Actor Positioning Based on Molecular Geometry in …gsolmaz/publications/ICC12... · in Aerial...

Actor Positioning Based on Molecular Geometryin Aerial Sensor Networks

Actor positioning strategy for aerial sensor networks

Mustafa Ilhan Akbas, Gürkan Solmaz and Damla Turgut

Department of Electrical Engineering and Computer ScienceUniversity of Central Florida - Orlando, FL

June 13, 2012

M. I. Akbas, G. Solmaz, D. Turgut (UCF) ICC 2012 June 13, 2012 1 / 17

1 Application scenario



2 Problem definition




3 System model




3 System model

4 Positioning method




3 System model


5 Simulation study




3 System model


5 Simulation study

6 Conclusion


Application scenarioVolcanic eruption such as the volcano Eyjafjallajökull in 2010Close-up observation of the volcano was impossibleUAV system with built-in sensors to investigate volcanic plume

Sink UAV

Actor UAV

Actor UAV

Actor UAV

Actor UAV

Sensor UAV


Problem definition

Problem◮ For effective data collection, positioning of UAVs is important◮ Most dynamic node positioning strategies limited to 2-D space◮ Popular 2-D strategies become NP-Hard in 3-D space

Objective◮ Dynamic positioning of the actors in three dimensional space with

local communication


System model

Challenges◮ UAV system has autonomous flight operation mode◮ Autonomous flight may reduce situational awareness and error

correction◮ The communication must be simple yet effective◮ The actors must be able to reorganize in case of a loss

System◮ WSAN of small UAVs with built-in sensor nodes◮ Larger and more powerful UAVs with actor nodes◮ Central, most powerful UAV serving as the sink node


System dynamics

Affiliation of sensor to actor nodes is executed as in SOFROP†

◮ Actors assigned with weight k◮ Sensor nodes initially get random weight values◮ Sensor nodes update weight = k - (hop count of sensor node)◮ Only data available for a sensor node s, are the direct neighbors

Neigh(s) and their corresponding weights w(Neigh(si ))◮ Sensor nodes maintain and update only local information

Nodes have spherical transmission ranges

Network among actors and the sink form the communicationbackbone

†M. I. Akbas, M. R. Brust, and D. Turgut. “SOFROP: Self-Organizing and Fair Routing Protocol for Wireless Networks with Mobile

Sensors and Stationary Actors” Elsevier Journal of Computer Communications, in early access

DOI:10.1016/j.comcom.2011.01.006, 2011.

†M. I. Akbas, M. R. Brust, and D. Turgut. “SOFROP: Self-Organizing and Fair Routing Protocol for Wireless Networks with Mobile

Sensors and Stationary Actors” In the Proceedings of IEEE Local Computer Networks (LCN‘10), pp. 456–463, October 2010.


“VSEPR theory” based approach

VSEPR (Valence Shell Electron Pair Repulsion) model is the mostsuccessful model for the molecular geometry prediction

Arrangement of electron pairs in valence shell of the central atomare due to the repulsion between them

VSEPR theory is adopted to build a self-configuring dynamicnetwork architecture


“VSEPR theory” geometriesPeripheral atoms mapped to actors and central atom to the sink

Linear

Trigonal planar

Tetrahedral

Trigonal bipyramid Pentagonal bipyramid

Octahedral Square Antiprismatic

Sink

Sink Sink

Sink Sink

Sink

Sink


Formulation of VSEPR geometries

Examples:

Positions of actors in Linear geometry:

pa1(x , y , z) = (r , 0, 0) pa2(x , y , z) = (−r , 0, 0)

Positions of actors in Trigonal planargeometry:

pa1(x , y , z) = (r , 0, 0)pa2(x , y , z) = (−r .sin(30◦), r .sin(60◦), 0)pa3(x , y , z) = (−r .sin(30◦),−r .sin(60◦), 0)

Positions of actors in Tetrahedral geometry:

pa1(x , y , z) = (0, 0, r)pa2(x , y , z) = (−r .a,−r .b, r .cos(109.5◦))pa3(x , y , z) = (−r .sin(109.5◦), 0, r .cos(109.5◦)pa4(x , y , z) = (−r .a, r .b, r .cos(109.5◦))

Sink

x

y

z

a 1 a 2

Sink

120

120

120 o

o

o x

y

z

a 1

a 2

a 3

Sink

a 1

a 4

x

y

z

a 2

a 3

109.5 o

109.5 o


Formulation of VSEPR geometries

Actor locations must be identified according to a reference point

Sink taken as the reference origin in XYZ coordinate system inflight

Transition between geometries must not be complex◮ Geometries formulated s.t. transition from one to another requires

least number of position changes⋆ When number of actors is between 1-3, actors located on a single

plane⋆ When number of actors is between 4-7, 2 actors located on z-axis,

others located on single plane with equal connection angles⋆ When number of actors is 8, actors located on 2 planes


Dynamic positioning

Transitions betweengeometries done with alightweight (depending onbasic rules) algorithm

There is no operation center orremote control

Changes and maintenancethrough local communicationonly

Affiliation of sensor nodes tothe actors are handled bySOFROP networkorganization

Transitions between geometriesif n < 4 then

Position on z = 0 planePositions with Θ = 360

nelse if 8 > n > 4 then

Positions:Θ = 360

n−2 on z = 0 plane

Θ = 90◦ for y -axis and z = 0 neighborselse if 13 > n > 7 then

Positions:Θ = 360

a and (z = h2 )&(x = 0 or y = 0)

Θ = 360a and (z = −h

2 )&(x = y or x = −y)

Θ = 90◦ for y -axis and z = 0 neighborsif n = 9 then

Θ = 360n/3 on z = 0 plane

end ifelse if ids = Minimum among neighbors then

Position on z-axiselse

Position at z = −r .cos(109.5◦) equally spacedend if


1-hop Coverage for Different GeometriesScalability improvement by

◮ Inreasing number of actors and sinks◮ Specifying VSEPR geometries for multiple sinks

The performance difference becomes apparent as the number ofactors exceeds 7

2 4 6 8 10 12 140

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

2.2x 10

6

Number of data collectors

1−ho

p co

vera

ge o

f net

wor

k ba

ckbo

ne (

m3 )

1−sink geometries2−sink geometries


1-hop Coverage of Actors

Our protocol vs. Random positioning with a central node

Our protocol performs better with an increasing difference as thenumber of actors increases

0 2 4 6 8 10 120

2

4

6

8

10

12

14

16

x 105

Number of actors

1−ho

p co

vera

ge o

f net

wor

k ba

ckbo

ne (

m3 )

VTBP

PRP


1-hop Coverage of Actors with 2 sinks

Our protocol vs. Random positioning with 2 sinks

Performance improves as the number of actors increases

0 2 4 6 8 10 120

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

2.2x 10

6

Number of actors

1−ho

p co

vera

ge o

f net

wor

k ba

ckbo

ne (

m3 )

VTBP

PRP


Average maximum and minimum weight values

Higher average weight values mean small hop-counts, so bettersharing of the sensor nodes

2 3 4 5 6 7 80

1

2

3

4

5

6

7

8

Number of actors

Wei

ght v

alue

Max. weight

Min. weight


Cardinality of actors

Fluctuation in cardinality reduces as number of actors increases

2 4 6 8 10 12 140

5

10

15

Simulation run

Ave

rage

car

dina

lity

LinearTrigonal planarTetrahedralTrigonal bipyramidOctahedralPentagonal bipyramidSquare Antiprismatic


Conclusion

Scalable heuristic algorithm for positioning of actors in aerialWSANs

Simulation results show our protocol provides high connectivityand coverage

Future steps:◮ Increasing scalability and extending our protocol to large networks◮ Exploring other concepts of VSEPR theory and molecular geometry◮ Real-life experiments with UAVs


Actor Positioning Based on Molecular Geometry in …gsolmaz/publications/ICC12... · in Aerial...

Documents

Transcript of Actor Positioning Based on Molecular Geometry in …gsolmaz/publications/ICC12... · in Aerial...