flight control system design and test for unmanned rotorcraft · flight control system design and...
TRANSCRIPT
Flight Control System Design and Test forUnmanned Rotorcraft
Flight Control and Cockpit Integration BranchArmy/NASA Rotorcraft Division
NASA Ames Research CenterMoffett Field, CA
Chad R. Frost
IEEE Santa Clara Valley
Control Systems Society Technical Meeting
September 21, 2000
UAV Control Design
Overview
• Background• Design Tools• Design Methods• UAV programs• Example
UAV Control Design
Background• A UAV is an uninhabited, reusable aircraft that is
controlled:– Remotely,– Autonomously by pre-programmed on-board equipment,
– Or a combination of both methods
• Currently >241 UAV systems developed by 31 countriesare operational or in test
• Numerous missions, current and proposed:– Military
– Civilian– Space
Source: Bob Keith, NATO UAV C2 Workshop 1999 (Unclassified).
UAV Control Design
Background
UAV Control Design
Background
• Many vehicle configurations, but rotary-wingedvehicles form a significant and growing portion
• Hover-capable UAVs offer unique capabilities, butcome with unique challenges
UAV Control Design
Background
• Significant industrial and military expertise exists in fixed-wing UAV development.
• Initial work on rotary-wing UAVs did not exploit thecapabilities of the configuration:– Lack of familiarity with rotorcraft issues
– Inability to foresee problem areas
• NASA involvement in rotorcraft UAV developmentsought to take performance to a new level.
UAV Control Design
Background
• Ames is NASA rotorcraft center:– Army / NASA Rotorcraft Division
• NASA: Aerospace Directorate
• Army: Aviation & Missile RD&E Center
– Flight Control and Cockpit Integration Branch
• Expertise in rotorcraft:– Flight control– Modeling– Simulation
• Design tools developed in-house
UAV Control Design
Design Tools
• CIFER®
– Comprehensive Identification from FrequencyResponses
• CONDUIT– CONtrol Designer’s Unified InTerface
• RIPTIDE– Real-time Interactive Prototype Technology
Integration/Development Environment
UAV Control Design
Design Tools
• CIFER®
– Extraction of mathematicaldescription of vehicle dynamicsfrom test data
– “Inverse” of simulation
– Robust software, widely used inaerospace industry
UAV Control Design
Design Tools
• CONDUIT– Evaluation and analysis of any
modeled system• Linear model from CIFER• Non-linear simulation code
– Control system design• Simulink or SystemBuild block-
diagram modeling
– Control system optimization• User-selected specifications
• Multi-variable, multi-objective FSQP
UAV Control Design
UAV Control Design
Design Tools
• RIPTIDE– Real-time simulation
environment– Can use models from
CIFER, CONDUIT, orstand-alone code
– Can use control systemdesigns fromCONDUIT
– Hardware-in-the-loopcapability
UAV Control Design
Design Tools
RIPTIDESHAREDMEMORY
A/C Model
Controller
Auto Code
Cockpit Displays
Out-the-window Displays
Pilot Inceptors
MatlabSIMULINK
MathModel
Development
CockpitDisplay
Development
CONDUIT
Elements of RIPTIDE real-time simulation environment
UAV Control Design
Design Methods
Design
Simulation
Flight Test
Development
Specifications
Flight VehicleDevelopment Cycle
UAV Control Design
Design Methods
• Typical sequence of control system development:– Collect data from vehicle
– Extract linear math model using CIFER
– Design control system using CONDUIT
– Optimize control system gains using CONDUIT
– Shakedown tests in RIPTIDE
– Fly control system on vehicle• If modeling done correctly, vehicle response should match
model predictions.
UAV Control Design
UAV Programs
• VTUAV• LADF• R-50/R-MAX• BURRO
Ames participation in:
UAV Control Design
Example: BURRO
• USMC demonstration program• Broad-area
Unmanned Responsive Re-supply Operations
• UAV to pick up loads frommoving ship, deliverautonomously to inland troops
UAV Control Design
Introduction
• Kaman Aerospace K-MAX– In production– Designed for load-lifting– 6,000 lb vehicle– 6,000 lb slung load capacity– Synchropter configuration– Servo-flap rotor
UAV Control Design
Introduction
• Army/NASA CRDA with Kaman to supportFCS development
• Three integrated tools– CIFER® : System identification– CONDUIT: Control system modeling, analysis,
optimization– RIPTIDE: Desktop real-time simulation
UAV Control Design
IntroductionDesign
Simulation
Flight Test
Analysis
ManualLittle or no optimizationStop when "good enough"
Not real timeNot "pilot in the loop"
ManualSubjectiveGuesswork
SlowExpensiveRisky Design
Simulation
Flight Test
Analysis
CONDUITMulti-objective optimization
CONDUIT (NRT)RIPTIDE (RT, piloted)
CIFERAutomatedAccurateObjective
Fewer hours requiredHigher initial confidence
Traditional
State-of-the-Art
UAV Control Design
Introduction
• Scope:– Hover / low-speed– Unloaded– Ground operator control
UAV Control Design
Aircraft Modeling
• Start with pilotedfrequency sweeps ofunaugmented K-MAX
• 8-DOF (rigid-body + 2rotor states) linearstate-space modelidentified from flightdata using CIFER®
Real Axis
Imag
Axi
s
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Unstable roll
Unstable pitch
Pitch
Lateral-directional
Heave
Eigenvalues, hover
UAV Control Design
Aircraft Modeling
• Verified in time domainusing CIFER®
Ref: Jason Colbourne, et al., American HelicopterSociety Forum, May 2000.
-10
1
Lateral Control Deflection
Flight data
Math model
-20
020
Roll Rate Response
-20
020
Pitch Rate Response
Time (Sec)0 1 2 3 4 5 6 7
UAV Control Design
Aircraft Modeling
• Sensor dynamics– Equivalent delays estimated
from manufacturer specs (25ms)– 2nd-order Padé approximations
• Actuator dynamics– Identified from bench-test
frequency sweeps– 2nd-order systems
(ω=20 r/s, ζ=.5)
– Rate- and position-limiting
UAV Control Design
Aircraft Modeling
• Aircraft, actuator and sensor modelsimplemented in Simulink block-diagram
1
Aircraft response
Sensor models
Linearized 8DOFAircraft dynamics
Flight controllaws Actuator models
1
Control input
UAV Control Design
Control Law Development
• Inner Loops– Attitude Command / Attitude Hold
• PID controller
– Heading Command• PD controller
– Altitude Rate Command• PD controller
• Outer Loops– Translational Rate Command
• PI controller (or position feedback)
• Modeled in Simulink
3
Attitude Rate
2
Attitude
1
TranslationalRate
Sensor models
PIDAttitude
controller
PITranslational
Ratecontroller
Actuator input
Velocity
Attitude
Attitude rate
Linearized 8DOFAircraft dynamics
Actuator model
1
Control input
UAV Control Design
Control Law Development
AttitudeCommand
Mode
VelocityCommand
Mode
1
Lateral_Servo_Command [stu]
.05
trim sensitivity (right)
.05
trim sensitivity (left)
1
stick sensitivityft/sec per stu
1
stick sensitivitydeg per stu
trigger
in out
discrete latchHolds input value
when FCC_State = Active (1)(Velocity hold)
trigger
in out
discrete latchHolds input value
when FCC_State = Active (1)(Attitude hold)
1/57.3
deg > rad
dpp_aKp
Proportional gainstu/rad
dpp_aKp_v
Proportional gainrad/(ft/sec)
dpp_aKi_v
Integrator gain
dpp_aKi
Integrator gain
s
1
Integratorlimited windup
.523 rad(30 deg)
s
1
Integratorlimited windup
(50%)
dpp_aTp
Feedback time constantrad/(rad/sec)
m
FCC_State
Trim +
Trim -
Direct
Total Control
Control Mixer
Attitude limit+/ .523 rad
(30 deg)
6
Command_Mode [ 1= Velocity 2 = Attitude ]
5
sensors
4
Trim_Right
3
Trim_Left
2
Stick Input[stu]
1
FCC_State
Roll angle
Vd
Roll rate
• Complete Simulink model:– 22 inputs– 32 outputs– 331 states (continuous and discrete)
– 27 tunable gains (“design parameters”)
• Includes nonlinearelements:
– Limited integrators
– Authority limits
– Mode switching
Lateral controller shown
UAV Control Design
Control Law Development• CONDUIT Optimization Engine:
– Multi-objective optimization using FSQP
– Adjusts design parameters (system gains)to meet requirements of specifications
– Specifications represented graphically, 3regions based on level of performance
• Categorizes specifications:– Hard
• must be met
– Soft• should be met, without violating Hard
specs
– Objectives• minimized after all specs satisfied
GM [db] P
M [d
eg]
Gain/Phase Margins
0 5 10 15 200
20
40
60
80
Level 1
Level 2
Level 3
UAV Control Design
Control Law Development
• Specification selection– Stability
– Performance and “Handling Qualities”– Objectives
• Rationale– Airframe originally designed as a manned vehicle– Safety pilot on board demonstrator vehicle
– Ground operator control will be VFR / simple tasks
UAV Control Design
Control Law Development
• Stability Specifications (Hard constraints)
GM [db]
PM
[deg
]
(rigid-body freq. range)Gain/Phase Margins
0 5 10 15 200
20
40
60
80
Real Axis
Eigenvalue Location
-1 -0.5 0 0.5 1
UAV Control Design
Control Law Development
• Performance Specifications (Soft constraints)
d_theta [deg]
Change in 1 Second [deg]Pitch Attitude
0 1 2 3 4Time (sec)
Normalized Attitude Hold(Disturbance Rejection)
0 5 10 15 20
-1
-0.5
0
0.5
1
Actuator Rate Saturation
Act
uato
r P
ositi
on S
atur
atio
n
Actuator Saturation
0 0.5 10
0.2
0.4
0.6
0.8
1
UAV Control Design
Control Law Development
• Handling Qualities Specifications (Soft constraints)
Eq.
Ris
e Ti
me
(Tx-
dot, T
y-do
t) [s
ec] Translational Rate Rise Time
1
2
3
4
5
6
7
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Other MTEs;UCE=1;Fully AttBandwidth & Time Delay (roll)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Dam
ping
Rat
io (
Zet
a)
Attitude HoldDamping Ratio
0
0.2
0.4
0.6
0.8
1
heave mode, invThdot, [rad/sec]
time
dela
y, ta
u_hd
ot, s
ec
Hover/LowSpeedHeave Response
0 0.5 10
0.1
0.2
0.3
0.4
0.5
UAV Control Design
Control Law Development
• Objective Specifications– Spec selection reduces actuator sizing,
component fatigue, and noise sensitivity
Actuator RMS
Actuator RMS
0 0.5 1
Crossover Frequency [rad/sec]
(linear scale)Crossover Freq.
0 5 10 15 20
UAV Control Design
Control Law Development
• Control system gains tunedusing CONDUIT– Initial tuning:
• 27 parameters• 33 specifications
– All specifications satisfied(Level 1)
– RMS actuator position andcrossover frequencyminimized
Real Axis
Eigenvalues (All)EigLcG1:
-1 -0.5 0 0.5 1-0.1
-0.05
0
0.05
0.1
GM [db]
PM
[deg
]
(rigid-body freq. range)StbMgG1: Gain/Phase Margins
H
0 5 10 15 200
20
40
60
80
GM [db]
PM
[deg
]
(rigid-body freq. range)StbMgG1: Gain/Phase Margins
0 5 10 15 200
20
40
60
80
GM [db]
PM
[deg
]
(rigid-body freq. range)StbMgG1: Gain/Phase Margins
0 5 10 15 200
20
40
60
80
GM [db]
PM
[deg
]
(rigid-body freq. range)StbMgG1: Gain/Phase Margins
0 5 10 15 200
20
40
60
80
Crossover Frequency [rad/sec]
(linear scale)CrsLnG1:Crossover Freq.
0 5 10 15 201
1.2
1.4
1.6
1.8
2
heave mode, invThdot, [rad/sec]
time
dela
y, ta
u_hd
ot, s
ec
shipboard landingFrqHeH2:Heave response
0 0.5 10
0.1
0.2
0.3
0.4
0.5
Actuator Rate Saturation
Act
uato
r P
ositi
on S
atur
atio
n
Actuator SaturationSatAcG1:
0 0.5 10
0.2
0.4
0.6
0.8
1
Actuator Rate Saturation
Act
uato
r P
ositi
on S
atur
atio
n
Actuator SaturationSatAcG1:
0 0.5 10
0.2
0.4
0.6
0.8
1
COL
LON (ACAH) LAT (ACAH)
PED
LON
COL, LON LAT, PED
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Other MTEs;UCE=1;Fully AttBnwPiH2:BW & T.D. (pitch)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Other MTEs;UCE=1;Fully AttBnwRoH2:BW & T.D. (roll)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Other MTEs (Yaw)BnwYaH2:BW & T.D.
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Actuator RMS
RmsAcG1:Actuator RMS
0 0.5 10
0.2
0.4
0.6
0.8
1
Time (sec)
HldNmH1:Normalized Attitude Hold
0 5 10 15 20
-1
-0.5
0
0.5
1
Dam
ping
Rat
io (
Zet
a)Attitude Hold
OvsPiH1:Damping Ratio
0 0.5 1
0
0.2
0.4
0.6
0.8
1
d_theta [deg]
Change in 1 secRisPiV1:Pitch Attitude
0 1 2 3 40
0.2
0.4
0.6
0.8
1
d_phi [deg]
Change in 1 Second [deg]RisRoV1:Roll Attitude
0 2 4 6 80
0.2
0.4
0.6
0.8
1
d_psi [deg]
Change in 1 Second [deg]RisYaV1:Yaw Attitude
0 1 2 3 4 50
0.2
0.4
0.6
0.8
1
LON
PITCHROLL
YAW
PITCH
ROLL YAW
COL
LAT
PED
COL
LAT
PED
UAV Control Design
Control Law Development
• Conditionally stable lat & lon
• Model predicts stable, well-damped responses
0 6
1Roll response
Time (sec)
Atti
tude
(de
g)
-40
-20
0
20
Gai
n (d
B)
Lateral Stability Margins (Initial):PM = 46.8 deg. (ωc = 3.75 rad/sec)
GM = 9.7 dB, (ω180 = 11.23 rad/sec)
1 10-400
-300
-200
-100
0
Frequency (rad/sec)
Pha
se (
deg)
UAV Control Design
Flight Test
• First flight test with CONDUIT-tuned gains• Aircraft responses did not agree with model
(lon and lat)
0 5 10 15 20 25 30 35 40 45-40
-30
-20
-10
0
10
20
30
40
Time (sec)
Rol
l Atti
tude
(de
gree
s)
Roll attitude command
Roll attitude
UAV Control Design
Flight Test
• Looking for source of discrepancy:– Lon and lat doublets flown closed-loop– CIFER® used to extract frequency responses– Actual sensor and actuator dynamics identified
• Equivalent time delay greater than originallyestimated
UAV Control Design
Flight Test
ComponentEstimated
Delay (ms)Actual Delay
( m s )
Actuators 5 0 107Sensors 2 5 5 3
Computer 2 0 6 0Filters 0 7 0
TOTAL 9 5 2 9 0
UAV Control Design
Flight Test
• Updated Simulink model with identified delays• Added delay results in highly constrained
system
1 10-400
-300
-200
-100
0
Frequency (rad/sec)
Pha
se (
degr
ees)
K-MAX BURRO
XV-15 Tilt Rotor
Narrowed range of stability
UAV Control Design
Flight Test
• Added lead filter to lon& lat attitude feedback
• FCS gains re-tunedwith CONDUIT
• CONDUIT successfullytraded off phasemargin for gain margin
-60
-40
-20
0
20
Gai
n (d
B)
1 10-400
-300
-200
-100
0
Frequency (rad/sec)
Pha
se (
degr
ees)
acceptable range of crossover frequency
CONDUIT-tuned result
UAV Control Design
Flight Test
• CONDUIT tuning results
GM [db]
PM
[deg
]
(rigid-body freq. range)Gain/Phase Margins
0 5 10 15 200
20
40
60
80
GM [db]
PM
[deg
]
(rigid-body freq. range)Gain/Phase Margins
0 5 10 15 200
20
40
60
80
Actuator Rate Saturation
Act
uato
r P
ositi
on S
atur
atio
n
Actuator Saturation
0 0.5 10
0.2
0.4
0.6
0.8
1
Actuator Rate Saturation
Act
uato
r P
ositi
on S
atur
atio
n
Actuator Saturation
0 0.5 10
0.2
0.4
0.6
0.8
1
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Bandwidth & Time Delay (pitch)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Bandwidth & Time Delay (roll)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
LON (ACAH) LAT (ACAH)
LON LAT
PITCH ROLL
After CONDUIT tuning
Baseline gains
UAV Control Design
Flight Test
CONDUIT results:
– Level 2 (8 specs)
– Reduced bandwidth
Real Axis
Eigenvalue Location
1 0.5 0 0.5 1GM [db]
PM
[deg
]
(rigid-body freq. range)Gain/Phase Margins
0 5 10 15 200
20
40
60
80
GM [db]
PM
[deg
]
(rigid-body freq. range)Gain/Phase Margins
0 5 10 15 200
20
40
60
80
GM [db]
PM
[deg
]
(rigid-body freq. range)Gain/Phase Margins
0 5 10 15 200
20
40
60
80
GM [db]
PM
[deg
]
(rigid-body freq. range)Gain/Phase Margins
0 5 10 15 200
20
40
60
80
Crossover Frequency [rad/sec]
(linear scale)Crossover Freq.
0 5 10 15 20
Actuator Rate Saturation
Act
uato
r P
ositi
on S
atur
atio
n
Actuator Saturation
0 0.5 10
0.2
0.4
0.6
0.8
1
Actuator Rate Saturation
Act
uato
r P
ositi
on S
atur
atio
n
Actuator Saturation
0 0.5 10
0.2
0.4
0.6
0.8
1
Bandwidth [rad/sec] P
hase
del
ay [s
ec]
Other MTEs;UCE=1;Fully AttBandwidth & Time Delay (pitch)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Other MTEs;UCE=1;Fully AttBandwidth & Time Delay (roll)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Bandwidth [rad/sec]
Pha
se d
elay
[sec
]
Other MTEsBandwidth & Time Delay (yaw)
0 1 2 3 4 50
0.1
0.2
0.3
0.4
Actuator RMS
Actuator RMS
0 0.5 1
Time (sec)
Normalized Attitude Hold(Disturbance Rejection)
0 5 10 15 20
-1
-0.5
0
0.5
1
Dam
ping
Rat
io (
Zet
a)
Attitude HoldDamping Ratio
0
0.2
0.4
0.6
0.8
1
d_theta [deg]
Change in 1 Second [deg]Pitch Attitude
0 1 2 3 4d_phi [deg]
Change in 1 Second [deg]Roll Attitude
0 2 4 6 8d_psi [deg]
Change in 1 Second [deg]Yaw Attitude
0 1 2 3 4 5
Eq.
Ris
e Ti
me
(Tx-
dot, T
y-do
t) [s
ec] Translational Rate Rise Time
1
2
3
4
5
6
7
heave mode, invThdot, [rad/sec]
time
dela
y, ta
u_hd
ot, s
ec
Hover/LowSpeedHeave Response
0 0.5 10
0.1
0.2
0.3
0.4
0.5
COL
LON (ACAH) LAT (ACAH)
PEDLON LAT
PITCH
ROLL
LAT
LON COL
LON
LAT
PED
COL
LON
LAT
PED
PITCH,ROLL
YAW
PITCH ROLL YAW
LON (TRC)LAT (TRC)
UAV Control Design
Flight Test
• Roll response muchimproved; modelresponses agree wellwith flight results
0 5 10 15 20 25 30 35 40 45-60
-40
-20
0
20
40
60
Time (sec)R
oll A
ttitu
de (
degr
ess)
Roll attitude command
Roll attitude
UAV Control Design
Flight Test• BURRO successfully demonstrated to USMC nine
months after start of development
UAV Control Design
Conclusions
•• Design space is very limitedDesign space is very limited– Aircraft dynamics
– Control system hardware– CONDUIT was able to extract the best achievable
performance within design limitations
•• High frequency dynamics were key driver of closed-High frequency dynamics were key driver of closed-loop performanceloop performance– CIFER was useful in identifying system elements
•• Advanced design tools allowed rapid developmentAdvanced design tools allowed rapid developmentof a successful UAVof a successful UAV– 9 month time span– Recovery from added delay
UAV Control Design
Current and Future Work
• Build 1 of K-MAX BURRO UAVsuccessfully demonstrated toUSMC
• Build 2 now in development
• 2000-lb loaded hover– 10-DOF EOM and CIFER ident
complete– FCS design complete
• 5000-lb case in development
• Envelope expansion to 70 KTASin progress
UAV Control Design
Questions?
For additional information:
http://caffeine.arc.nasa.gov
http://uavinfo.homepage.com