+ All Categories
Home > Documents > UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen...

UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen...

Date post: 04-Jan-2016
Category:
Upload: helena-dennis
View: 213 times
Download: 0 times
Share this document with a friend
Popular Tags:
21
UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt
Transcript
Page 1: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

UAV Navigation by Expert System for Contaminant

Mapping

UAV Navigation by Expert System for Contaminant

MappingGeorge S. Young

Yuki Kuroki, Sue Ellen Haupt

Page 2: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

GoalsGoals

BackgroundBackground• Source and wx information needed for contaminant modeling• Long et al.(2008) demonstrated the use of Gaussian puff to back- calculate the source characteristics via a Genetic Algorithm

BackgroundBackground• Source and wx information needed for contaminant modeling• Long et al.(2008) demonstrated the use of Gaussian puff to back- calculate the source characteristics via a Genetic Algorithm

ConstraintsConstraints• Number of sensors & time to solution

ConstraintsConstraints• Number of sensors & time to solution

MissionMission• Identify a total of 4 parameters (source strength, source location (x,y) and wind direction) describing the release using mobile sensors

MissionMission• Identify a total of 4 parameters (source strength, source location (x,y) and wind direction) describing the release using mobile sensors

Page 3: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Dispersion Dispersion modelmodel

Dispersion Dispersion modelmodel

Gaussian Gaussian plumeplume

Gaussian Gaussian plumeplume

GaussianGaussian puffpuff

GaussianGaussian puffpuff

noisenoisenoisenoise

Identical twin experiment

System ComponentsSystem Components

ModelModel inverterinverterModelModel

inverterinverter

Genetic Genetic AlgorithmAlgorithmGenetic Genetic AlgorithmAlgorithm

Nelder-Mead Nelder-Mead downhill simplexdownhill simplexNelder-Mead Nelder-Mead

downhill simplexdownhill simplex

ObservingObserving systemsystem

ObservingObserving systemsystem

Fixed Fixed concentrationconcentration

sensorsensor

Fixed Fixed concentrationconcentration

sensorsensor

Autonomous Autonomous aircraftaircraft

Autonomous Autonomous aircraftaircraft

Page 4: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Dispersion modelDispersion model

2

2

2

2

2

2

2

2

5.1 2exp

2exp

2exp

2exp

2 z

er

z

er

y

r

x

r

zyx

r

HzHzyUtxtQC

Gaussian Puff• An instantaneous release

Gaussian Puff• An instantaneous release

true conc

1 2 3 4 5 6 7 81

2

3

4

5

6

7

8

1

2

3

4

5

6

7x 10

-6

Gaussian plume• A time averaged continuous emission• wind speed, eddy diffusivity are const• Mass is conserved

Gaussian plume• A time averaged continuous emission• wind speed, eddy diffusivity are const• Mass is conserved

2

2

2

2

2

2

2exp

2exp

2exp

2),,(

yzzzy

yhzhz

u

QzyxC

C: the concentration, Q: the emission masst: the length of time of the release itself t: the time since the release U: the wind speed : the standard deviations h: source height

C: the concentration, Q: the emission masst: the length of time of the release itself t: the time since the release U: the wind speed : the standard deviations h: source height

Page 5: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Hybrid Genetic Algorithm (GA)

Hybrid Genetic Algorithm (GA)

Mutation

Mate Selection

Mating

Optimization with a GA

Optimization with a GA

Evaluatecost Converge?

Initialize population

Solution

no Yes

Exchange informationBetween parents

Combine best of last generation

Nelder MeadeDownhill Simplex

Fine-tune GA solution

Page 6: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

GA TuningGA Tuning

1. What we did?• Determine best combination of GA parameters

1. What we did?• Determine best combination of GA parameters

Pseudo-Runtime= pop*it# Pseudo-Runtime= pop*it#

102

103

104

105

106

107

10-3

10-2

10-1

100

101

Pseudo Runtime

Err

or

Ma

gn

itu

de

Error Sensitivity to GA parameters for snr = 5

0.010.020.040.080.160.32

105

10-4

10-3

10-2

10-1

100

Pseudo Runtime

Err

or M

agni

tude

q=1 Error Sensitivity to GA parameters for snr = 5

0.010.020.040.080.160.32

2. Concerns?• Minimizing CPU timeMinimizing CPU time• Increasing accuracyIncreasing accuracy

2. Concerns?• Minimizing CPU timeMinimizing CPU time• Increasing accuracyIncreasing accuracy

3. Best combination?• Population size = 40Population size = 40• Mutation rate = 0.32 Mutation rate = 0.32 • Iteration counts = 640Iteration counts = 640

3. Best combination?• Population size = 40Population size = 40• Mutation rate = 0.32 Mutation rate = 0.32 • Iteration counts = 640Iteration counts = 640

Page 7: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Experimental SetupExperimental Setup

• Wind direction 270 degrees

• Random source location in upwind half of domain

• Single fixed sensor in downwind half of domain

• UAV takes off from upwind corner of domain

– Worst case position

– Launches on first detection by fixed sensor

• UAV speed is 4 times wind speed

Page 8: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Autonomous AircraftAutonomous Aircraft

Why use aircraft?Why use aircraft?• Equipping the UAV with GPS & concentration sensor • Avoid the cost of a dense array of fixed sensors

Why use aircraft?Why use aircraft?• Equipping the UAV with GPS & concentration sensor • Avoid the cost of a dense array of fixed sensors

Why autonomous?Why autonomous?• AI required for rapid decision making• Ensemble of manned aircraft would be too expensive

Why autonomous?Why autonomous?• AI required for rapid decision making• Ensemble of manned aircraft would be too expensive

Why virutalWhy virutal• Test in a fully controlled environment• Test UAV naviagtion algorthims without societal risk

Why virutalWhy virutal• Test in a fully controlled environment• Test UAV naviagtion algorthims without societal risk

Page 9: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Information FlowInformation Flow

• UAV AI needs observed & modeled concentration fields to navigate

• UAV AI needs observed & modeled concentration fields to navigate

• GA needs UAV wind & concentration observations to locate source

• GA needs UAV wind & concentration observations to locate source

• Forward model needs wind and source locaton to predict concentration field

• Forward model needs wind and source locaton to predict concentration field

Page 10: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Expert System DesignExpert System Design

•Plume Puff Difference

– How many passes

through plume?

– How much separation

in space?

– How many passes

through puff?

– How much separation

in time?

– Why the difference?

Amount of data needed

Page 11: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Plume Expert SystemPlume Expert System

• Plume decision logic

pass1actual source

sensorpass2

-700 700

700 Route 2

-700 700

Route 1

300m

Route 3

-700 700

Page 12: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Puff Expert SystemPuff Expert System

• Puff decision logic

aircraftaircraft u

anNt

22

windpuff u

axt

0coscossinsin 2121

1tanx

nN

2tantan n

ax

Origin

Sensor

pass1

pass2

Pass1 Max Conc

N

y

a

n

Mean wind direction

(-7000,7000)

Page 13: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Flight Track – Plume Example

Flight Track – Plume Example

Page 14: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Flight Track – Puff ExampleFlight Track – Puff Example

Page 15: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Testing ArchitectureTesting Architecture

Identical twin experimentCreate data

NoiseContaminate data

Collect data

• Monte Carlo testing of UAV non-collaborative ensemble• Pseudo-random initial population and sensor location

Hybrid GA optimizing

Ensemble size

• Ensemble median to back calculate source and wind dir.• Monte Carlo mean of ensemble median will be shown

Page 16: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Plume ResultsPlume Results

245

250

255

260

265

270

275

280

285

290 wind direction distribution

Ensembel member

win

d di

rect

ion

Route1

Route2Route3

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

4 strength distribution

Ensembel member

stre

ngth

dis

tribu

tion

Route1

Route2Route3

-60

-40

-20

0

20

40

60

80 x distribution

Ensembel member

x di

strib

utio

n

-30

-20

-10

0

10

20

30

40 y distribution

Ensembel member

y di

strib

utio

n

4 10 20 50 4 10 20 50

Wind Concentration

X Y

0.2m0.3m

0.05 0.02 [kg/s]

Page 17: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Puff ResultsPuff Results

268

268.5

269

269.5

270

270.5

271

271.5 wind direction distribution

Ensembel member

win

d di

rect

ion

0.75

0.8

0.85

0.9

0.95

1

1.05

1.1

1.15

1.2

1.25 strength distribution

Ensembel member

stre

ngth

dist

ribut

ion

-6015

-6010

-6005

-6000

-5995

-5990

-5985 x distribution

Ensembel member

x di

strib

utio

n

-300

-250

-200

-150

-100

-50

0

50

100

150

200 y distribution

Ensembel member

y di

strib

utio

n

4 10 20 50 4 10 20 50

Wind Concentration

X Y

3m55m

0.3 0.02 [kg/s]

Page 18: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

ConclusionsConclusions

ExperimentalExperimentalSetupSetup

GaussianGaussianPuff UAVPuff UAV

DiscussionDiscussion

• Idential twin• 1 fixed sensor• Single UAVor • UAV ensemble• No cooperation

• 2 flight legs• 1 UAV• UAV navigation by expert system• GA optimization for source & dir• 1400m domain• Results improve

• 6 flight legs• 20 UAVs• Median Solution• 14km domain • Greater tracking challenge• Most UAVs succeed

GaussianGaussian plume UAVplume UAV

• UAV Ensemble • Expert system naviagaion

Solves single-sensor source

characterization

Page 19: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Future WorkFuture Work

Goal: Compensate for the tight time constraints inherent in emergency management

• Cooperation between Multiple UAVs• Improve Gaussian Puff Model Navigation• Actual UAVs• Field Test

Page 20: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

AcknowledgementsAcknowledgements

• The second author was supported by Japan Ground Self Defense Forces during this study

• Thanks to J. Wyngaard, K. Long, A. Annunzio, A. Beyer-Lout, L. Rodriguez for insights and advice

Page 21: UAV Navigation by Expert System for Contaminant Mapping George S. Young Yuki Kuroki, Sue Ellen Haupt.

Questions?Questions?


Recommended