22
Maintaining a Linked Network Chain Utilizing Decentralized Mobility Control AIAA GNC Conference & Exhibit Aug. 21, 2006 Cory Dixon and Eric W. Frew

8/22/20061 Maintaining a Linked Network Chain Utilizing Decentralized Mobility Control AIAA GNC Conference & Exhibit Aug. 21, 2006 Cory Dixon and Eric

  • View
    213

  • Download
    0

Embed Size (px)

Citation preview

Maintaining a Linked Network Chain Utilizing Decentralized Mobility Control

AIAA GNC Conference & Exhibit

Aug. 21, 2006

Cory Dixon and Eric W. Frew

8/22/2006 2

Long Range Sensing:Chaining with Small UAs

Operational Range determined by the limiting value

Endurance / Fuel Range

Communication Range

Fuel range >> communication range for single UA

Limited size for antenna and electronics

Limited available power

Team of UAs

Can utilize ad hoc communication network

Extends communication range using relay nodes

Adds robustness to aircraft loss

8/22/2006 3

Chaining Problem for Mobile Vehicles

Radio Chaining: maintaining a communication link along a chain of vehicles using only locally measured

communication performance metrics.

Chaining Objectives

Maintain communication along a chain of vehicles

Increase operational range of the UAV team by using a chain of airborne relays

Maximize UAV spacing to minimize the number of required UAV relays

Maximize link throughput

Chaining Applications

Long-range sensing and communication

Increasing search area for Search & Rescue missions

Provide communication to disconnected networks

8/22/2006 4

AUGNet:Ad Hoc UAV Ground Network

SoekrisModel 4511

100 MHz

OrinocoPCMCIA 802.11b

Antenna

Adjustable 100 mWto 1 watt bidirectional amplifier

GPS receiver

. .92 @ FS1

82

81

858483 on

& off. .

92 @ FS182

81

858483 on

& off

FS2

85. .FS2

85. .

Disconnected Networks

FS2

85

FS2

85

UAV-UAV Chain

UAV Swarm

8/22/2006 5

Communications & Control:A Closed Loop System

Wireless ad hoc network communication

performance and vehicle mobility

control form a closed loop system.

Integrate communication performance into control architecture and use

mobility control to maintain/improve communication performance.

C

SNR

edesiredSNR

w Pos

Ad Hoc Network

R

R R

R

+

-+

RR

R

R

Network Performance

Team Control

Network Topology

Vehicle Mobility

Network Performance

Team Control

TopologyVehicle Mobility

8/22/2006 6

References & Related Work

Communication as Control PrimitiveControlled mobility to Improve Network Performance

(Goldenberg et al., 2004)

(Dixon and Frew, 2005) – “Leashing of an Unmanned Aircraft to a Radio Source”

Connectivity & Limited Range Communications

(Beard and McLain, 2003)

(Spanos & Murray, 2004)

Vehicle Control in a Sampled EnvironmentCooperative Level Set Tracking (Boundary Tracking)

(Hsieh et al., 2004), (Marthaler & Bertozzi, 2003)

Cooperative Gradient Climbing

(Bachmayer et al., 2002), (Ogren et al., 2004)

Adaptive Sampling Utilizing Vehicle Motion

(Fiorelli et al., 2003)

Path (Route) Tracking for Nonholonomic VehiclesOptimal Control of Bounded-Curvature Vehicles

(Soures et al., 2000)

(Balluchi et al., 2005)

8/22/2006 7

-1000 -500 0 500 1000

-1000

-500

0

500

1000

X-Location [m]

Y-L

ocat

ion

[m]

SNR Field Lines

-1000 -500 0 500 1000

-1000

-500

0

500

1000

X-Location [m]

Y-L

ocat

ion

[m]

SNR Field Lines

Channel Capacity and Signal-to-Noise Ratio

Shannon Channel Capacity

Radio Propagation EnvironmentReceived power is directional and link dependent

Interference is dependent upon the location of the UAV

Exponential power decay and fast fading (noisy channel)

frrKrPxN

rPrxSNR jijijijij

jj

jijij

)()(

)(

)(),(

SNRBC 1log2

Communication Range

Radio EnvironmentRadio EnvironmentThroughput vs. RangeThroughput vs. Range

iNi

iNi

SNRC

minmaxminmax

8/22/2006 8

UA Kinematic Model

UAV Motion – Unicycle Model0 < VMIN ≤ VO ≤ VMAX

steering input: |u| ≤ uMAX

1

11

11

1 2sin

2cos

k

kko

kko

kkk

Tu

TuV

TuV

TXXX

u

Vy

Vx

T

T

sin

cos

8/22/2006 9

Motion of Vehicle Motion of Vehicle (Discrete Sampling Time)(Discrete Sampling Time)

SNR Gradient Field SNR Gradient Field (no localized noise)(no localized noise)

Change in Measured SNRChange in Measured SNR

Path GradientPath Gradient

SNR Path Gradient

rSN

KrS ii ˆ2/)1(

0

0

11

111 coscos

sinsin

kkk

kkk

k

Tkkk Tu

TuVxxx

*

=

cos, kkkkk xSxSS

C

SNR

edesiredSNR

w Pos

Ad Hoc Network

R

R R

R

+

-+

RR

R

R

T

=> kkk xSg

8/22/2006 10

Balancing SNR Link Gradients

Re-cast Control ProblemControl motion of an orbit center point (i.e. point mass)Autopilot system tracks orbit point

Gradient Estimates for Each Link

Feedback ForceMove the point mass by generating forces based on link gradient.

Scaling Parameter Ki

a=F/mPos

Ad Hoc Network

R

R R

R

RR

R

R

F(Si)

Orbit Point

sensorF

])[][(][ kxkSkgGorbitkorbitk

ii

m

vCFa o

iii GKf

linksi

ifF

linksjj

ilinksj

j

i S

SSK

)(

)(

else0)min(arg if1 i

iSi

K

8/22/2006 11

Simulation:Two Static Nodes with Two Helper Nodes

8/22/2006 12

Extension to Multiple Nodes

50 100 150 200 250 300 350 400

2

4

6

8

10

12

14

16

Time [s]

SN

R

Node 1Node 2Node 3Node 4

kdcPT

p

a = 2a > 2a => ∞

q

Energy Optimal Placement

8/22/2006 13

Extremum Seeking Algorithms

Time derivative of Cost Function

Path Gradient

Low-pass filtering generates control update

+

r cos(g )

J(p )

X

oo

+

r sin(g )oo

x-k s

Q^

x-k s

Q^

ss + wh

Fo

ss + wh

ss + wh

Y

X

Y

ws + wl

l

ws + wl

l

QX

QY

j

p kp

i

V

kP JJ p ,

k

kPP JJ

p

p

)cos()sin(

tt

Vo

ok

p

)(cos)cos()sin(

)cos()sin()(sin)( 2

2

tJttJ

ttJtJVtJ

oyoox

ooyoxP

y

xP J

JVJLPF

2)(

Ad Hoc Network

R

R RR

RR

R

R

J(Si,Sk)

Pos SNR

J(p )j

k

jkjijkji SSSSJ 11),(

Objective Map: J(pi)

8/22/2006 14

Extremum Seeking:Simulation

8/22/2006 15

Research Questions & Algorithm Improvements

Stability & SensitivitySimulations have shown the controller to be stable, but a formal proof is still required

Wireless communication channels are noisy (fast fading) and will require the SNR signal to be smoothed

Effects of node mobility

Estimation of SNR (performance) fieldEstimate the field to improve gradient estimation

Radio source localization, and noise source detection and localization

Initialization and Node Task AssignmentWhen to introduce a relay node?

In what position of the chain should it fill?

Additional Control Parameters to ConsiderGPS position to improve tracking of highly mobile nodes

Link importance and communication requirementsReal-time vs. delay tolerant data

Bandwidth requirement and node utility

4/27/05 16

Conclusion & Future Work

SNR as Control InputDoes not require any additional communication

Is extensible from one node to many nodes

Finds the energy optimal location regardless of environment

Provides a robust measure of link quality and bandwidth

Simulation ResultsShow that the SNR, sampled at 1 Hz with GPS, can be used to leash a single aircraft to two mobile nodes.

To obtain a leashed chain with multiple aircraft requires additional information to be shared, such as position, to aid in proper chain ordering

Future workExperimental testing utilizing AUGNet platform, i.e. Ares UAV and MNR radios.

Adapt ES algorithms and methods to provide research base

Next Talk – Phase Transitions for Controlled MobilityWhen should a relay node be introduced into the chain to maintain a communication throughput requirement?

The SNR controllers presented provide a robust control method that is capable of leashing disconnected nodes in the presence of localized

noise disturbances.

8/22/2006 17

http://RECUV.Colorado.edu

Questions and Comments are Welcomed!

Thanks for [email protected]

8/22/2006 18

8/22/2006 19

u

Vy

Vx

T

T

sin

cos

TV

g maxmax

tan

11

111 coscos

sinsin

kkk

kkk

k

Tkkk Tu

Tu

u

Vxxx

FrorKrP jijijijij log1)(log10)(

jjjijij xNrPrxSNR ,

T

iPi V

xxrSS

,

][],[][ kxkSkS

][][][ kxkSkgi

][

][

kg

kgG

i

ii

iii GKF

j

iji S

SSK

m

kvCFka oi ]1[

][

Tkakvkv

TkaTkvkxkx

oo

ooo

][]1[][

2/][]1[]1[][ 2

2/0 To Vv

NSBC 1log2

8/22/2006 20

Communications & Control:A Closed Loop System

Wireless ad hoc communication performance and vehicle mobility control form a closed loop system.

MANETS

Fault Tolerant Networks

Swarm Intelligence

Formation Flying

Data Ferrying

Mobile Infostations

Networked Control Systems (NCS)

Distributed Cooperative Control

Network Performance

Team Control

Network Topology

Vehicle Mobility

Network Performance

Team Control

TopologyVehicle Mobility

8/22/2006 21

Closing the C2 Loop on SNR

Communication Performance as a Control PrimitiveIntegrate communication performance into control architectureExploit mobility control to maintain/improve communication performance

Closed Loop UAV Steering ControllerAssume vehicle has low-level autopilot system controlling altitude and airspeedUse the SNR of each neighbor link to form the feedback signal Generate bounded steering commands for use by an autopilot

C

SNR

edesiredSNR

w Pos

Ad Hoc Network

R

R R

R

+

-+

RR

R

R

n

i iiK

k i

K

k ii GamFS

ktg

ktgG

1

1

1

1

8/22/2006 22

i

Tracking a Communication Performance Metric

Maintain communication link?Traditionally (position based):

Range ≤ RangeMAX

Communication performance motivated:

Throughput ≥ ThroughputMIN

Communication Performance FieldCan view performance as a continuous, measurable fieldDistribution of field does not need to be known a priori

Performance FieldPosition Based

Chaining ObjectivesThroughput ≥ ThroughputMIN defines a communication regionMaximizing sensor coverage reduces region to outer bound

ri

ji

ii