Académique Documents
Professionnel Documents
Culture Documents
RonaldChoe,EnricXargay,andNairaHovakimyan
ld Ch
i X
d i H ki
CoordinatedScienceLaboratory,UniversityofIllinoisatUrbanaChampaign
email:{choe19,xargay,nhovakim}@illinois.edu
http://naira.mechse.illinois.edu/
AerospaceDecisionandControlWorkshop::GeorgiaTech::11June2012
Outline
Aircraft LossofControl
9 Characterization
9 Prevention
Previous Research: AirSTAR Project
j & L1 Adaptive
p
Control
9 Main challenges
9 Flight test results
iReCoVeR: LOC Prevention through Adaptive Reconfiguration
9 Limitations of the current technology
9 iReCoVeR architecture
9 Fullscale aircraft validation: U of Is Beckman Institute
Concluding Remarks
AircraftLossofControl
AmericanAirlines587
BelleHarbor,NY
Wakevortices&piloterror
p
AmericanEagle4148
Roselawn IN
Roselawn,IN
Asymmetriciceaccretion
Inconsistentairspeedreadings
&spatialdisorientation
LOCeventscanresultfromawiderangeofcausalandcontributingfactors
g
g
thatoccurindividually orincombinationsequentiallyintime
LossofControlCharacterization
Quantitative LossofControl Criteria:
Airplane flight dynamics, aerodynamics, structural integrity, and flight
control use.
LOC=Excursion
outside 3 envelopes
outside3envelopes
Wilborn&Foster2004
LossofControlPrevention
Integrated architecture for LOC prevention:
*AIRSAFEconcept
AirSTAR &L1AdaptiveControl:
FlightTestEvaluations
Flight
Test Evaluations
andResearchModelingTasks
NASAAirSTAR >M
Predictable::Repeatable::Testable::Safe
Controllawobjectives:
ll
b
Provide predictable aircraft response to help
the pilot avoid excursions outside the wind
tunnel data envelope in the presence of
aircraft impairment.
IsA/Ccontrollable
here?
Safe
envelope
Source: NASA
Highriskflightconditions,someunabletobe
testedintargetapplicationenvironment.
5
5.5%geometricallyanddynamicallyscaledmodel
5 % geometrically and dynamically scaled model
82inwingspan,96inlength,49.6lbs(54lbsfull),53mphstallspeed
Modelangularresponseis4.26faster thanfullscale
Modelvelocityis4.26timesslower thanregularscale
NASAAirSTAR ::Challenges
Innerloopstatefeedbackcontrollerfortrackingangleofattack,rollrate,and
sideslipanglecommands
Challenges:
SingleCASdesignfortheentireflightenvelope(includingstallandpoststallhigh conditions)
Compensationforstructuraldamage/actuatorfailureswithoutFDImethods
Compensationforunmatcheduncertainties variationsin,,V dynamicswithflightcondition
Strictperformancerequirements:
S i
f
i
Highprecisiontracking
Reducedworkload
Predictableresponse!!!
P di t bl
!!!
Hardwarerequirements:
Eulerintegrationat600Hz
L1
AFCS
FlightControlLawEvaluationMatrix
EvaluationTask
1st straightleg
2nd straightleg
Turns
Scope
Latencyinjection
(5msec/5sec)
FaultEngaged
RollDoublet
FaultEngaged
PitchDoublet
FaultEngaged
NominalStability
(Cm&Clp )0%
FaultEngaged
RollDoublet
FaultEngaged
PitchDoublet
N/A
NominalStability
(Cm&Clp )50%
FaultEngaged
RollDoublet
FaultEngaged
PitchDoublet
DisengageFault
RobustStability
(Cm&Clp )75%
FaultEngaged
RollDoublet
FaultEngaged
PitchDoublet
DisengageFault
RobustStability
(Cm&Clp )100%
(neutrallystable)
FaultEngaged
RollDoublet
FaultEngaged
PitchDoublet
Disengage Fault
DisengageFault
Robust Stability
RobustStability
(Cm&Clp )125%
(unstable)
FaultEngaged
RollDoublet
FaultEngaged
PitchDoublet
DisengageFault
RobustStability
Poststall tracking
NoFault
NoDoublet
NoFault
NoDoublet
N/A
RobustPerformance
EvaluationTask
Downwind/straightleg
Upwind/straightleg
Turns
Scope
Offsettolanding
(nominal)
Achieve
goodtrim
Nofault
1st:Practicelanding
2nd:Evaluationlanding
: Evaluation landing
N/A
NominalPerformance
Offsettolanding
(Cm&Clp )100%
(neutrally stable)
Achieve
goodtrim
FaultEngaged
Evaluationlanding
DisengageFault
RobustPerformance
Offsettolanding
(C
(Cm&Clp
& Cl )125%
) 125%
(unstable)
Achieve
goodtrim
FaultEngaged
g g
Evaluationlanding
Di
DisengageFault
F lt
R b tP f
RobustPerformance
L1SupportTasksonModeling
ResearchTask
Airdatavane
calibration
Subtask
vanecalibration
vanecalibration
1st straightleg
Variable
Variable
Constant
Flatturn
2nd straightleg
Deployment
Flights
Repeat
SSep2010
20 0
May 2011
28,56
Repeat
Sep2010
May2011
29,31,56
Regainaltitude
Sep2010
31,35,52
Multistep
Unsteady
aerodynamic
modeling work
modelingwork
Explorationof
departureprone
edges
Poststall tracking
Schroedersweep
Multisine
Rollforcedoscillations
Rollwavetrain
Regainaltitude
May2011
49,50,53,
56,57
sweepfromlow
angles,throughstall,to
departure
Controlsurface
wavetrains
Regainaltitude
May2011
54,55,58
Xargay, Hovakimyan,
Xargay
Hovakimyan Dobrokhodov,
Dobrokhodov Kaminer,
Kaminer Cao & Gregory,
Gregory L1
L1 adaptive control in flight,
flight in
Intelligent and Autonomous Aerospace Systems, ser. Progress in Aeronautics and Astronautics Series,
J. Valasek, Ed. American Institute of Aeronautics and Astronautics, 2012, to appear.
UnsteadyAero::HighAOATracking(September2010)
Modeling unsteady aerodynamics by emulating the dynamic motion in the
wind tunnel determining efficacy of GTM to be a flying wind tunnel
Target
a get AOA
O = 18
8 deg post
poststall
sta
Injected inputs for L1 FCL to track Step, Schroeder, Sinusoids
Step Input
StepInput
Schroeder Input
SchroederInput
22
cmd
18
20
16
14
12
10
22
20
cmd
20
cmd
18
18
22
Sinusoids Input
SinusoidsInput
16
14
12
10
16
14
12
10
4
650
655
660
665
670
Time, sec
675
4
680 740
750
760
Time, sec
770
4
780 800
810
820
Time, sec
Awellcontrollableaircraftduringstallandpoststallflight
DanMurri
AirSTAR GTMT2researchpilot
830
840
FlightTestEvaluation&ModelingSupportSummary
Adaptive FCS that provides nominal aircraft performance and takes care of large
changes in aircraft dynamics;
Enabled
bl d operation near stallll and
d departure
d
f longer
for
l
periods
i d off time,
i
allowing
ll
f
for
data collection for a wide range of flight conditions.
L1adaptivecontrollawprovides:
L1
adaptive control law provides:
tighteracquisitionoftargetflightconditions
precisiontrackingcapabilityacrosstheflightenvelope
gracefulperformancedegradation
graceful performance degradation
targetflightconditionsarebeyondachievablevalues
controlsurfacesarepersistentlysaturated
+15flights
15 fli h
Source:
NASA LaRC (internal review)
OngoingCollaborationsinEurope
CessnaCitationII
Businessjetaircraft
Business
jet aircraft
TUD,TheNetherlands
DA 42
DA42
Twinseat,propellerdrivenaircraft
TUM,Germany
iReCoVeR:
Loss of ControlPrevention
LossofControl
Prevention
throughAdaptiveReconfiguration
LimitationsoftheCurrentL1Technology
Stability might be compromised in the case of extremely severe failures:
For example, roll instability for 125% Cm/Clp degradation
Rollaxis p_cmd doublet
100
125%
15
10
5
0
-5
-10
-15
cmd
80
p
pcmd
60
40
20
0
-20
20
-40
-60
-80
740
750
Time, sec
760
770
125%
740
750
Time, sec
760
770
Controllawreconfigurationisneededforpredictableaircraftbehavior
Control
law reconfiguration is needed for predictable aircraft behavior
withoutadversepilotaircraftcoupling
IntegratedSystemsforLOCPrevention
Control reconfiguration (active L1 controller):
Online readjustment of controller gains;
Use of unconventional control configurations;
g
;
iReCoVeR
Pilot error:
Development of quantitative and computational metrics
and models;
Environmental and task conditions leading to error;
Situation awareness:
AlexKirlik
UIUC
Lui Sha
UIUC
iReCoVeR
Control
Reconfiguration
ResilientFlightController
Highperformance flight control law:
Shortterm aircraft stabilization;
Improved
p
maneuverabilityy margins
g at challenging
g g flight
g conditions or in the
event of moderate faults and failures;
Consists of:
Baseline controller: designed for a nominal aircraft model with the purpose
of optimizing system performance;
L1 adaptive augmentation: provides improved resilience by compensating
for the undesirable effects of system uncertainty;
FlightEnvelopeProtection
Ensures that the aircraft stays within its safe operational envelope by
overriding, limiting , or shaping commands generated by the pilot;
Minimi
Minimizes
es adverse
ad erse pilotaircraft
pilot aircraft coupling
co pling and
d stops
t
th pilot
the
il t from
f
making control inputs that would put the aircraft in a hazardous state;
Approaches:
L1MPC;
L1 adaptive control with input and/or output constraints;
Wilborn&Foster2004
FaultDetectionandIsolation
Responsible for detecting and isolating adverse conditions:
Sensor failures;
Structural damage;
g ;
Ice accretion (symmetric and asymmetric);
Approach:
MichaelBragg
Michael
Bragg
UIUC
Challenges:
Choice
h
off the
h number
b and nature off offnominal
ff
l models;
l
Quantification of the runtime required for the FDI module to converge;
Selection of triggering signals for the FDI module.
FlightEnvelopeDetermination
Responsible for determining an estimate of the operational envelope of
the possibly impaired aircraft:
Relies on a library of pre
precomputed
computed safe flight envelopes;
Estimate is obtained by simple interpolation;
Requires FDI to reliably isolate the adverse condition.
Computationally
Computationally
efficient
Vehicle
impairment
Safeenvelope
Updated
safeenvelope
FlightEnvelopeProtectionandLOCPredictionneedtobeupdated!
Pilotneedstobeinformed!(Situationawareness)
VerificationandValidation
University of Illinois Beckman Institute:
Fullscale aircraft humanintheloop experiments;
Large
g outthewindow display
p y screens;;
XPlane 9.40:
9
9
9
9
9
9
9
Stateoftheart
S
f h
eyetracking
ki technology;
h l
Information displays reconfigured via software.
FlightSimulator
BeckmanInstitute
Eyetrackingtechnology
Beckman Institute
BeckmanInstitute
ConcludingRemarks
Integrated system for LOC prevention:
Resilient controller, FDI module, flight envelope determination and
protection systems;
p
y
;
Reconfigurable control for improved stability and maneuverability margins;
The system constantly predicts the current flight envelope and imminent
LOC developments, and prevents and recovers from upset conditions.
L1 Adaptive Control:
Predictable aircraft behavior under significant system uncertainty as well
as limited control authority;
Graceful performance degradation under increasingly severe adversity.
adversity
Situation awareness:
Avoid adverse pilot
pilotaircraft
aircraft coupling by providing precise and timely
situation awareness to pilots during automated compensation of faults;
Should not lead to an increase in pilots workload.
Thankyou!
{choe19,xargay,hovakim}@illinois.edu
{choe19
xargay hovakim}@illinois edu
http://naira.mechse.illinois.edu/