On the Coverage Problem in Videobased Wireless Sensor Networks
Stanislava Soro
Wendi Heinzelman
University of Rochester
Outline






Motivation
Problem statement
Overview of DAPR
DAPR in video-based WSNs
Simulation results
Conclusions
10/3/2005
BASENETS'05
Motivation

Telepresence application for VWSN



10/3/2005
enables user to experience being fully present at
a physically remote location
network consists of wireless nodes equipped
with very low-power cameras
user can navigate and virtually move around in
the monitored space
BASENETS'05
Motivation (II)

Distinct features of video-based WSN
over traditional WSN



10/3/2005
Very large amount of highly correlated data
Capturing images of objects that are not
necessarily in camera’s vicinity
Sensing range is replaced with FoV (field of
view)
BASENETS'05
Problem of interest…

Coverage preservation in WSNs:


PEAS, DAPR, CCP….
How do already existing coverage protocols
for WSNs behave in video-based WSNs?



10/3/2005
We assume floorplan monitoring – monitoring of
scene in one plane
Each point of monitored area should be covered
by at least one camera
We analyze how an application-aware routing
protocol (DAPR) behaves in this design space
BASENETS'05
Overview of DAPR in WSN

DAPR-Distributed Activation based on
Predetermined Routes
Coverage preserving protocol that avoids the data routing through
critical nodes

Proposes application-aware approach – each node’s importance for
sensing application is evaluated
C(Sj) – area monitored by sensor Sj
Monitored area is divided into grid, where the center of each grid cell is
given as (x,y)
Total energy for monitoring location (x,y):
(x,y)




Etotal ( x, y) 
10/3/2005
 E (S
Sj:( x , y )C ( Sj)
j
BASENETS'05
)
Overview of DAPR in WSN (II)
S3 E(S3)=5
S4
E(S4)=10
C
E
D
G
B
F
A
Eto ta l ( A) 
Eto tal ( B ) 
S1
E(S1)=1
1
1

1  10 11
1
1

1  5  10 16
Etotal ( E ) 
S2 E(S2)=2
Application cost of node S1
10/3/2005
1
1

1  2  10 13
Eto ta l (C ) 
Eto tal ( D ) 
BASENETS'05
1
1

1 2 3
1
1

1 5 6
Etotal ( F ) 
1
1

1 2  5 8
Etota l (G ) 
1
1

1  2  5  10 18
Overview of DAPR in WSN (III)

Application cost:
1
C AA ( S j )  max(
) ( x, y ) C ( S j )
Etotal ( x, y )

Link cost between two nodes:
Clink ( S i , S j )  C AA ( S i )  Etx  C AA ( S j )  E rx

Cost of a route from node to sink:
Croute(S src ) 
10/3/2005
C
(Si , S j )
link
( Si ,S j )p ( Ssrc,Sdst )
BASENETS'05
DAPR in camera-based WSNs

Two planes


Cameras’ plane: location of point given as (x,y)
Cameras’ FoV plane: location of point given as (xc,yc)
10/3/2005
BASENETS'05
DAPR in camera-based WSNs (II)

Every location (xc,yc) on monitoring plane
characterized by total energy:
Eto tal ( x, y ) 

 E (S
S j:( xc , yc )C ( S j)
)
Final application cost:
C AA ( S j )  max(

j
1
), ( xc , yc )  C ( S j )
Etotal ( xc , yc )
Total routing cost for every camera:
Croute( S src ) 
10/3/2005
C
(Si , S j )
link
( Si ,S j )p ( Ssrc,Sdst )
BASENETS'05
Traditional energy-aware routing

Willingness of every node to route data:
1
C EA ( S j ) 
E (S j )

This cost does not consider the importance
of a node for sensing application
10/3/2005
BASENETS'05
105
105
100
100
Coverage (%)
Coverage (%)
Comparison of application-aware
routing in WSN and video-based WSN
95
90
95
90
Cea(si)
Caa(si)
85
8
8.5
9
Cea(si)
Caa(si)
9.5
10
10.5
Time
11
11.5
x 10
4
85
0.8
0.9
0.95
1
1.05
1.1
Time
Traditional wireless sensor
network
10/3/2005
0.85
Video-based wireless
sensor network
BASENETS'05
1.15
1.2
1.25
x 10
5
Application-aware routing in wireless
sensor networks


Requested part of the scene determines the
locations of all potentially active sensor nodes
The application cost tells us


how redundantly the node is covered
how important node is as a router
BS
10/3/2005
BASENETS'05
Application-aware routing in videobased WSNs


Mismatch between cameras’ physical positions and cameras’
FoV
Here, the application cost evaluates the node:
only from the coverage perspective

but NOT from the routing perspective
Example: a node can be well covered (small application cost), but located in
scarcely deployed area – makes it important as a router

BS
network’s
plane
scene plane
10/3/2005
BASENETS'05
Application-aware routing in videobased WSN (II)

Hotspot problem appears more easily



Potentially active nodes can be far from each other
Select to be active a node with smallest cumulative path
cost – usually node closest to the base station
Energy-aware cost outperforms application-aware
cost


10/3/2005
Balanced energy spent among the nodes – prolongs the
lifetime of each node
The loss of nodes is more uniform over the area
BASENETS'05
Combined application and routing cost

Every camera node validated through two separate
cost functions
105
 C EA ( S j )  C AA ( S j )
( xc , y c )  C ( S j )
1
1

 max
E (S j )
Etotal ( xc , yc )
100
Coverage (%)
Ctotal ( S j )
95
90
85
0.8
Cea(si)
Caa(si)
Ctotal(si)
0.85
0.9
0.95
1
1.05
Time
10/3/2005
BASENETS'05
1.1
1.15
1.2
1.25
x 10
5
Combined application and routing cost

Reduces the energy consumption,
compared to application-aware
routing
With a change in number of nodes, the
same relation between three protocols
persist

CEA(Sj)
Average
power/path
(mW)
0.1091
CAA(Sj)
0.1251
Ctotal(Sj)
0.1121
Time until 95% of the network is covered
18
x 10
16
4
Cea(si)
Caa(si)
Ctotal(si)
14
12
10
8
6
4
2
0
10/3/2005
BASENETS'05
150
100
Number of camera-nodes in the network
200
Conclusions



Application-aware routing protocol gives different
results in traditional and video-based WSNs
Found that coverage and routing problem exist as two
separate problems in video-based WSNs
Further study of this problem





Explore further combined cost function
Explore how other coverage preserving protocols behaves in
video WSNs
Three dimensional coverage problem
Consider collaboration of cameras
Consider the ability of cameras to capture image with different
resolution
10/3/2005
BASENETS'05