdelay;but theelevatordeflectionsfor thefirst 30secondsin theflightswithdelayaredifferentfrom therespectivevaluesin
theflight withoutdelay.Dueto theexceptionalnatureof the bestflight of pilot P3withdelay,theaverageflight of P3is includedin Figure1to bediscussedasanothertypicalflight withdelay.Fourof thefive flights in Figure 1 maybe treatedastypicalstraightlevelflights; for themthefirst timegaprangesfrom 350msto 600msandthesecondtime gaprangesfrom 2300msto 5700ms.Oneof thereasonsfor largedifferencesin thefirst andsecondtimegapvaluesfor different flightsmay
be thatpilot's observationsandactionsarecontinuousratherthandiscretein nature;the pilot doesnot needto wait until elevatordeflectionor airplaneresponsecorrespondswith certain stick
movement.Thelargestmagnitudesof correlationcoefficientsandthefirst andsecondtimegapsarenot significantlyaffectedby the500 mstransportdelay.This is surprising,especiallywhenwenotethatthermsvaluesof fluctuationsof altitudeandelevatordeflectionin flightswith transportdelayare
approximatelytwicethecorrespondingvaluesin flightswithno delay.Therespectivermsvaluesare
6
providedin Figure1.It is understoodthata changein elevatordeflectioncausesapitchingmomentthatbringsabouta rateof changeof airplanealtitude.Therefore,for a moreclear viewof certainfeaturesof thesameflights,to substantiatethe_5-hcorrelationsof Figure1,the 5-hdot(hdot= dh/dt)
correlationsareprovidedin Figure2.
In Figure 2, all the flights, exceptthe best flight with transportdelay for pilot P3, show anappreciablenegative_5-hdotcorrelationat zero time gap.This is expectedbecauseupon noticingpositivehdotabovetherequired10,000ft. altitude,thepilot movesthestickto deflectthe elevatordownward,and upon noticingnegativehdot below10,000ft. he movesthe stick to deflect theelevatorupward.Thenegativevaluesof _5-hdotcorrelationsat zerotimegapin Figure2 arelessinmagnitudethanthecorresponding_-h correlationsin Figure 1. This is understoodbecauseuponnoticingpositive hdot for altitudebelow10,000ft. andnegativehdot for altitudeabove10,000ft.,thepilot doesnotneedto changetheelevatordeflection.
In thebestflight of P3 withouttransportdelay,pickedup hereasa typical flight, the correlationreacheszerowhenthetimegapis 400ms.Forthesameflight, thisreinforcestheobservationthatthelargestmagnitudeof negative_5-hcorrelationoccurredatthesametimegapasseenin Figure1.In allthegraphsof Figure 1 andFigure2, it isnoticeablethatthe_i-hdotcorrelationbecomeszeroat the
sametimegapsatwhichthe_5-hcorrelationshavelocalmaximumor minimumvalues.
In Figure2, letusconsiderthetwoflightsof pilot P4.Forthe_-hdotcorrelationsthenumberof zerocrossingsin20secondsfor theflight with transportdelayis twicethatfor theflight withouttransportdelay.A similarfeatureis not foundin thetwokindsof flightsfor pilotP3.It maybereiteratedherethatpilot P3is acertifiedandexperiencedpilot whereastheexperienceof pilot P4 is limitedto flightsimulatorsonly.FromFigureI, werecallthatthefirst timegapgoesfrom zeroto a valuewherethemagnitudeof negativecorrelationis largestandthesecondtimegapgoesfrom theend of thefirstgapto a valuewherethe correlationiszero.Let thesubsequenttimegapsrepresentthe subsequentzerocrossingsof thecorrelationstartingfrom thefirst zerocrossing.Perhapsthetimegapsbetweentheconsecutivezerocrossingswouldprovidean indicatorof the pilot's performance,largergapsrepresentingbetteror moresteadyperformanceprovidedthatthe deviationin altitudestayssmall.Suchan indicationis jeopardizedto acertaindegreebecausesticksignalin theflightsunderstudyiscontaminatedby electronicnoise.Forelevatorandailerondeflections,thenoisesignalis estimatedat
3%of a typicalstickmovement.Forrudderdeflection,thesignalis significantlylargerthan3%.Noattemptis madeto suppressor filter outthisnoisebecauseit requiresthepilot to remainvigilantif hewishesto maintainthe straightlevel flight. To find the subsequentzero crossingsof the _5-hcorrelations,thegraphsareobtainedfor timegapsfrom0 to 100seconds.Forthe sameflightswhichareincludedin Figure1,the100secondgraphsareshownin Figure3. As suggestedin thesectiononFlightDataProcessing,careis neededin interpretingthelongtimegapcorrelations.In Figure3, for
7
thetime gaprangefrom zeroto 100seconds,for pilot P3,a flight with transportdelayhas5 zerocrossingsandaflight withoutdelayhas16zerocrossings.Onthecontrary,for pilot P4,a flight withtransportdelay has 24 zero crossingsand a flight without delay has 7 zero crossings.UsingLanchaster'sestimate(referenceEtkin and Reid 1996, p. 172), the time period for phugoidoscillationsat450kts.or 760ft./sis 104sec.whichimpliesnearly2 zerocrossingsin 100seconds.Itis notclearif thephugoidoscillationshavea role here.Basedon rmsvaluesof h and8 fluctuationsonly,theperformanceof bothpilotsisnearlythesame.Havingfewerzerocrossingsin flightswithoutdelay as comparedwith the ones with delay is intuitively more acceptable.Therefore, theperformanceof pilot P4is somewhatpredictablebut thatof P3is surprising.It maybe interestingtoreiteratethatpilot P3is a licensedandexperiencedpilot but theexperienceof P4is limited to flightsimulatorsonly.
Figure4 showsthecoefficientof crosscorrelationbetweenailerondeflectionsandairplaneheadingversustimegapfor thesamefive selectedflightswhichhavebeencoveredin Figure 1.For four ofthe flights the graphshavesimilarfeatures.For all the five flights, the largestmagnitudeof thenegativecorrelationoccursat thetimegapvaluesbetween450msand725ms.This timegaprangeiscomparablewith 350to 650ms,whichis thefirst timegaprangefor 8-hcorrelationsin Figure1. Forzerotimegap,anappreciablenegativecorrelationexistsin all thefiveflights.Forbothpilots,thermsvaluesof ailerondeflectionsfor flightswithdelay areappreciablylargercomparedwith thoseforflightswithoutdelay.Thermsvaluesof airplaneheadingdo nothavethe samecharacterasthatofthermsvaluesof ailerondeflections.
Figure5 providesthecoefficientof crosscorrelationbetweenrudderdeflectionandairplaneheadingversustimegap.In thebestandaverageflightswithdelayfor pilot P3 appreciablechangesappearinthecorrelationwith timegap.In therestof thethreeflights thecorrelationisratherinsensitiveto thetimegap.Therawdatain AppendixA helpsto tracethis insensitivityto an inactivityon thepilot'spart in usingtherudder for headingcontrol.Thepilot P3hasusedruddermovementmuchmorethanpilot P4.For theaverageflight of pilot P3withdelay,therudder-headingcorrelationin Figure5hassimilarfeaturesastheaileron-headingcorrelationin Figure4. Thissimilaritymaybean indicatorof coordinateduseof rudderandailerons.
Figure6 hasthecoefficientof crosscorrelationbetweenelevatordeflectionand aileron deflectionversustime gap.Thecorrelationfor the five flightsdoesnot seemto haveany commonfeaturesexceptthatthecorrelationis rapidlychangingbetweenpositiveandnegativevalues.
Figure7 hascoefficientof crosscorrelationbetweenelevatordeflectionandrudderdeflectionversustimegap.Fourof thefive flightsshowappreciablenegativecorrelationsfor zerotimegap,although
sucha featurebetweenthemovementsof elevator and aileron may not be worth emphasizing. In
general, the correlation graphs for the five flights do not have any common features.
Figure 8 presents the coefficient of cross correlation between rudder deflection and aileron deflection
versus time gap. For different flights, at zero time gap, the different values of correlation coefficient
appear to represent the pilot's coordinated use of rudder and aileron. No common features are
discernable between different flights, perhaps, because of the lack of use of the rudder by the pilots.
The hardware facility in the present simulation system, perhaps, demotivated pilots from using the
rudder. The system hardware does not have rudder pedals, instead, the rudder pedal movement is
simulated by the rotation of the joystick as described in the section on "The Simulation System."
An overall consideration of all the correlation graphs suggests that for reasonably well defined flights,
the cross correlations between pilot input and airplane response versus time gap reveal interesting
features of the pilot dynamics. When the aileron-heading and the rudder-heading correlations for a
flight are considered together, they reveal the presence or absence of coordinated use of ailerons and
rudder by the pilot. Consideration of correlations amongst the parameters that represent longitudinal
motion of the airplane and correlations amongst the parameters that represent lateral motion of the
airplane support the generally used approach of separate studies of longitudinal and lateral dynamics
of an airplane in flight.
Additional research work may be recommended to find certain characteristic features of cross
correlations between pilot input and airplane response. Such characteristic features would then help in
determining different progressive stages of pilot training on simulators.
Pilot II h.' I II h:P1 73.8
P2
P3
P4
51.2
7.5
9.4
0.60
0.51
0.23
0.25
77.1
66.4
18.2
18.7
1.02
0.20
0.45
0.72
Table I: Comparison of the Pilot's Performance in Their Best Flights With and Without Transport
Delay, h' and _i' are the rms values of altitude changes and elevator deflection changes, respectively,
subscript n is used for flights with no delay and d is used for flights with 500 ms delay.
9
... I ...|. --
"_; _ _if, I I--'I--'--__] ni_
P3 _B. hdop
'P3 B.hdep" •
0. 4 COBRmax -0. 103773 0 tau m3.8
COItlm£n m-0.356171 0 t.au -0.4
svoif a 10004.4 avoDEP m -2.59663
rmsB - 7.55242 rngDKP - 0.2320110.21'
)" oJ.-0.2k
-0.4
• i * i * i i i i
2 4 6 8 10 12 14 16 16 20
TAfJ
P4_B ohdep
'P4 B.hdep' *
0-4_:ORRnex .0.236S67 • t;au -8.3
[C::OllmJLn m-0.342996 @ t:au -0.6
zaveB e, 10006 ,voDUrp m -1.90002
-o.4_
0 2 4 6 8 10 12 14 16 18 20TAU
P3_BD.hdep
0.8{" 'P3_OD.bdop' •_Ollmez m0.642921 0 tau m16.4
-0.41'
-0.61'
-0.1 I'
0
* a a , , i J i u
2 4 6 $ 10 12 14 16 18
TAU
2O
II
P4_BD.hdep
0.6. 'P4_BD.hdop" ,
:Ollmmz n0.120?24 • tau m6.3
_OllnAn m-0.47902S 0 tau u0.35
0.4 svoH - 10000.9 sveDEP - -1.7435
rnsB m 18.6935 rmsOlP - 0.72454
0.2
_OlR 0-0.2
-0.4
-0.6
_2oTAU
0.6
0.4
0.2
colt, 0
-0.2
-0.4
P3_AD.bdop
'P3ADohdop'
Oglmax m0.325452 0 tau m19.975
=Ollm£n m00.464615 • tau 00.425
tvoB m 10005.5 evoDKP - -1.55348
rmmi m 21.7423 rmmDKP - 0.462317
-0.6 .
i I I ,
0 2 4 6 8
i * * i i
10 12 14 16 18
TAU
20
Figure I: Cross Correlation Coefficient Between 5 and h
vs. x for the Five Selected Fiights of Pibts P3 and I)4. B
is the pilot's best flight without additional transport
delay. BD and AD are the pilot's best and average
flights with additional transport delay of 500 ms. (5 =
elevator deflection, h = airplane altitude, x = time gap).
10
0.4
0.2
*"ORB 0
-0.2
-0.4
J i
0 2 4
P4_Bb.cLhe
'Pl,Bb.aiho' •
CORRmlx mO.36BgB5 0 rio m3.575
n I i , i i J
6 $ 10 12 14 $6 18 20
TAU
i
0.61
0.4
0.2
301i 0
-0.2
-0.4
-0.6
0
II IIP3_BDb.c lho
'PqBOb.cihl' •
;Oltimax m0.398765 O tiu m3.45
i n i n n J I i i
2 4 6 8 10 12 14 16 18 20
T&U
P4BDb.clht
L 'P4_BDb.aihe' '0.8 OERmsx -0.660987 • teu -1.525
_OIL_ltin m-0.296106 0 tau mO
0.4 _v_a_ - 0.115164 IvoOKP - -1.7435
_o_" 8. 032S6 *..D,, - 0.724540.4i( t
0.2 II/l\_
_Oil 0
-0.2
-0.4
-0.6
-0.8
i i J i a | | I o n
0 2 4 6 8 10 12 14 16 18 20
TAU
__.____lL I I I II II I "_i
0.4
0,3
0.2
0.1
-0.2
-0.3
-0.4
n
i
_3_AOb.c lha
" P3_ADb. c .I.lle " ,
:Olllmax m0.3'1986_. M tau m4,'lS
i I , n I I i I I
2 4 6 S 10 12 14 16 18 20
TAU
Figure 2: Cross Correlation Coefficient Between 8 and
dh/dt vs. • for the Same Selected Flights Which Arc
Covered in Figure I. (8 = elevator deflection, h =
airplane altitude).
II
P3Bc.hdep
'P3_Bc.hdep' -
0.4 :ORRmax mO.245911 0 tsu m92.575
CORRm_n --0.356171 e tau mO.4
avefl m 10004.4 aveDEP - -2.59663
-0.4
i J i
20 40 60 60 1OC
TAU
P4 Bc. hdep
"P4 Bo.hdep' •0.8
_ORRmax n0.625233 0 cau o94.975
0.6 :ORRmXn m-0.34346 I) tau o85.75 d_kLveH m 10006 aveDJEP - -1.98802 /0.4_..msH u 9.38642 rnsO£P - 0.245737
0.21"
_OBR 0 [/-0.2
-0.4['
-0.61'
-O.8 I'
0
i
20 4O 60 8O 10(
TAU J
P3_BDe.bdep
t 'P3_BDc.hdop' •0.6 _OBBmax m0.6429_Z _ tau m16.4
DEP - -1.55635
m 0.45004
0.4_
COBB 0
-0.2
-0.4
-0.6
-0.8
0 20 40 60 80 101
TAU
P4_BDo .hdep
0.6 'P4_BDo.hdep' 4
:OBBaax m0.207152 i) t_u o45.775
.'OBBm£n m-0.479025 Q t:su mO.35
0.4 sveB m 10000.9 avoDEP m ol.7435
rmstf o 18.6935 :maDEt_ . 0.72454
0.2
:OBID. 0
-0.2
-0.4
-0.6
0 20 40 60 80 100TAU
.--,--,--.L _
0.5
P3_ADc.hdep
'P3_ADc.hdep' •
_ORBma: m0.349134 e tau m24.575
COBBmXn --0.78089 0 tau n43.575
svoB m 10005.5 eveOEP • -[.55348
ruaB m 21.7423 rmaDEP - 0.462317
60 80 10(
TAU
-L
i
0 20 4O
Figure 3: Cross Con'clarion Coefficient Between 8 and h
vs. x for z up to 100 Seconds. For the same flights,
goes up to 20 seconds in Figure 1. (5 = elevatordeflection, h = airplane _lrimde).
12
P3Do.pdmp
'P3_Bo.pdap' •
0.3 _ORRmex .-0.0594986 • tau -19.975
CORRILn 100.25259S 0 taU .0.725
0.2 pveeSZ - -0.340902 *veOAe - -0.00677262_aBESZ m 0.544824 FIsDAP - 0.046264
..iF
con o F................
-o.:, F
* * i r r i r * *
0 2 4 6 8 10 12 14 16 18 20
TAU
II
P4_Be.pdap
0.2 'p4_Be.pdap _ "
:Ollmlx --0.0995051 0 tau mlE.10.15 :ORRnAn --0.151271 0 _eu mO.7
lvOPSI " -0.S10125 aveDAP - -0.000792701
0.05
:Oil 0 .........
-0.05
-0.2
-0.15
-0.2
* i t i t * J i i
0 2 4 6 8 10 12 14 16 18 20
TAU
.- ms mn::lP
P3_BDI.pdtp
0.4_ ' P3_BDo. pdap* •
m-0.279572 0 t*u -0.
F:::::" 0.206,39.+.oA,. -oo..160.2 0.482938 =nsDAP 0.101864
::3/-o.4F • i i i i i i , t
0 2 4 6 8 10 12 14 16 18
TAP
0.4
0.2
00.4
20
P4.BDe. pdsp
'P4 BDe.pdap* "
:O|Rnex ,0.236366 e tau -10.375
:,BRian --0.372959 • ¢au .0.65
lvePSZ m 0.135562 aveDAP - -0.0352947
_usPSI - 0.881634 rnsD&P - 0.0135028
* i , i i * i i ,
2 4 6 8 10 12 14 16 18 20
TAU
P3_kDe.pdsp
0.4 'P3_ADe.pdap' •
:Ollmax -0.164808 9 _au -6.3250.3 _Ollm£= m00.290042 0 tsu m0.45
IVIRIZ e 0.260222 nveDA_ - -0.0412342
0.2 rIIPIZ m 0.50166 rlsDAP m 0.209219
-0.2
-0.3
-0.4 J , , i i i i i
0 2 4 6 8 10 12 14 26 18 20
TAU
, I
Figure 4: Cross Correlation Cocfficicm Between 8, and
¥ vs. % for the Same Selected Rights Which Are
Covered in Figure 1. (8, : aileron deflection, xp =
airplane h_ding).
13
o.41
0._11
0.2
0.1
¢011 0
-0.1
-0.2
-0.3
°0.4
P3_Bd.pd=p
"PS_Bd.pdwp" *
:OIRmox m-0.212113 0 tau m4.775
:OBRm,n m-0.320394 0 tau m18.22S
svePSZ " -0.340S02 *voOIP = 0.00345877
rmmeSZ - 0.544824 :moDRP m 0.00462673
0 2 4 6 8 10 12 14 16 18 20TkU
m_
P4_Bd.pdrp
'P4_Bd.pdwp' •0.4
:ORIImax m-0.264649 0 tau 012.15
0.3 :ORnmln m-0.329402 @ tsu 14.7Shve_OSZ i -O.510125 aveDIIP -- 0.003)709
0.2 :nsPSZ = 0.276421 rnsDlIP • 0.00264825
0.1
:011 0 ................
-0.1
-0.2
°0.3
-0.4
0 2 4 6 8' 10 12 141 16 18 20
TAU
mm_KJ
0.6
0.4
0.2
COB& 0
-0.2
-0.4
-0.6
0
P3_BDd.pdrp
'P3_BDd.pd_p' •
:OIImaz m0.420246 0 tsu m11.575
=ORlmlnJl_3582 0 ts_lmO_
tveP$_lJZ . _UJle_6eJmSsJmqJ_eDKPDRp 01_.00285273
* , , * * , | * ,
2 4 6 8 10 12 14 16 18 20
YAU
0.2
0.11
_ORIE 0 I
-0.1
-0.2
p4_DDd.pdrp
'P4BDd.pdrp" '
:ORBmax m-0.0441563 0 tau m4.325
:ORRa_n m00.187214 • tau mOtvoPSZ 8 0.135562 *veDSP = 0.000455656
_mIPSZ m 0.881634 rnsDEP = 0.00253255
, , , * | ,
2 4 6 8 10 12 14 16 18 20
TAU
b_qa_s-- l_
0.2
0.1
r'OIIt0
*0.1i
°0.2'
P$_ADd.pdrp
'P3_ADd.pdrp' ,
=ORBmsz m0.0621841 • tau m6.4S
:O1imLn =-0.155183 O tau =0.575
avePJZ m 0.260222 avuDRff m *0.00741358
=mSPJX m 0.50106 rmJORP m 0.0104627
* * , , , , * , ,
2 4 6 8 10 12 14 16 18
TAU
2O
!I
Figure 5: Cross Correlation Coefficient Between _, and
¥ vs. x for the Same Selected Flights Which Are
Covered in Figure I. (8, = rudder deflect/on, _ = airplane
heading).
]4
P3Bg.sdep
0.3 _OmRmaz -0.181784 0 taU --5.525
_ORKmla m--O.257131 0 taU m17.575
_voDAP " -0.0067?262 *veD£P " -2.59663
0"2_,DA, - 0.0R264 rnsD_P - 0.2320_
o-,k ^j\ / X /
i J i i i i _ i i i
0 2 4 6 8 10 12 14 t6 18 20
TAU
P4_Bg. adep
_0111max m0.0850131 g tau m10.425_OIIIlmlll m-0.117406 8 tau m16.725
0. I _veDAIP m -0. 000792701 IIveDEP - -1. 98802
p:m,DAP - O. O_.87 r_lq_45737
o.os f. _ / _ •
:0../ A.,/ i __!°P/-- I/ ....
'' V ',-o. os V , ._
"0"1 r
-0. 15
0 2 4 6 6 10 12 14 16 18 20
?AUI
1F3_BDq.adep
"@$_BDq.adep' ,o
FOilmax m0.108808 0 tsu -12.35
FORImLn m-O.Z2456§ 0 tmu m7.925
0.1 _voDAP i -0.0246216 avoD£Jl_ -1.55635
:.,t v V-_ .15F
* J i , + i r J J
0 2 4 6 8 10 12 14 16 18 20
T&U
cOl1 0
-0.11.
-0.2
P4 BDq adep
Lo,.2...o 17,31 . t.u .12 8 '.4-.h,-.d-.' .
0"2_OIImLII m--0.167236 81 tau mZ0+qrl_§
laveDA P . -0.0352547 aveDEP -i-_.743S
0.1 'mmD_lt_ m 0.0835021 rnsDf[P - fb.7_454
, , i i i i i * i
2 4 6 8 1.0 12 14 16 1,8TAU
2O
i
m
P3_ADqsdep
.15' "P3_ADq. Idop" ,
=Ollmlx m0.103305 0 tSU mO
1 _O|ls£n moO.IllS32 I tau m7.6]_lfoDAP - -00412342 avoDEP - -2S5348
sDA_ m 01052SS rnaDE_ m 0.46231?
CC • 0 ..................................
- .05
0.1F
- .25i l l i l i i _ l
0 2 4 6 8 10 12 14 16 18 20
_XU
Figure 6: Cross Correlation Coeffi¢icnt Between _ and
8, vs. z for the Same Selected Flights Which Are
Covered in Figure 1. (_5 = elevator deflection, _, =
aileron deflection).
lS
II_I_L i ----
P3Bf. rdep
0.2 :Omimsz m0.18150$ 0 tJu -3.6
;OJIJIJl+i,I+_I¢O. '1299+'J 6 I_ t, mu 1.1) l,,veDRIp"'b 0"%.00345577 .veDIIP - -2_663 I
_'IIll - O. 46]:673 =.sOEP . 0 ,_0.I
:011 0 ]
/"0"11-0.2
* I n i i i i i n
2 4 6 8 10 12 !4 16 18 20
?AU
P4_B_.rdop
0.4 _u mS,47..'Pq..Bf.rdop'_" ' ,,
0,3k_O,q/Im.tn m-0_302588 I tau ..0.2
__035709 _v.DSrp . -I..98502
-...ooo![,...,............:.-,.-+-.-.----0
-0
0 2 4 6 IS 1O 12 14 16 . 18 20
?AU
__...... -_--_+_
P3_ID f. rdep
0.4 _
-_OIIIID ---0.351578 B tdIU mO.§5
.v.DIP - 0.00285273 .v.DEp . -1.55635
-0.4
0 2 4 6 $ 10 22 14 16 18 20
TAU
O.
0.1
O.
0.0:
'11 ,
0.0t
-0, :
-0.;
P4_BDt.:dep
, 'P4_BDf.rdep' ,
Oltmax -0.0307454 e cau m3.05
20111£a a00.160276 0 _au m[0.TSIveDBP -- 0.000459656 sveDgP e "1.7435
PmoDIP " 0.002932S9 rnnDgP i 0.72454
TAU
P3_ADt.rdep
'P3oAD£.rdep' .0.4 _OIRmaa ---0.136746 @ ta_ -0.025
:OIInLa --0.345831 I tsu m8.425IVIDIP m "0.00741358 aveOEP i -1.55341
0.2 _IIDIP i 0,0104627 :IIDIP m 0.462317
11 0_ ...................................................................
0 2 4 6 8 10 12 14 16 18 20?&U
Figure ?: Cross Correlation Coefficient Between 8 and
8, vs. _ for the Same Selcct_.d Flights Wh/ch Arc
Covered in Figure I, (8 = elevator deflection, +, = mdclcr
deflection).
16
_ll BE]B
P3 _8h. adrp
0.:3 Oilimax _0.253288 • tau -0.25
0.2
0.1
COJt| 0
-0.1
-0.2
-0.3
'P3 Bh.adrp' ,
mmJLn m--0.0162541 10 tau u19.97 c;
AP w -0.00677262 svJl_nP - 0.00345877
AP m 0.046 0.00462673
• 2 4 6 8 10 12 14 16 18 20
TAD
®U [ I
P4Bb.md_p
:O||m*x -0.0763326 0 tsu -18.075 |;OiRmln m-0.0205183 @ tmu m2.175 _ |
,veDA, - -0.0007,2701 ,veDB_ w O_i_03_ ;
0.05 _-.o^P - 0.02,5287 ..,o. - 0.0¢z._._. .;,. f
-0.05
-0.1
i , i i
0 2 4 6 8
I i i | i
10 12 14 16 18 20
TAt/
j inl i
P, BDh.adrp
0.3'P, BDb.ad=p' ,
:OBRma_ -0.208236 • tau _0
0.2_ORRmln m-O.137686 @ tau m3.425
_veDAP o -0.0246115 sveDRP m 0.00285273
_pDAP m O.101864_T_msDRP m 0.00460978
-0.2_
-0.3F
0 2 4 6 8 10 12 14 16 28T&U
P4_BDh. adwp
0.,F' P4_BDh. 8drp '
C01tlmax m0.215876 0 tau ,,0
0.2FOItEmLm m-0.0583524 • t:atu m16.175laveDAP w -0.0352,47 aveDRP - 0.000455686
_msDAP o 0.083,028 rmsOmP m 0.00293255
)aa o [ 3t_%awrqm_d_4_ _ _: _q
-0.11.
-0.2 F
-0.3
0 2 4 6 8 10 12 14 16 18 20
TAU
mm
P,_ADh.adrp
0.4F
'P$_ADb.sd=p ° ,
0 3_Ol_max m0.277801 ff tau mO
" _ORRm£a m-0.167684 • ttu m17.,5Ldrm,oDAP m "0.0412342 avoDRP m -0.00741358
O-2_sDAP m 0.109285 ::8DRP m 0.0104627i !
_Oll 0 ..........................
-0.1
-0.2F
-0.31.
-0.4k , m i i i J i i
0 2 4 6 8 10 12 14 16 18 20
TAU
Figure 8: Cross Correlation Coefficient Between _, and
5, vs. _ for the Same Selected Flights Which Are
Covered in Figure 1. (8, = rudder deflection, 8, : aileron
deflection).
17
CONCLUSION
In the study of straight level flights on a simulator with pilot in the loop, the root mean square values
of altitude variations for flights with a 500 ms additional delay are twice or larger compared with their
respective values for flights with no delay. This is in agreement with the other recent studies which
indicate deterioration in flight with 300 ms transport delay. The graphs of cross correlations between
pilot input and airplane response versus time gap reveal interesting features of the pilot dynamics. A
time gap of 350 to 600 ms appears to include three steps of altitude control, these are pilot's reaction
time, time gap between stick movement and resulting elevator deflection, and the time gap between
elevator deflection and affected altitude change. This observation, however, requires further
exploration because introducing a 500 ms additional transport delay does not make any visible
impact on the correlation graphs of a flight.
The correlation between elevator deflection and time derivative of altitude becomes zero at the same
time gap when the correlation between elevator deflection and altitude has a local maximum or
minimum value. An examination of the aileron-heading and the rudder-heading correlations suggest
that similarities or lack of similarities between the two kinds of correlations correspond to the
presence or absence of coordinated use of aileron and rudder by the pilot. Additional work is
recommended to investigate cross correlations between pilot input and airplane response to search for
their characteristic features which would determine different progressive stages of pilot training on
simulators. The UNIX based simulation system that has been installed and used for the present work
is suitable for a future study of the Pilot Induced Oscillations. The installation of rudder pedals
suitable for the UNIX based simulator system is strongly recommended as an improvement in the
system.
18
NOMENCLATURE
Symbol
ms
rms
t
"u
"um
H (t)
avg H
h (t)
h'
hi+j
V (t)
A
avg A
5 (t)
_ (t)
5r (t)
CORR (x)
DEP
DAP
DRP
HDOT
PSI
SEL
MEL
INST
AIAA
NASA
PIO
CPU
SGI
Description Units
Milliseconds.
Root mean square.
Time. instant
Time gap. seconds
Time gap for the maximum absolute value of negative correlation, seconds
Instantaneous value of airplane altitude, feet
Time average value of altitude for a 3 minute flight, feet
H (t) - avg H, fluctuating part of altitude, feet
rms value of h.
h at the (i + j) time interval of a flight, each interval is 0.025 ms. feet
Heading, zero indicates Northward direction.
(t) - avg _, fluctuating part of heading.
Elevator deflection, positive for upward deflection.
Time average A for a 3 minute flight.
5 at the i'h time interval of a flight, each interval = 0.025 ms.
rms value of 5.
A (t) - avg A, fluctuating part of elevator deflection.
Fluctuating part of aileron deflection.
Fluctuating part of rudder deflection.
Coefficient of cross correlation between two signals with time gap x.
Stick deflection for elevator or deflection of elevator, according to context.
Stick deflection for aileron or deflection of aileron, according to context.
Pedal movement for rudder or deflection of rudder, according to context.
Time derivative of H. ft. / sec.
Heading, zero indicates Northward direction, degree
Single-Engine Land (aircraft type).
Multi-Engine Land (aircraft type).
Instrument Rating (FAA-certification).
American Institute of Aeronautics and Astronautics.
National Aeronautics and Space Administration.
Pilot Induced Oscillation.
Central Processing Unit.
Silicon Graphics, Inc.
milliseconds
19
REFERENCES
. McRuer, D. T. and Graham, D. (1981), "Eighty Years of Flight Control: Triumphs and Pitfalls of
the Systems Approach," Journal of Guidance, Control, and Dynamics, Vol. 4, No.: 4, pages 353 -
362.
, Ashkenaz, I. L. (1984), "Twenty Years of Handling Qualities Research," Journal of Aircraft, Vol.
21, No.: 5, pages 289 - 301.
3. Washizu, K., and Miyajima, K., (1963), "Controllability Limit of Human Pilot," AIAA Journal,
Vol. 3, No.: 5, Pages 941 - 947.
4. Smith, R. H., (1963), "Comment on Controllability Limit of Human Pilot," AIAA Journal, Vol. 3,
No.: 10, Pages 1980 - 1981.
5. Smith, R. H., (1977), "Theory for Longitudinal Short Period Pilot Induced Oscillations," Air
Force Flight Dynamics Lab, Wright Patterson AFB, OH, AFFDL-TR-77-57.
,
.
Bjorkman, E. A.,1986, "Flight Test Evaluation
Induced Oscillations," Air Force Institute of
AFIT/GAE/AA/86J-1, December 1986.
of Techniques to Predict Longitudinal Pilot
Technology, Wright Patterson AFB, OH,
Hess, R. A., and Kalteis, R. M., 1991, "Techniques for Predicting Longitudinal Pilot Induced
Oscillations," Journal of Guidance, Control, and Dynamics, Vol. 14, pages 198 to 204.
, Cardullo, F. M., and George, G., 1993, "Transport Delay Compensations: An Inexpensive
Alternative to Increasing Image Generator Update Rate," AIAA Paper No. 93-3563-CP.
, Middendorf, M.S., and McMillan, G. R., 1991, "The Effect of Simulator Transport Delay on
Performance, Workload, and Transport Activity During Low Level Flight," AIAA Paper No. 91-
2965-CP.
10. Williams, J. P., and Johnson, Y. A., 1997, "Pilot Input and Airplane Response Cross Correlations
on a Flight Simulator With and Without Transport Delay." Presented at the AIAA Southeastern
Regional Student Conference, Atlanta, GA. (A copy of the paper is included in Appendix B).
20
11.Norlin, KenA., 1995, "Flight Simulation Software at NASA Dryden Flight Research Center,"
AIAA Paper No. 95-3419.
12. Roskam, J., 1995, "Airplane Flight Dynamics and Automatic Flight Controls," Parts I&II, DAR
Corporation, Lawrence, KS.
13. Goto, N., and Matsuo, T., 1988, "Identification of Pilot Dynamics in a System with a Choice of
Feedback Structures," Journal of Guidance, Control, and Dynamics, Vol. 11, No. 2, pages 159-
166.
14. Etkin, B. and Reid, L. D., 1996, "Dynamics of Flight Stability and Control," Wiley, New York.
21
APPENDIX A
The graphs of the raw data for the five selected flights of pilots P3 and P4. These
flights are discussed in the main text.
B represents the pilot's best flight without additional transport delay.
BD represents the pilot's best flight with additional transport delay.
AD represents the pilot's average flight with additional delay.
The x-axis in all the graphs shows time from 0 to 180 seconds. The control input
signals and the airplane response signals are plotted on the y-axis.
H = airplane altitude (ft.).
HDOT = dh/dt (ft./sec.).
PSI = airplane heading (radians in this appendix, degrees in the main text).
DEP represents joystick movement (in.) to deflect the elevator.
DAP represents joystick movement (in.) to deflect the ailerons.
DRP represents rudder pedal movement (in.) to deflect the rudder.
On a proportional scale, DEP, DAP, and DRP may also be interpreted as the
deflections of elevator, ailerons, and rudder, respectively.
A1
l__B_Izt P4_Bxlat
"4O 3
5_1. A
._!t_V' w'_ u"!l vW "it _ v u _ - v
0 ,;n ]00 1_0
• sO 3
'[/I - A.-,^]-"-." L
=" r,A^fl. ^ ^ A$,
-$.
0 ,_n IGI 1 sn
dO 3
i .... i .... , , * ,
]__BDAat
.... i .... i .... ! • , ,Inn 150
oJ _
O 50 100 150
dO 3
P4 BDAat
lo-[_VI/'VVIIV "V'vv_,IvV U"' "11_ '_ I'ib' II v_l-2o'.y, _ - v! v U i"
O _ 100 I_0
002- ; i /;_
O 50 I00 1_
dO 3
I__ADAat
"-,, ........ i ........O 50 100 1_0
"=: A fh,ooi - _ _j'_,
Figure AI: Airplane altitude, time derivative of altitude,
and airplane heading vs. time for the five selected flights.
For nomenclature, please see the first page of Appendix
A.
A2
P_J_B_lat P4_Bxlat
-2o+"54"_°+f| ' --"_. :" L A .| ., 1_ • _ IL . . t _, J_ • .
3a 'I1/'P' Iv l'J'_ i'_v- V'vl/"_/_'v _V "v-v-- "v 1'-" Y ''-'_ _I
o35-! ............. ]
0 50 100 150
03-02 -;
Ot-
O0 2
-01 £
-02 -
JO )ltO ]1_1
-I0 ,,
-2_
0 r.n 1nn 1._
D3_
0.3-
- ;! " , . . ",' . , . . " . . " . . , . , , J
0 50 Inn I sn
0co.
001-" j . --me
0m -+_-ore 2
-o_ -"
0 os 100 Imn
P3_BDAat
0 SO 100 150
--o11'
P4_BDAat
O,5O0
-05,ol 0,-lJ-20
-25
i • - i i ' " _ --w
0 30 100 1';_
002,_0M_
OO0-
-00t
.0j_ -
i0 .... _3 .... Inn 150
1-" I .... D]_03
:-_'_-:................... t ....... b
0 ,.n 1_0 1.._
o,.(_ --o,,_
O _ 100 1_
0 50 I00 I._I
P3_ADAat
o:L" i -"
0 _ 100 1_
04Y , -- Da..e
.i} _,._, _.-._,,,,--+,_.+u+,+re,w,,w,,-_ --.,-_.+._
.o,+! : r ! I0 _ I00 I_0
-,,--.II .,kJ -- J _+I l l I_II0I'1"1"
ii i -+......:---0 50 100 I_
Figure A2: Pilot inputs to move the elevator, aileron,
and rudder, respectively vs. time for the five selected
flights. For nomenclature, please see the first page of
Appendix A.
A3
APPENDIX B
Copy of the paper "Pilot Input and Airplane Response Cross Correlations on a Flight Simulator With
and Without Transport Delay. " The paper was presented by Jason Paul Williams and Yusef Ali Johnson at
the AIAA Southeastern Regional Student Conference, held on April 10 and 11, 1997 in Atlanta, GA. The
work reported here was performed as part of the present project under the NASA Dryden grant number
NAG 2 - 4006.
B1
PILOT INPUT AND AIRPLANE RESPONSE
CROSS CORRELATIONS ON A FLIGHT SIMULATOR
WITH AND WITHOUT TRANSPORT DELAY
Jason Paul Williams
Under_aduate Senior
Yusef Ali Johnson
Undergraduate Senior
Aerospace Science Engineering Department
Tuskegee University
Tuskegee, AL 36088
Presented to:
The AIAA Southeastern Regional Student Conference
April 10, 11 1997
Atlanta, GA
•____....aper has bee'? re_'iewed and appr°ved f°r presentati°n by: ___
Dr. Vascar Godfrey Harris, Professor Dr. Syed Firasat Ali, Associate Professor
This paper has been reviewed and approved for presentation by: (')
Dr. S yed F. Ali
AIAA Faculty Advisor
B2
Abstract
Four volunteer pilots were required to maintain given altitude and heading on a computer based stationary flight
simulator of the F-15E fighter airplane. One set of flights was obtained without any additional delay between the
stick input and the appearance of airplane response. In another set of flights, a 500 ms additional delay was used.
The deviations in altitude and heading and the fluctuations in the altitude indicated that the flights with 500 ms delay
were appreciably deteriorated. This is in agreement with the other recent investigations. The cross-correlations
between elevator deflection and altitude reveal interesting aspects of pilot's performance. In particular, these
correlations reveal that different pilots exercised a variety of techniques for maintaining straight level flights. The
C++ source code, which is used for the computation of the desired correlations, is also included.
B3
Introduction
For human operators in a feedback control system, their ability to adapt to variations in control tasks is generally
acknowledged as a significant advantage. Baron _ says, "With training and proper feedback, pilots are capable of
adapting their control strategies to a wide range of task variables." Baron, however, reminds us that this adaptive
capacity is not unlimited. The transport delay or a delay between input signals and their response is one of the most
important factors that needs studies for different flight situations to understand whether they fall within the pilot's
adaptive capacity. The papers published by Cardullo et al 2, Lusk et al 3, and Middendorf et al 4 are some of the recent
studies on the effect of simulator transport delay on pilot's performance and on the ways of compensating delays.
The present paper offers a study of the airplane responses to stick inputs with and without an additional transport
delay. Four volunteer pilots conducted straight level flights on an F-15E flight simulator. The pilots used the joy
stick to maintain a given altitude and a given heading. Two sets of flights were conducted. In one set the standard
simulator was used and in the other a 500 ms additional delay was programmed between the stick input and the
appearance of airplane response on the monitor. The deflections of elevator and ailerons were treated as pilot's
inputs, and the altitude and heading of the airplane were treated as the responses. In the present study the average and
rms values of the elevator deflection and altitude signals as well as the cross correlations between these two signals
are presented as the statistical tools to compare pilot's performances. A C++ code to compute the desired cross-
correlations from the flight data is provided in the appendix. The present paper is included in the final report on a
flight simulation research project at Tuskegee University sponsored by NASA Dryden Flight Simulation Research
Center under the Grant NAG2-4006, June 1994 to January 1997.
Back_mound
The simulation software, which represents the dynamics of flight of an F-15E fighter, has been installed by NASA
Dryden Flight Research Center on an SGI 4D/420 IRIS workstation. On the IRIS alone, the simulator could be
operated by a mouse. To operate the simulator by a joystick, its graphic software has been moved to an SGI
INDIGO 2 workstation. The two workstations are connected through ethernet. Thus the F-15E simulation software
used in the present study operates on three central processing units, two on an SGI IRIS workstation and one on an
SGI INDIGO 2 workstation. The pilot in the loop operation is comprised of the simulation software, the pilot, and
B4
a BG System joystick called a FlyBox. Norlin 5 has provided a description of the software design, model
development, and simulation capabilities of the system together with a description of other NASA Dryden Systems.
The F-15E simulation is coded with a FORTRAN shell and C support routines and operates on a UNIX - based
platform. The integration scheme in the simulation has been optimized for real time operation. The code has the
flexibility to program a desired transport delay between the stick input and the airplane response on the monitor. In
a simulated flight, the digital data on forty-two different flight parameters can be recorded at every 25 ms time
interval. The flight parameters include the altitude, heading, and velocity, as well as the deflections of elevator,
rudder, and ailerons. Provisions for turbulence and gust, although available, were not used for this study.
Experimental Protocol
Four volunteer pilots conducted the required flights for the experimental study on the F-15E airplane real time
simulation. Three of the them were experienced licensed pilots and one of them had no prior experience of flying.
In this paper, the volunteer pilots are named as P1, P2, P3, and P4. The ages, flying hours, and license status of the
pilots are PI: 69 yrs, 3200 hrs, SEL, MEL, INST; P2:48 yrs, 100 hrs, SEL, MEL, INST; P3:23 yrs, 830 hrs,
SEL, MEL, INST; and P4:26 yrs, no flying experience.
In every test flight a pilot was required to maintain straight level flight for three minutes duration heading toward
North, with pre-adjusted throttle for flying at 450 knots indicated at 10,000 ft. altitude with autotrim off. During the
flight, the pilot inputs to the joystick were limited to elevator and aileron deflections only. A measure of
performance, named rsws score, was calculated for a flight from its altitude and heading data. An rsws score is the
square root of the sum of the weighted squares of maximum deviations in heading and altitude. Let A h in ft. be the
maximum absolute deviation from the required 10,000 ft. altitude. Let AW in degrees be the maximum absolute
deviation from the required Northward heading. Then
rsws = sqrt [(AW) _ + (A h/10) 2] (1)
It is obvious that a smaller value of the rsws score represents superior pilot performance. Two kinds of test flights
were conducted. One kind of flight simulated a normal flight without inserting additional delay between stick inputs
and the appearance of airplane response on the computer monitor. The second kind of flight had a 500 ms transport
delay programmed between the stick inputs and airplane response. For each of the two kinds of flights, a pilot was
B5
trained and evaluated by conducting several three minute flights and by calculating the respective rsws scores. An
asymptote in performance consisted of six consecutive flights of the same kind in which the pilot did not break his
own rsws score record. The flights of each kind that represented the asymptote in performance for every pilot formed
the data base for further analysis. In addition to the rsws scores, other criteria were also considered to compare
different flights. For the present paper, attention is focused on the elevator deflection, A as input signal and altitude
of airplane, H as the response or output signal. The time average values of A and H are avg A and avg H. The
fluctuating parts of the two signals are 8 (t) = A (t) - avg ,5 and h (t) = H (t) - avg H, where t represents time. The
rms value of 8 and h are denoted by 5'and h'. The cross-correlation between _5and h for arbitrary time gap x is
denoted as CORR (x) and defined as:
t=T
1 jCORR(_) = T 5' h' _i(t) h(t + "_) dt
t=0 (2)
The simulator provides data at every 25 ms interval, thus a 3 minute flight has 7201 data points for each signal.
The cross-correlation is therefore calculated as follows:
( -3
CORR(j) h(7200 --j) 5' h' i÷j
i= 1 (3)
To cover a time gap, ":, of 20 seconds for correlation, CORR (j) is calculated forj = l, 2 ..... 7201. For each kind of
flight for every pilot, the correlation graphs, CORR (_) vs x are obtained for only three flights amongst the six
flights that determine the asymptote in performance; these three flights are the best, the worst, and the average based
on the rsws scores. An initial estimate of 20 second maximum time gap, x, was considered sufficient for
correlation. Upon considerable computations we learned that longer time gaps would have been desirable.
Flight Data Analysis
In Table I, P1, P2, P3, and P4 are the four volunteer pilots who conducted two kinds of straight level flights of three
minutes duration. The column 1 for each flight has rsws scores. A lower value of rsws indicates better pilot's
performance The columns 2 and 3 have rms values of elevator deflection in degrees and altitude fluctuations in ft.,
B6
respectively; their lower values also indicate better pilot's performance. From Table I it is apparent that 500 ms
transport delay has caused deteriorated pilot's performance. Middendorf et al 4has noted that 300 ms delay resulted in
significant deterioration of pilot's performance. Lusk et al 3 have conducted an elaborate study on the delay
compensation. With 300 ms delay they observed more significant deterioration in heading performance than in the
altitude performance. Consideration of pilot P2 in Table I brings up an interesting aspect of pilot's strategy. Unlike
other pilot's, the pilot P2 has smaller rms value of elevator deflection for the flight with transport delay than the
flight with no transport delay. To obtain improved understanding of pilot's performance, cross-correlations are
obtained between the elevator deflection and altitude. From amongst the flights that represented asymptotic
performance of a pilot, only the best, the worst, and the average, are processed to report the cross-correlations. Thus
for the two kinds of flights by every pilot, a total of twenty-four cross-correlation curves are reported. Figures 1, 2,
3, and 4 show the cross-correlations between the fluctuating parts of elevator deflection, _5(t) and the altitude h(t + _)
for the time gap, x ranging between 0 and 20 seconds. The positive value of _ represents upward deflection of the
elevator that gives decrease in lift on the tail and which should result in nose up tendency of the airplane. Thus the
positive 5 is expected to result in positive h. Twenty-three out of twenty-four correlation graphs show negative
correlations for the time gap values near zero. While conducting their test flights, the pilots had observed that in the
event of leaving the stick free, the airplane had a tendency to climb up. To arrest the climb up tendency, the pilot
moves the elevator down that gives rise to negative correlation. Thus the negative values in the first part of the
correlation curves reveal lack of trim. Let "E_ be the time gap at which the negative correlation has the maximum
absolute value. The xmis ranging between zero and 600 ms and it does not seem affected by the transport delay. We
would expect xm to provide a measure of the pilot' s reaction time delay but its apparently unpredictable variations do
not meet the expectations. It may be noted here that Roskam 6 assumes 100 ms reaction time for test pilots and 120
to 200 ms for other pilots. For larger time gaps, most correlation curves show appreciable positive values. It is
suspected that they are caused by the Phugoid oscillation of the airplane, reinforcing the desire to extend time gap
delays to larger values.
B7
.
References
Baron, Sheldon, "Pilot Control." Human Factors in Aviation, Academic Press Inc., 1988, pages 347-385.
2. Cardullo, Frank M., and G. George, "Transport Delay Compensations: An Inexpensive Alternative to Increasing
Image Generator Update Rate," AIAA Paper No. 93-3563-CP, 1993.
3. Lusk, Steven L., C. D. Martin, and J.D. Whiteley, "Time Delay Compensation Using Peripheral Visual Cues
in An Aircraft Simulator," AIAA Paper No. 90-3129-CP, 1990.
4. Middendorf, Mathew S., and G. R. McMillan, "The Effect of Simulator Transport Delay on Performance,
Workload, and Control Activity During Low Level Flight," AIAA Paper No. 91-2965-CP, 1991.
. Norlin, Ken A., "Flight Simulation Software at NASA Dryden Flight Research Center," AIAA Paper No. 95-
3419, 1995.
. Roskam, Jan, Airplane Flight Dynamics and Automatic Flight Controls, DAR Corporation, Lawrenece,
Kansas, 1995, page 765.
B9
Conclusion
For all pilots observed in the study during straight level flight on a simulator, the additional transport delay of 500
ms, between the stick input and appearance of airplane response on the monitor, results in appreciable deterioration
of the flight. This is in agreement with the other recent studies which indicate deterioration in flight with 300 ms
transport delay. The cross-correlations between elevator deflection and altitude signals support a belief that in a non-
perturbed flight, different pilots exercise a variety of techniques for maintaining straight level flight. It is proposed
that cross-correlations between the elevator deflections and aileron deflections shall provide an improved
understanding of pilot's technique on a flight.
Acknowledgments
This work is part of a research project on "A Low Cost Simulation System to Demonstrate Pilot Induced
Oscillations," sponsored by NASA Dryden Flight Research Center under grant number NAG2-4006. We wish to
thank Dr. Syed Firasat Ali, PI on the project, for providing us with the opportunity to participate in flight
simulation research and for his guidance and encouragement. From NASA Dryden, Larry Schilling's support and
encouragement, and Ken Norlin's expert help on the simulation software remained vital for maintaining the progress.
Thanks are due to Dennis Ezell, whose contributions to the project have been significant in initiating flight
simulation research at Tuskegee University and in motivating fellow aerospace engineering students to engage in this
research. We are thankful to Dr. Amnon Katz from the University of Alabama and Dr. Vascar G. Harris from
Tuskegee University, who are both pilots and aerospace engineering professors, and who remained available for
valuable guidance and contributions. The advisement on software issues received from Dr. C.L. Chen, Abraham
George and Ernest Kashiri is sincerely appreciated.
B8
S.vmbol
t
"t
'Cm
A
avg A
8
8'
5 (t)
h i÷j
h'
h (t)
H
avg H
hu
A_g
rsws
CORR(x)
SEL
MEL
INST
Nomenclature
Description
Time
Time gap
Time gap for the maximum absolute value of negative correlation
Elevator deflection, positive for upward deflection.
Time average of heading for a 3 minute flight.
8 at the time interval of a flight, each interval is 0.025 ms.
rms value of 8.
A (t) - avg A, fluctuating part of heading.
at (i + j) the time interval of a flight, each interval is 0.025 ms.
rms value of h.
H (t) - avg H, fluctuating part of altitude.
Flight altitude.
Maximum absolute deviation in altitude, from 10,000 ft.
Time average value of altitude for a 3 minute flight, ft.
Heading, zero for Northward heading.
Maximum absolute deviation in heading, from zero deg.
Square root of the sum of the weighted squares of maximum
deviations in heading and altitude.
Coefficient of cross-correlation _5(t) and h (t + "_).
Single-engine Land (aircraft type)
Multi-engine land (aircraft type)
Instrument Rating (FAA certification)
Units
instant
seconds
seconds
degree
degree
milliseconds
degree
milliseconds
feet
feet
feet
feet
degree
degree
B 10
Pilot Flight Without Transport Delay Flight With Transport Delay of 500 msi 2 3 1 2 3
P1 73.8 77.1
P2 51.2
26.4 0.60
29.3 0.51
3.2 0.23
3.7 0.25
66.4
52.2 1.02
33.5 0.20
5.8 0.45
6.3 0.72
P3 7.5 18.2
P4 9.4 18.7
Table I: Comparison of Pilot's Performance in Flights With and Without Transport Delay.
1 - Average rsws for the best 5 flights, 2 - 5' deg in the best flight, 3 - h' ft in the best flight.
Bll
0.4
0.2!
=C |R
-0.2
-0.4
Pl_B.bdep
:01ho_ -0.Z_374S • _,, -_5.35 'PI B.hde_' "_QI|sLm 0-0.366042 | _*u o0.65
sveB e 9947.47 avoDKP m -2.05784
rnsJ e 73.J0|6 rmmDEP _ 0.603530
i i i i
S _0 25 20
T&U
PI_BD.hdep
o.s_oH.._ -o,xx3_24 0 _,. -7.4 'PI_BD+bd_p' "
_O|Bm_l 0-0.509733 0 _,m o0.3
0.4,vei - 9,6,.2 ...0,, . -1.621,3kles - 77.0513 rmaO[P o _.0_$76
0.2_
=CRI 0i ........ _ ..........
::::t/-o.6F
l , , e
0 5 L0 15 20T&U
Pt_A.hdep
• Pt x._a_ '0-0.372224 0 _eu 10.35
-0.4_
0 S L0 15 20
T&U
iml I
P1 AD.hdep
'ez_,_./,aep 'O.II ,
_'01mmetm 10.363204 0 _il mL$,975
0.6 _OR|mLI o-O. 62468 • ciu mO
Iv'el ,u 5605.67 avQDEP i -0.692137
0.4 ©mmB ,u L0]S.78 z'moOl[P m t.32_35
0.2COil 0 ............................................
"0" 2 _id/I/r
-6.4
-0.6
-0.8
* i i i
0 _1 10 25 2O
'tAU
I |1
0.4
0.3
0.2
0.1
:ORe' 0
-0.1
-0.2
o0. $
-0.4
Pl_W.hdep
• •:Oilu*z -0.266511 • Zlu el$.ST5
I i I i
S 10 _5 20
TAU
O.l
0.6
0.4
0.2
COlII 0
00.2
-0.4
-0.41,
*0. II i i
0 S 56
'Pl_WD.hdep '
PI WD.hdep "Q
:Ollmil m0.263664 • tam m19.$73
;OIRILI m-0.545517 Otiu mO
avoi - _563.01 eveD[P m 0.2201t5
_mll m 1650.04 rmeD[P u 2.25504
i
15 20
TAU
Figw_ L: Cross-CorrclmJons Between Flevazor Deflect*on and Altitude of the Airplane for Pilot P !
B, A, and W ax_ th© pilot's best, average, and worst flights without addidot_l ummspon delay
BD, AD+ and WD me me pilot's besL _Veml_ and worn tlbdB widz addilioul mmpo_ _y of _ _.
B 12
B. hdep
0. t _ 'P2 _B.hdep ' ,
_01lmax _0.30|113 Q _au -t7.6
0 S I0 _S 20
_AU
P'Z _B D._dep
• ,p.z 8D.a_po .0,4 _OlllQx n0.326244 g tlu _Zg.975
0 3 _OlRmza oO.OS43?OS 0 _lu .n• svoi o 10000.1 Jv_lmmlFrO.225584
O.ir _[_ " Q'2025'G
o.tb
&A o_ ..............................................
-0. LI"
-0.2 I'
-0.3 J"
-0.4 t"
0
i * i I
5 10 15 20
TAU
' PI_A.hdep'
o. o ' _ _A___ hdep ' i
:OItulx a0.213912 • _su ,1_._75
0,6 _OlluLs --0,$76133 • _8_ -O
IVI61 10275.8 IvoDK_ e -1.22585
0.4 rue• e 45S.?S2 cnaOEP e 0.72752_
:011 0 ...........
-0.2
00,4
-0.6
-0,6
T&U
02
0.1.
Ill 0
-0.1. i
-0. _t
0
__,'_).hdep
' ' .:0IInQI _O.l_2TSl g _lu _25.57$
_01lmL8 e-4.0427427 0 _lu eO.3avon • 10106.2 *veOEP • O.$7.lAlqJ"'_
i | , i
S 10 10 20TAg
_w. hdep
o.IL 'P2 w.hdep'-Ol|llx _0.07S5126 • _au _22.075 --
_0lln_: o-0.0410_
0.05 FUn• • 75 cuIOK_ m 1.007_
:ORB 0 -/ ....
/-0.05
-0.1
,| | n a
0 S 10 IS 20
T&U
lm
0.3
0.2
0.11
toll 0
-0.1
-0._
-0.3
P2_WD.hde_
' -- ------_P'_-WD-hd_':Ollnlu mO.OlO024 I _lU eI_.$5
:OllmLu --O._55077 6 _m *t. OS
jv+l m 10_16 *vIDE+ * 0.$11516
Pull * 110.123 +uIOgP • 0.$20114
__ ..........
0 S 10 lS 20
TAU
RiPm= 2: Cmu-Con'=hmo_ Between E;leva_r Dcflsc_on and AJrltude of the Airpla_ (or Pilot ! _.
B. A. and W as= the pilot's bes_, aver'_=, and worn tli¢t_s without addirloaal a-_asp_t delay
BD, AD, and WD az_ the pilot's bc_t, avenq_, Ind worn t'lieh_ with additioma_ tr_ulpo_ d_l_y of 500 ms.
B 13
,+
' FJ_B,bdep '
' p3_B.hde'p ' .0.4[COll|nax 10.203773 0 1:*_ -,3.8
ICOlln£n m-0.356171 • r..au ,m0.4
levell i L0004.4 *voDI[p e -2.55663
0 2_:msll * ?.$5242 :maOf[P _ 0 232011
:oS=.o.21.oF........................................
-0 41
i i i i
0 S 10 15 20
TAU
__BD.bde'p
o mF ' l__BD.bdc"_. OCOIIIII _0 642921 I _.*1 _14..4
0 ,=_:Olllm_Ln _-0 221474 0 tsv *" [dVoli _ 20003.7 *veO[P_t. SS6"tS
0 4p''" " 18 2_- 0.45004
o=0 l I 0 .........................................................-0 Z/ _
-0.41
-0.6d
I-0.Sl
i• * t
5 10 15 20
TAg
.r-÷:-, L
_bdep
_O|lleE .0.1S5334 I _*u .$.42S
IC:011mLn .-0.233303 0 t,lu *0.3?50 2bhveli m $gSqJ 38 4veDit t -3 04715
°*f/II 0 ............................ _ ........
-0 ]L-0.l
0 5 10 1S 20T&U
_,+,O.hdep'P3_AD._p •O.6F
:OIImax _0.3294S2 0 _eu _1997S
:01lm_l i-0.464815 • tsu e0.4250.4 _veE m 1000S.5 aveDEP • -1.SS348
0.2fnli _ 21.7423 _
oo..olS .......................................
-0.0F
a n i n
S 10 15 20TAg
V
0.6
0.4
0.2
:ORe' 0
-0.2
-0.4
-0.S
0
P3 :w. bdep
'P3 W.hdep' -
_O|Imax _0.45?64_ 0 _ou eL_.025
/
i i i i
5 1.0 15 20
_AU
0.6
0.4
O.Z
|ll 0
00.2
-0.4
-0.6 l,
0
___.bdep
' P3_WD.hdep ':Olln4x *0.42157 I :au -L?.S2S
/i i i
S 20 25 20
?&O
F_gum 3: Crous-Con'elmons Between Elevator Deflection and Altitude of the Airplane for Pilot P3
B. A, and W m'e the pilot's bcsL avera_, and worst flights without additional transpofl delay
BD, AD, and W'D am the piloCs best, avera_, and worst flights with additional translxx_ _ o(500 ms.
B 14
' P4 B.bde_ '
0-4_:Ollumant -0.236567 • tau .I.3 'p4 B.bdep' '
o -1.58802
O.2ImO E . 5._. 0.24(;737
+°t / ..............................0 2;.J
0 5 10 15
TAU
m_
' P4 BD.bde, p '
•0.6 'P4 BD. ' .Olantaz mO.120724 O tam o8. 3
:Ontltnm£nm a,00.479025 0 tau eO.3S
0.4 IveO o 10000.9 eveOl[P o -1.7435rill o 18.1_J3S rmoOEff e 0.72454
0.2
=OI1.10
-0.2
*0.4
-0.6
* a
0 5 10
| |
15 20
TAU
0.1
0.6
0.4
0.2
:O11, 0
-0.;1:
-0.4
-0.6
-0.1
P4 ,_,..hc_,,,p
' P4 A.bdep ' •=Ollmlz o0.640456 • tau o16.275
:Oilutu o-0.0466645 • t_
mv*_ - L0001.8 aveOj,,l_er--l.bJx_.
i i i i
$ 10 2S 20
T&U
0.2]
COBB 0l
o0.2!
-0.4
P4_AO.bdep
0.4_OISm*z o0.13S736 • _au _&.15
:OaR._a o00.34|851 • ta_ m0.4
I yea • 10013.1 aveOEP m -2.06157
" mar - 15.1606 =nsOll o 0.514265
0 5 10 15 2Ot&U
• ,r+_
' P4_W.hd_ '
01 '_
_ii_i2, --N,_3-a,. ,,0.47sO.2_vell *, 5_Jd'_.22 _"_DEL j' ,J -1.1661
0.1.
c:oaa 0
-0.1,
-0.2
-0.3
0 $ 10 lS _0
T&II
0.4
0.1
:OB, II 0
-O. 5,
00.4
P4_WD.hdep
' P4_WD.hdep_ ' O:OlamQz oO._477S$ • t.u o7
:Oa|mAm o04.$77555 • tau _0.175Ivel * L001.1.2 svoOiff - -1.)1759
i i i
S 10 IS 20
f&g
Hgurc 4: Cmss-Corrclaxions Becwe_,n Elcvazor Deflection and Altitude of the Airplane for Pilot P4
B, A, and W amethe piloCs be.st, average, and worn flights withom additional u'anspo_ delay
BD, AD, and WD am r_e pilot's bcsL a_ and work tliglzs with additional u'anspo_ delay of _0 ms.
B 15
Appendix
The complete program written in C++ is enclosed. The program reads the F-15E simulation data after the data file
has been converted from binary to ASCII. The program was compiled on IRIX 5.3 C/C++ compiler 4.0.2 on the
SGI Indigo2 workstation. Pages 15 to 17 contain the Correlation Source Code. Pages 18 and 19 have the
Correlation Header. Page 20 has the Script Source Code. The program prints out the calculated correlation values
and the jargon for plotting the graphs on a GNU Plot Program. The Script file separates the jargon from the data,
and it creates a script file for the GNU Plot Program.
B 16
#include <fsUrea:. h>
#include <s:riz_._#include <mar_h. h>
|include "corrJ..._.;_"_define i 7200
void main( )
(
c.hax fileName(80} ;
char FileName(BO] ;
char bu£fe:(8O] ;
char :exu(255] ;
char CAll0] ;c_a: CB[IO] ;
char FileDaca(80! ;
char FileLoad(80] ;
in¢ chart;
inc i. ii = O, j = O.k = O;
in: n(10! ;inn w:
in= of:
in,- po+,",r-:double [Y;
double d;double e;
double Numb:
double y_;double y1:
double xmin:
double y_J_:double max:
double ymax:
/* co¢-_elanion header "/
/* maxL:u, number or points "/
I* inpun film "I
/* oun_u= film "/
I" buffer "/
I* noces character vari_ule "/
I" name of A(n + cau) °I
/" _ of S(_) "1
/* number of c_els "/
/*¢_Ca poin¢er */
/e ¢oz_elanion Cy_e "//* maxiumcoz_ela=ion factor
/* enrer maxiumnumber of poin= */
/" _au value acmaxiumcozTela=ion °/
/" _au value ac miniumcoE_elanion */
/* dana buffer */
/. maxium ¢orEela_ion "/
/" m_J.um correla=ion "/
/" miniumx value of _raph "/
/* _iumy value of grap_ "/
/- maxiumx value of _:aph */
/_ maxium y,_lue of _aph "/
*/
CORR DA_oSCO:'r. coEr;
reRame:
¢OUC << " Film: ";
tin >> fileName, /* film of the ,¢ii */
¢ou_ << " EnCIE _.he maxi,,,- ¢ozTmla_ion factor ";
c_ >> cf; /* encer ma._.um coz':elacion facno: "/
cou_ << " EJ_er number of points *;
tin >> poin¢; /, enrer maxi,mntmber of point */
cou_ << " Please choose and co:Tale=ion (H_D_=I, PSX_D_F=2. PSI..=P,F=3}: ";cLn >> w; /" coE_ela_ion salec:ion */
/t opens _he dana file. if non file exisu _hen _he file name is asked
for a_ain -/
ifsuream fin(fileName, ioe::norreaue);
if (_fin)
(cou_ << * Unable _o open :: " << fileName << ° "** file no= found *** in\n';
noun << " Please re-enrer _he filename, in%n';
gOCO rename;}¢ou= << " in Filename • << fileS, me << • was found, is" :
/t reads -,meede¢l _a_;en Tel= by :he qe¢_ara proq_.'a., "Ifor ( i = O; i < 5: i++ )
(f_J_ >> bu£fe=:
}
fin >> chart: /* ¢e_da _ nmobe= of chanele */
couc << "in This file has • << ¢ban << " channels, in';
I" reads and soc'_s _J_e number of channels "I
B 17
re: ( i- 0: i < (chart+2): i++ )(
fAa >> buffer:if (![sc_c:_tbuffer.'T')_| ( _[0l = j;}if (!(su:c=pCbuffe=.'H'))) ( all] = j;}if (!(sc:c=p(buffe:.'?SZ')))( n[2] = _;}if (!(eC:==p(bu_fec,'OAP'))) ( a[3] = j:}if (!(eC=c=p(bu£fec,'OEP*))) ( n[4l : j;}if (i(eC_-=mp(buffer.'DRl_')}) ( n[51 = _;}
}
/" reads and sorus uhe claca co _ieE co=Tecc arTays "/while ({fin>> Numb) && (k < (point)) ]
£f ( ii == -(01if ( ii == nil]if ( ii == -[21if [ ii := n(3|if ( ii == n(4|if ( ii .m n[S]li++_if (ii=m (than+l)}
( DATA.T[k| = Numb; }( DATA.K[kl = Numb;}( DATA.PSZ[k| • Numb; }( DATA.DAP[k| = Rumb: }{ DATA.DEP[k] = Numb; }{DATA.DP-P[k] = Num_; }
(ii = 0; k++1}
I" closes C.he i_c file "lfin.close( ) ;
/* sees valuel i_ic&lize CO the coEEolaCion heedere/DATA. of•of ;DATA.www_DATA. k•k;DATA. COEE () ;
/e finctS _he maxium u;_er and loweE points of _hm ¢oEralacion "/¥I '• DATA.COR_A[O] ;5=0;yu • DATA.C0KPA(0] ;em0;for ( i • Os i < Of; i++ )(
if(yl • (DATA.COR_A[i|) |(yl • (DATA.C0_UPA[A| };dsi" (.025) ;})foc ( i • I; i < of; i++ )(
if(yu < (DATA.COR3A[i]) ) (yu ,, (DATA.COKPA[i] ] ;eeL" (.025) ; }}if(fahs(yu) < fa_s(y_] ) (U=y_;} cou, c <<U;if(fahs(yu) • fabe(yl))(U=_u-} couc <<U:
= ((DATA.TAU[cf-1]) ".2+DATA.TAU[cf-I] );
y,d.u - -(fabe(U+U'.SO) | ;ymax -(labs( U+U'.50)];
Rename:cou,_ <,c • Film: ";cin •> rileName: /- name of _he oucpuc file "/C0u_ << " EuneE one llne of co.enos for da_a file :\n';cin.i_nowe(1,'\n*); /" enter notes on r.ha _aph "/c4 n .geCline ( r.oxl:. 255 ) ;
/" sects the ohm file name "/• scz_rl (Fl_eDa_a,Flleb_me) ;
8tLT=I;_ (F_.leLoad, Fllel_hune) ;eclat (F1].el..oed,. • .gl_'} ;if (w==l.|(
s_'cac (Y_leDa_a, ".haaS" ) ;s£Ecac _CA, "/_'l ;sr.Ecac (C_,*DEI"} ;
B 18
)
(
}if(
)if(
)ifC
(w=m2 }
e_:a_ (FileDa_a, " ._dap" ) ;s_rcat (CA. "PSZ') ;s_cau (C3, "D_') ;
(w==])
s_Ecat (FileDa_a, " ._dzlP" );at,Teen (CA, •PSI" );s_E=at (C_,'_RP'} ;
(w==4)
s_ca_ (FilaOaca, " .ed=_') ;s_zrcat (CA, "D_') ;strca= (_. "DP_'} ;
(w==5)
eC_:at (FileDa_a, • .eden') ;strcat (CA, "U£P') :m_cat (C:m. "_') ;
/- opens output file °/ofs_=eam fout(FileData, ioe::noreplacs);if {!four)(
ccna_ <¢ " Unable to create and _rci_m _ocouL <c • Please re-enteE _he fil_. \n\n';
goLo R@n,ame:)
/" wE£uem scripn _az_en "/foul << "% mec title \''<cFileDa_a<<'\'\n';fourfour:foulfOuLfoulfourfou,.foulfou=fou_
• fou_fouc
• << FileJ_sme << " "'* file aLread_
<c "| met xla.]oel \'TA_\''<<'\n°:<c -% met ylabel \'COU.\' O,-10"<c'_n';<< "0 met la_m_ 1 \'CORRmax ='<<yu<¢" _ _au =.<<e<c'\' at "<< _:Lin'.9 <<',"<< "t met label 2 \'COl_min ='<<yl<<" _ _au ='<cd_<'\" at "<< x=_'.9 <<.,"<¢ "# see label 3 \'a_'<<CA<<" = .<¢DATA.aveA_<" ave'<<CB_<" = "<<_kT_.a_<< "# ee_ label 4 \'_'_S'<ccA<<" = "_<DATA.x_uA<<" r_s'_<C_¢<" = "<<CIkTA._'J<< "O p_o¢ (x='<< _An <<':'<¢ _wx:<<'| ['<< ymA_ <<-:'<_'ymax<<'| \0-<<FileData<<'\'\n';<<'l cauti] corrCi] • << "\n';<<-1 • << fileN_me << "\n';<<'| * << _e.xt << "\_';<<'# =au(i] cor_(i] • << "\n';
I- writes remulLs, to ou=;n;t file "Ifor [ t = 0; i < cf; i++ )(
fou_ << DATA.TAU(i] <<'\_" << DATA.C0RPA(i] << "\n';
)
fou_.cloee(}; /- closes ouLpuc file "couc << "\r_ilename "<< FileDa_a << " was c=eated. \n';c_ut << "\nNc,_ run ec_ip_ and use .<cFileDa_a<<" scEip_ film.\n';
B 19
-- # include <=ar._. ]_# 4,,elude < fscream. _>
edefi=e i 7200
class CORR(public :
Ant w:inc cf;Ant z:dmzble T[I|. HILl. PSZ[I|. DAP[I]. DEP[I]. DRP{I|. TAU[I]. CORPA(I|;
double C, h. psA, clap, dep. ct_p;
dingle areA. ave6;
double sqdA. s_ :
dingle coz'p, coL'n:double m. sadS;double mqA. s--,qS:double smOOCh, smooch;double A[I|;double S[1] ;double difB[l| ;double difA[l] ;_2e corpA;double co='AA tdingle cauz
COZy() ( )-COP._( ) { }void Scot= ( )(
T(z]-c, H(z]-h, PSZ(z]-ps£, DUCzl,xbm, D_(zl-de_, _(=]'_'_:)vold coc_ ( )(
/" sees all sum variables Co ze_places chert co_'=ecc signals Co be corela_ea Co A and Band sums A and B "1
• smA = O, s,,,m= O, smsqA • O, smsq6 • O, smcorn . O, sm=o:_ i O;
££ (w--_)(
£or ( i - O; £ < k; i++ )
(A(i] ,, H(i|;B(£] = DEP[i|;smA = smA + A[i| ;sin6 • s.US + B[/]:
)}£_ (w-=2)(
£oc ( £ - 0; i < M: i+_ )(
A(A] = eSZ(il;B(/] = DB4P(:LI;i = saA + &(/l ;sa8 - laa + 8[£1 ;
)}
{
B 20
};
for ( i : 0; i < k: i+÷ )(
A(£| ,, PSZ(i];B(i| : DA_Ci|;smA = areA _ A(i] ;stub = s ''a ÷ BCi];
)}£_ (w=-4 !{
for ( i = 0; i < k; i_ ){
A(tl ffi oe.,eCil;_(i] ffi v_[£1;smA = smA + a[£] ;rids = sma + 8(£];
))i£ (w==5)(
for ( £ = 0; i < k; i*-- ){
A(£] = nkPCil;sCil = DEZ'(£I;smA = m ÷ A[i] ;am8 = sin8 ÷ B[£];
))
avaA = smA/(k) ;avsB = smBl (k);for (i = 0; i < k;i*-)(
dlfB[l] =(B[i] -araB};d/fA[i| = (A[i] - areA) ;scab = powidifB(i]. 2.0);scala = pow(difA[£l. 2.0) ;samoa. = =naqA - sc_A:_S = smsqS .,. sqclB-
.}_msc_ • mqA/(k} ;ImsO • smsqSl (k) ;mS = (sqrc(mnsqS) } ;rmJA = (s_-_:(:m.sqA) } ;for (j = O; _ < cf ; _,'....){
f0= (£ = O; £ < (k-j÷L);£_-+)(
¢orp = difB[£l * cttfA(i._l;811C0_ ,= 8me'O_lP 4- ¢OZ'p;
}cau = j * 0.025:
I* average of A */I" average Of B "I
/* difference between B(:L] and aveB "//* difference bem,'ean S[£] and. ayes "I
/* Ems value of A "//* Z'ms value of B "/
¢o_A = smco_-_/(Ck - j) " rms8 " _A}; /* ¢orralacion cooficionc *Ism"Oz_ : 0;TAU[_] = cau:C0RPA(_] : co_A; /" correlation coe£icien: a_ay "/
B21
OtacJ.uda <fa_._eaa.h>ll-clude <s_zi_g. b._void mean( )(
char file_ame (80 ],into i.-:
II enter filenamo
cia >_. file_ame:COU_ << " \n Filen_une " << file_ame << " _ found. \n" ;ifsr.:eam f_[fila_amo, io8: :not=ear'.e) ;
s_rl[l| = "\0":n,, O;wb_le (file_a,-e(n] != '.'}(s_rl[O] = file_an,e{n] :smrca_ ( fileLoad, scrA} ; n++:}
smrcam (££1eLoad. ".;If" ];couc << " \n Filename ' << fileLoad << • was creamed. \n\n" ;
ceu_ << • Now run gnuplo_ and load \' "<< fileLoad<<'\" \n';ofs _ream fou= ( fileLaad. £=s: :no=e_Lace ) ;
fOE (i=0 pi<8 ;i_-}{fln.ge_(buf£er.2] ; fin.ge_llne(WRITE,2SS] ;fou_ << WRITE<<'\n';
}
huffer[255], WRITZ_255]=' ",_ileLoad[80|,sCrl[80|;
B 22