June 2002 ENJ - Georgia Tech 1 UAV Research at Georgia Tech Eric N. Johnson Eric N. Johnson Lockheed Martin Assistant Professor of Lockheed Martin Assistant Professor of Avionics Integration, Avionics Integration, Georgia Tech School of Aerospace Engineering Georgia Tech School of Aerospace Engineering Presentation at TU Delft Presentation at TU Delft June 3, 2002 June 3, 2002
47
Embed
UAV Research at Georgia Tech - Eric N. JohnsonUAV Research at Georgia Tech Eric N. Johnson Lockheed Martin Assistant Professor of Avionics Integration, Georgia Tech School of Aerospace
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
June 2002 ENJ - Georgia Tech 1
UAV Research at Georgia TechEric N. JohnsonEric N. Johnson
Lockheed Martin Assistant Professor of Lockheed Martin Assistant Professor of Avionics Integration, Avionics Integration,
Georgia Tech School of Aerospace EngineeringGeorgia Tech School of Aerospace Engineering
Presentation at TU DelftPresentation at TU Delft
June 3, 2002June 3, 2002
June 2002 ENJ - Georgia Tech 2
Outline
•• Previous WorkPrevious Work–– MIT and Draper LaboratoryMIT and Draper Laboratory–– Ph.D. Thesis Work: Advanced Control for the XPh.D. Thesis Work: Advanced Control for the X--3333
•• Current ResearchCurrent Research–– Adaptive Guidance and Control for Hypersonic VehiclesAdaptive Guidance and Control for Hypersonic Vehicles–– Aggressive Maneuvering for UAVsAggressive Maneuvering for UAVs–– DARPA Software Enabled Control, and the GTMax UAVDARPA Software Enabled Control, and the GTMax UAV–– Aerial Robotics CompetitionAerial Robotics Competition
June 2002 ENJ - Georgia Tech 3
Draper Small Autonomous Air Vehicle (DSAAV) in 1996
IMUSD MotionPak
32cc Engine
D-GPSNovAtel RT-20
Sonar Altimeter
Camera/Tx
Compass
486 ComputerRF Modem
Receiver/Servo Interface
Power Distribution
Battery
6 ft Rotor
Modified TSK BlackStarTotal Weight 23 Pounds, 10 kg
June 2002 ENJ - Georgia Tech 4
DSAAV at the 1996 Aerial Robotics Competition
•• Organized by the Association for Unmanned Vehicle Organized by the Association for Unmanned Vehicle Systems, International (AUVSI)Systems, International (AUVSI)
•• Epcot Center, Orlando, FloridaEpcot Center, Orlando, Florida
•• Research Project Sponsored by NASA MSFCResearch Project Sponsored by NASA MSFC–– Thesis Advisor: Anthony J. Calise, Georgia TechThesis Advisor: Anthony J. Calise, Georgia Tech
•• Exploring Flight Control Technologies Applicable to Exploring Flight Control Technologies Applicable to XX--33 & Future Reusable Launch Vehicles (RLV)33 & Future Reusable Launch Vehicles (RLV)–– Reduce Analysis Required per MissionReduce Analysis Required per Mission–– Increase Tolerance to Failures and EnvironmentIncrease Tolerance to Failures and Environment
•• Feedforward Neural Networks Feedforward Neural Networks with a Single Hidden Layer are with a Single Hidden Layer are Universal Approximators.Universal Approximators.
•• The Sigmoidal Activation The Sigmoidal Activation Function has Internal Function has Internal Activation Potential ‘a’.Activation Potential ‘a’.
Single Hidden Layer Neural Network
( )σ ⋅
( )σ ⋅
( )σ ⋅
( )σ ⋅
o
o
x1
x2
xN1
o
y1
y2
yN3
V W
N2
N1 N3
June 2002 ENJ - Georgia Tech 9
( )[ ][ ] V
W
V||'WζV
W||ζxV'W
Γ+−=
+−Γ−=
ζσxζσσ
T
T
κκ
D
DAdaptation Law:Adaptation Law:
Define:Define:
0Q,QPAPA
PζT >−=+
= beTrm
Error Dynamics:Error Dynamics:
((A is Hurwitz)is Hurwitz)
∆−=
−+=
ννννν
xadpdcrm
��
( )∆−+= adrmrm bee νA�
Neural Network Adaptation
( ) ( )zzσzσ
∂∂='
(Diagonal Matrix)(Diagonal Matrix)
−−
=xxxx
erm
rmrm
��
June 2002 ENJ - Georgia Tech 10
Issues
•• Capability is LimitedCapability is Limited–– Saturation (Including Axis Priority), Rate LimitsSaturation (Including Axis Priority), Rate Limits
•• Not Feedback LinearizableNot Feedback Linearizable•• Sign of Control Effectiveness Becomes ZeroSign of Control Effectiveness Becomes Zero
–– Discrete Control (e.g., RCS Thrusters)Discrete Control (e.g., RCS Thrusters)
•• Need to Make a Flight Certification CaseNeed to Make a Flight Certification Case–– Show Adaptation Extremely Unlikely to Show Adaptation Extremely Unlikely to CauseCause Loss of VehicleLoss of Vehicle
•• Assumptions for Stability Need to be Extremely MildAssumptions for Stability Need to be Extremely Mild•• Require Recovery from Temporary “Faulty” AdaptationRequire Recovery from Temporary “Faulty” Adaptation
June 2002 ENJ - Georgia Tech 11
NN Adaptive Control with Pseudo-Control Hedging (PCH)
PDControl
PDControl
DynamicInversionDynamicInversion
NeuralNetworkNeural
Network
PlantPlantReference Model
Reference Model
-
++
TrackingTrackingErrorError
CommandCommandEstimateHedge
EstimateHedge
ActuatorActuatorcmdδ
hedgeν
δν
x
rmν
pdνadν−
June 2002 ENJ - Georgia Tech 12
Implications
•• “Shelter” Adaptive Element from the Adverse Effects “Shelter” Adaptive Element from the Adverse Effects of Plant Input Characteristics: of Plant Input Characteristics: –– Linear Dynamics, Latency, Saturation, Rate Saturation, etc.Linear Dynamics, Latency, Saturation, Rate Saturation, etc.
•• Achievable Adaptation Performance is Increased Achievable Adaptation Performance is Increased DramaticallyDramatically
•• Adaptation is Correct During SaturationAdaptation is Correct During Saturation–– Adaptive Element Can Recover from “Faulty” Adaptation Adaptive Element Can Recover from “Faulty” Adaptation
•• Enables Correct Adaptation When Not in Control of Enables Correct Adaptation When Not in Control of PlantPlant
•• Performance Improved Over Performance Improved Over Existing DesignExisting Design–– Attitude Error is LowerAttitude Error is Lower–– Hinge Moments Look GoodHinge Moments Look Good–– NothingNothing is Scheduled!is Scheduled!
BaselineBaseline
NNNN
June 2002 ENJ - Georgia Tech 15
-120
-60
0
60
120
0 50 100 150 200
time (sec)
attit
ude
erro
r (de
g)
roll pitch yaw
Ascent Phase Multiple Actuator Failures
-150
-100
-50
0
50
100
150
0 50 100 150 200
time (sec)
roll pitch yawBaselineBaseline
NNNN
•• Half of Aero Surfaces Fail Half of Aero Surfaces Fail HardHard--Over at 60 secOver at 60 sec
•• (All Right(All Right--Hand Surfaces Hand Surfaces Give Uncommanded Left Give Uncommanded Left Turn)Turn)
•• Occurs Near Max Q Occurs Near Max Q (60 Seconds)(60 Seconds)
FailureFailure
June 2002 ENJ - Georgia Tech 16
Ascent Phase Multiple Actuator Failures NN Controller
•• Saturates on All Three AxesSaturates on All Three Axes
•• Vehicle Rolls Three TimesVehicle Rolls Three Times
•• Full Recovery Once Full Recovery Once Dynamic Pressure Dynamic Pressure DropsDrops
EffectorsEffectors
June 2002 ENJ - Georgia Tech 17
•• Adaptation is “Correct” Adaptation is “Correct” During SaturationDuring Saturation
•• No Knowledge of No Knowledge of Failure Used Failure Used (Not Even in the (Not Even in the Hedge!)Hedge!)
Ascent Phase Multiple Actuator Failures NN Controller
Roll Axis PseudoRoll Axis Pseudo--Control SignalsControl Signals
-0.20
0.20.40.60.8
11.2
0 50 100 150 200 250
time (sec)
del vad
June 2002 ENJ - Georgia Tech 18
Subsequent Research Involving PCH
•• XX--33/RLV Attitude Control33/RLV Attitude Control
•• Adaptive Tracking and Control Adaptive Tracking and Control (Inner and Outer Loops) for RLV(Inner and Outer Loops) for RLV
•• Reconfigurable Flight Control for Civillian Aircraft Reconfigurable Flight Control for Civillian Aircraft (Training While Not in Control)(Training While Not in Control)
•• PCH is Used ToPCH is Used To–– Modify the Command Trajectory to Create the Feasible Reference Modify the Command Trajectory to Create the Feasible Reference
Trajectory (And Leave it Alone if Not at Limits)Trajectory (And Leave it Alone if Not at Limits)–– Protect Outer Loop Adaptation From Inner Loop DynamicsProtect Outer Loop Adaptation From Inner Loop Dynamics–– Protect Inner Loop Adaptation From Limited Control Authority Protect Inner Loop Adaptation From Limited Control Authority
(As Before)(As Before)
InnerLoop
InnerLoop
CommandTrajectory
OuterLoop
OuterLoop
NeuralNetworkNeural
Network
vx, θ
PCHPCH
June 2002 ENJ - Georgia Tech 21
Application to Rotorcraft Maneuvering
Yamaha R-MaxSimulation Results:Fly in a Circle While
Pirouetting-80 -60 -40 -20 0 20 40 60 80
-80
-60
-40
-20
0
20
40
60
80
Eas t
Nor
th
Network ON
Vel = 15 ft/sYaw = 45 o/sec
-80 -60 -40 -20 0 20 40 60 80-80
-60
-40
-20
0
20
40
60
80
Eas t
Nor
th
Network OFF
1st time around
“Circle”Network ON
Better Each Time
“Pentagon”Network OFF
June 2002 ENJ - Georgia Tech 22
Software Enabled ControlSponsored by DARPA
•• Develop softwareDevelop software--enabled control methods for enabled control methods for complex dynamic systems with application focus on complex dynamic systems with application focus on intelligent UAVsintelligent UAVs
•• SupportSupport--thethe--development and implement a plugdevelopment and implement a plug--andand--play, realplay, real--time software architecturestime software architectures
•• VTOL UAV hardwareVTOL UAV hardware--inin--thethe--loop simulation and flight loop simulation and flight testingtesting
June 2002 ENJ - Georgia Tech 23
Non-VolatileMemory Services
Serv
ices Scheduling
ServicesEvent
ServicesNaming
Services
PersistenceServices
TimerServices
TimeServices
Real-Time ORB
OS and Hardware Interfaces
ApplicationComponent
ApplicationComponent
ApplicationComponent
Bold Stroke Open Systems Architecture
•• RealReal--time CORBAtime CORBA--based Integration of Distributed, based Integration of Distributed, Heterogeneous ComponentsHeterogeneous Components
•• Utilizes Object Request Broker (ORB) Architecture Utilizes Object Request Broker (ORB) Architecture Developed by Washington University and BoeingDeveloped by Washington University and Boeing
•• Avionics and Simulation Tools Developed Over the Avionics and Simulation Tools Developed Over the Past YearPast Year
•• HardwareHardware--inin--thethe--Loop Simulation and Ground Loop Simulation and Ground Testing Started in November 2001Testing Started in November 2001
•• Navigation System Ground Tests Navigation System Ground Tests Completed February 2002Completed February 2002
•• Flights Testing (With Avionics)Flights Testing (With Avionics)Began March 2002Began March 2002
(RPM, Voltage, Pilot Inputs)(RPM, Voltage, Pilot Inputs)
•• Data LinksData Links–– 11 Mbps Ethernet Data Link11 Mbps Ethernet Data Link–– RSRS--232 Serial Data Link232 Serial Data Link
June 2002 ENJ - Georgia Tech 27
GTMax Hardware Integration
•• Exchangeable modules:Exchangeable modules:–– Flight Computer ModuleFlight Computer Module–– GPS ModuleGPS Module–– Data Link ModuleData Link Module–– IMU/Radar ModuleIMU/Radar Module–– Unused Module (Growth)Unused Module (Growth)–– Sonar/Magnetometer AssembliesSonar/Magnetometer Assemblies–– Power Distribution SystemPower Distribution System
•• Each module has selfEach module has self--contained power regulation contained power regulation and EMI shieldingand EMI shielding
•• Vibration isolated main Vibration isolated main module rackmodule rack
June 2002 ENJ - Georgia Tech 28
Onboard Avionics Architecture
Flight Computer
Serial Extension Board
FreewaveDGR-115
DC/DC
5V
5V
Battery 12V
Auxi
liary
Mod
ule
Aironet MC4800
EthernetHub
NovAtel RT-2 GPS Receiver
Auxiliary Computer /Payload
Radar Altimeter
ISIS-IMU
12V
DC/DC
DC/DC
HMR-2300Magnetometer
DC/DC
Sonar Altimeter
Power DistributionModule
Generator
Yamaha AttitudeControl System
RC Receiver YACS IMU
12V5V
RS-232 SerialEthernetDC Power
Dat
a Li
nkM
odul
eG
PS M
odul
eIM
U/R
adar
Mod
ule
Flig
ht C
ompu
ter
Mod
ule
5V
12V
June 2002 ENJ - Georgia Tech 29
Baseline Onboard Software
•• NavigationNavigation–– 17 State Extended Kalman 17 State Extended Kalman
•• Hardware In the Loop Hardware In the Loop Simulation CapableSimulation Capable
•• The Desktop Computer The Desktop Computer Simulation Utilizes Simulation Utilizes –– Actual Flight SoftwareActual Flight Software–– Actual Ground Control Station Actual Ground Control Station
SoftwareSoftware–– Flight Test Verified Dynamic Flight Test Verified Dynamic
Model of HelicopterModel of Helicopter–– Flight Test Verified Model of All Flight Test Verified Model of All
Sensors/ActuatorsSensors/Actuators–– Scene Generation CapabilityScene Generation Capability
June 2002 ENJ - Georgia Tech 31
Software in the Loop (SITL)•• Test algorithms within the simulationTest algorithms within the simulation•• Generate emulated sensor data from an aircraft simulation Generate emulated sensor data from an aircraft simulation
(including errors)(including errors)
Vehicle Model
SensorDrivers
SensorEmulation(w/ Error Model)
ActuatorDriver
SensorData
StateEstimate Control
ActuatorSimulation
State Control
SensorRaw Data
ActuatorRaw Data
Desktop Computer
TrajectoryPlanner
OtherSystems
FlightController
NavigationFilter
June 2002 ENJ - Georgia Tech 32
Hardware in the Loop (HITL)•• Flight software runs on the onboard computerFlight software runs on the onboard computer
•• Onboard computer “thinks” it is flying the vehicleOnboard computer “thinks” it is flying the vehicle
Vehicle Model
SensorDrivers
SensorEmulation(w/ Error Model)
ActuatorDriver
SensorData
StateEstimate Control
ActuatorSimulation
State Control
SensorRaw Data
ActuatorRaw Data
Desktop Computer
Flight Computer
TrajectoryPlanner
OtherSystems
FlightController
NavigationFilter
June 2002 ENJ - Georgia Tech 33
GTMax Flight Operations
•• Network Connections Network Connections Available At Ground Control Available At Ground Control Station from HubStation from Hub–– Multiple Laptops Can Multiple Laptops Can
Communicate with Onboard Communicate with Onboard Computers SimultaneouslyComputers Simultaneously
•• Due to Generator, Due to Generator, Endurance Limited by Endurance Limited by Onboard Fuel (> 1 hour)Onboard Fuel (> 1 hour)
•• Ground Equipment Can Ground Equipment Can Operate on 115VAC or Operate on 115VAC or 12VDC and Has Battery 12VDC and Has Battery BackupBackup
June 2002 ENJ - Georgia Tech 34
Flight Testing in McDonough, Georgia
June 2002 ENJ - Georgia Tech 35
First Tests w/ Baseline Controller
•• Neural Network Adaptive Controller on First Flight Neural Network Adaptive Controller on First Flight Test Day (April 10, 2002)Test Day (April 10, 2002)
•• Even With LargeEven With LargeModel Errors, Model Errors, System Was AbleSystem Was AbleTo Control theTo Control theHelicopterHelicopter
June 2002 ENJ - Georgia Tech 36
Results With Baseline Controller
Step Input of Altitude Command:
1 130 1 135 1140 114 5158
160
162
164
166
168
170
time (sec )
altit
ude
(ft)
s tep input of pos ition c ommand, do wn
P os ition Es tima teP os ition Command
June 2002 ENJ - Georgia Tech 37
Flight Control Reconfigurations
•• Switched Between Neural Network Adaptive Switched Between Neural Network Adaptive Controller to Much Simpler Conventional Inverting Controller to Much Simpler Conventional Inverting Controller and BackController and Back
•• Real Time and Real Time and Closed LoopClosed Loop
1 80 190 2 00 210 2 20 230 2 40 250325
330
335
340
345
350Con trol Rec onfig ura tion
East
(ft)
time (sec )
P os ition Es tima teP os ition Command
Reconfiguration at 206.24
June 2002 ENJ - Georgia Tech 38
Collective control failureCollective control failure
Failure detection and control reconfiguration with RPM control
With fault tolerantand reconfigurablecontrol system
With fault tolerantand reconfigurablecontrol system
Simulated Main Rotor Actuator Failure
Without fault tolerantand reconfigurablecontrol system
Without fault tolerantand reconfigurablecontrol system
June 2002 ENJ - Georgia Tech 39
Tail rotor failureTail rotor failure
Without fault tolerant and reconfigurable control system
Without fault tolerant and reconfigurable control system
Tail Rotor Failure (in Simulation)
Gain altitude usingmain rotor collective
Control reconfiguration using main rotor controls
Translatory descent to a clear area
Control reconfigurationfor autorotation
With fault tolerant and reconfigurable control system
With fault tolerant and reconfigurable control system
Autorotation and landing
June 2002 ENJ - Georgia Tech 40
International Aerial Robotics 2001-
Launch AreaLaunch Area
Two Lights Two Lights Identify Identify BuildingBuilding
3 km3 km
Vehicle orVehicle or SubvehicleSubvehicle(s) Enter Building(s) Enter Building
Building and an Building and an Entry Point FoundEntry Point Found
>1m>1m
Transmit an Image of “Point of Interest” Transmit an Image of “Point of Interest” Inside BuildingInside Building
In Less Than 15 Minutes:In Less Than 15 Minutes:
Image ReceiverImage Receiver(& Other(& Other GoundGound Components)Components)Sign OverSign Over
EntryEntry
June 2002 ENJ - Georgia Tech 41
International Aerial Robotics Competition
•• Unmanned and Autonomous Unmanned and Autonomous (No Active Human Operators, no Tethers)(No Active Human Operators, no Tethers)
•• Some Components Can Remain on the Ground Some Components Can Remain on the Ground (e.g., Additional Computers, Navigation Aids)(e.g., Additional Computers, Navigation Aids)
•• Launch and Recovery Need Launch and Recovery Need NotNot Be AutonomousBe Autonomous
•• Mission is Divided into “Levels”Mission is Divided into “Levels”
•• Each Teams Gets 60 Minutes To Fly (...Per Year)Each Teams Gets 60 Minutes To Fly (...Per Year)
•• Rules Change Once a Mission is CompletedRules Change Once a Mission is Completed
•• Level 1: Follow Prescribed Waypoints for 3kmLevel 1: Follow Prescribed Waypoints for 3km
•• Level 2: Locate Building and Find an EntryLevel 2: Locate Building and Find an Entry
•• Level 3: Enter the Building Level 3: Enter the Building –– Can Be a Different Vehicle or Can Be a Different Vehicle or SubvehicleSubvehicle That Used AboveThat Used Above–– Can Launch Near Target StructureCan Launch Near Target Structure
•• Level 4: Image Desired Location Within Building and Level 4: Image Desired Location Within Building and Transmit Transmit –– Complete In < 15 Minutes (Launch to Data Retrieval)Complete In < 15 Minutes (Launch to Data Retrieval)
•• Contest is Over Once Somebody Does Level 4Contest is Over Once Somebody Does Level 4
June 2002 ENJ - Georgia Tech 43
2001 Airplane: ¼ Scale Cub
June 2002 ENJ - Georgia Tech 44
Winning Flight in 2001, Level 1
-76.44-76.435
-76.43-7 6.425
-7 6.42-7 6.415
38.14
38.14 2
38.144
3 8.14 6
38.148
38. 15
38.15 2
38. 1540
500
Longitude
Latitude
Alti
tude
Automatic FlightManual Takeoff/Landing
Waypoint 1
Waypoint 2
Waypoint 3
Waypoint 4 & Holding Pattern
“Runway” & Ground Station
Under Automatic Flight:Under Automatic Flight:Distance Traveled: 3.1 mi / 4.9 kmDistance Traveled: 3.1 mi / 4.9 kmTime: 3 min 9 secTime: 3 min 9 secAverage Speed: 58 mph / 93 Average Speed: 58 mph / 93 kphkphMax Distance from Ground Station: ½ mi / 0.8 kmMax Distance from Ground Station: ½ mi / 0.8 kmAverage Altitude: 397 ft / 121 mAverage Altitude: 397 ft / 121 m
June 2002 ENJ - Georgia Tech 45
Plans for 2002
•• Level 2Level 2–– Add a Video Camera and Image Processor Add a Video Camera and Image Processor
(Donation from Texas Instruments)(Donation from Texas Instruments)–– Switch GPS to DSwitch GPS to D--GPS For Level 2 Accuracy (NovAtel)GPS For Level 2 Accuracy (NovAtel)–– Update Ground Station Software and Develop Image Processing Update Ground Station Software and Develop Image Processing
Software Software –– Possibly Also Switch to GTMaxPossibly Also Switch to GTMax
•• Level 2+Level 2+–– Design, Building, and Testing for SubDesign, Building, and Testing for Sub--Vehicle: Drops From Airplane Vehicle: Drops From Airplane
and Enters Building and Enters Building –– Designs for Operation Inside Building (Levels 3 & 4)Designs for Operation Inside Building (Levels 3 & 4)
June 2002 ENJ - Georgia Tech 46
Some Potential Areas of Collaboration
•• VTOL and FixedVTOL and Fixed--Wing UAV Flight TestingWing UAV Flight Testing–– Lessons LearnedLessons Learned–– Simulation Models and SoftwareSimulation Models and Software–– GTMax as a Research Flight Test PlatformGTMax as a Research Flight Test Platform
•• Studies at Georgia TechStudies at Georgia Tech
June 2002 ENJ - Georgia Tech 47
UAV Avionics at the2002 Digital Avionics Systems Conference•• AIAA/IEEEAIAA/IEEE
•• October 27October 27--31; Irvine, California31; Irvine, California
•• NEWNEW Applications of Avionics: Applications of Avionics: Uninhabited Air Vehicles (UAV) & Missiles Track:Uninhabited Air Vehicles (UAV) & Missiles Track:–– Avionics systems for UAVs, intelligent systems for vehicle autonAvionics systems for UAVs, intelligent systems for vehicle autonomy, omy,
operation of UAVs in controlled airspace, payloads, missiles, anoperation of UAVs in controlled airspace, payloads, missiles, and d guided munitionsguided munitions
–– 5 Sessions5 Sessions–– Paper Acceptance Still Possible for New Track, But Act FastPaper Acceptance Still Possible for New Track, But Act Fast
•• Contact: Eric N. JohnsonContact: Eric N. Johnson404404--385385--2519, Eric.Johnson@2519, [email protected]