Maximizing Network Lifetime

waralligatorΚινητά – Ασύρματες Τεχνολογίες

21 Νοε 2013 (πριν από 3 χρόνια και 8 μήνες)

76 εμφανίσεις

Maximizing Network Lifetime

via 3G Gateway Assignment

in Dual
-
Radio Sensor Networks



LCN 2012, 10/24/2012


Cisco Systems
:
Jaein

Jeong

Australian Nat’l
Univ
:
Xu

Xu
,
Weifa

Liang


CSIRO: Tim
Wark

Third party
network

Introduction

A Remote Monitoring Scenario


Deployed far away from the monitoring center


Network Model: Dual
-
Radio


Goal: Maximize Network Lifetime

Sensor network

Sensor

Gateway

Monitoring Center

IEEE 802.15.4 link

3G link

Base station

(Low Power

, 3G)

2

Introduction

Challenges


Explore main components of energy cons.


For gateways


For slave nodes


Identify gateways among all deployed sensors


m gateways


Network lifetime maximized


Route data to m gateways energy
-
efficiently


Throughput requirement


Delay requirement

3

Organization

1.
Modeling


Energy consumption


Network lifetime

2.
Heuristics


Establish routing trees


Determine network lifetime

3.
Performance Evaluation

4.
Related Work

4

1. Modeling

System Parameters

Parameter

Value



5

Parameter

Value

Graph G(V, E)

V

Set of sensor nodes (N = |V|)

E

Set of links between sensors (M = |E|)

r
s

Data generation rate

Location of
sesnors



Parameter

Value

Graph G(V, E)

V

Set of sensor nodes (N = |V|)

E

Set of links between sensors (M = |E|)

r
s

Data generation rate

Location of
sesnors

Gateway /
Slaves

Gateway

IEEE 802.15.4 and 3G radios

Slave nodes

IEEE 802.15.4 radio

m

Number of gateways

Parameter

Value

Graph G(V, E)

V

Set of sensor nodes (N = |V|)

E

Set of links between sensors (M = |E|)

r
s

Data generation rate

Location of
sesnors

Gateway /
Slaves

Gateway

IEEE 802.15.4 and 3G radios

Slave nodes

IEEE 802.15.4 radio

m

Number of gateways

QoS

metrics

α

Network throughput

D

Delivery

delay

1. Modeling

Energy Cost

Flash

memory

buffer

3G radio

802.15.4
radio

MCU



o
s
G
buf
n
g
E
D
v
d
r
P
P
P
v
ec







/
1
)
(
)
(
)
(
3
)
(
)
(
v
d
r
P
v
ec
s
n
s









otherwise
0
gateway

a

is


if
)
(
slave

active
an

is


if
)
(
)
(
v
v
ec
v
v
ec
v
ec
g
s
s
Param

Desc

P
n

Pwr

by radio,

MCU

P
3G

Pwr

by 3G

P
buf

Pwr

by buffering

d(v)

#
-
descendants

E
o

Overhead for sync

6

1. Modeling

Network Lifetime


Time before the base station is no longer able to
receive data from
α

percentage of sensors


Round

1

τ

Network Lifetime:
L

Round

2

τ

Round

r

τ

Round

R

τ

Round

R+1

τ‘

(<= τ)

7

1. Modeling

Network Lifetime

gateways

of
set
:
GW
nodes

ing
transmitt
of
set
:
)
(
'
g
V
r
nodes

active

of
set
:
)
(
'
g
V
r
GW
g


R
r
g
V
v
v
e
v
e
r
GW
g
c
r





1
),
(
'
,
)
(
)
(


v
v
e
r

of
energy

residual
:
)
(
v
v
e
c

of
n
consumptio
energy
:
)
(
v
v
e
v
e
c
r

of

lifetime

residual
:
)
(
)
(










)
(
'
,
)
(
)
(
min
'
1
g
V
v
v
e
v
e
R
GW
g
c
r




1
1

,

)
(
'







R
r
N
g
V
GW
g
r

1

τ

2

τ

r

τ

R

τ

R+1

τ‘


8

1. Modeling

Problem Definition


Periodic assignment of
gateways


Identify m gateways


Selecting nodes to send
data to these gateways


Route data from these
nodes to gateways

1

τ

2

τ

r

τ

R

τ

R+1

τ‘


9

Organization

1.
Modeling


Energy consumption


Network lifetime

2.
Heuristics


Establish routing trees


Determine network lifetime

3.
Performance Evaluation

4.
Related Work

10

2. Heuristic

Establishing the Routing Forest


Routing trees should span
at least
α

*N nodes.


1)
Identify the smallest set
of active nodes

2)
Partition the active
nodes into m subsets

3)
Find the routing tree

1

2

3

4

5

6

7

8

9

10

11

1

2

3

4

5

6

7

8

9

10

11

1

2

3

4

5

6

7

8

9

10

11

11

2. Heuristic

(1) Identifying Active Nodes


Choose sensors with
high
e
r
(v)



m
-
component
constraint


CC(G[V’]) <= m


Or, some nodes may not
reach a gateway.

v
v
e
r

of
energy

residual
:
)
(
high

low

12

2. Heuristic

(1) Identifying Active Nodes

identified

be

to
nodes

active
:
'
V
)
'
(
)
'
(
)
'
(
)
'
(
1
N
r
j
r
i
r
r
v
e
v
e
v
e
v
e









(v)
e
V
v'
r
by

s
Sort

'
, v
',
v
i
N

1
min

from

nodes

first

Choose


N
α
i


min
components

ed
:#-connect
CC(G[V'])
m
CC(G[V'])

1

2

3

4

5

6

7

8

9

10

11

13

2. Heuristic

(2) Partitioning active nodes into m subsets

1

2

3

4

5

6

7

8

9

10

11

1

2

3

4

5

6

7

8

9

10

11

CC(G[V’])=
m’

m
’<= m

CC(G[V])=m

Partition

G[V’] into G[V]

14

2. Heuristic

(2) Partitioning active nodes into m subsets


F = {S
1
, S
2
, …,
S
m

}
collection of vertex sets.


Select a set with the
largest #
-
vertices,
S
l


Partition
S
l

into S
l1
, S
l2

s.t
.
||S
l1
|
-
||S
l2
|| is
minimized.


Repeat until m’ = m.


1

2

3

4

5

6

7

8

9

10

11

S
l

S
l2

S
l1

15

2. Heuristic

(3) Finding routing tree


max
-
min tree


Find max
-
min tree T
i
(v)
for each connected graph
G
i

and given root v.



The tree T
i

rooted at a
node with the longest
lifetime is selected [9].

1

2

3

4

5

6

7

8

9

10

11

S
l2

S
l1

[9] W. Liang and Y. Liu. On
-
line data gathering for
maximinizing

network lifetime in sensor networks.
IEEE
Trans. on Mobile Computing, 6:2

11,
2007.

16

2. Heuristic

Determining the Network Lifetime


For each tree T
i
, evaluate
l
min

at round r.





If
I
min

>
τ


L = L +
τ


e
r
(v) =
e
r
(v)


τ
*
e
c
(v)


e
delta


If
l
min

<=
τ


τ
’ =
l
min


L = L +
τ



Terminate the loop












m
i
T
V
v
v
e
v
e
l
i
c
r
1

,
)
(

,
)
(
)
(
min
min
1

2

3

4

5

6

7

8

9

10

11

1

τ

2

τ

r

τ

R

τ

R+1

τ‘


17

2. Heuristic

Complexity


O(MN
2
) for N = |V|, M = |E|


Proof

1)
Finding #
-
connected components:

O(M) using BFS or DFS

2)
Partitioning active nodes:

O(N
3
logN) [8]

3)
Building a max
-
min tree rooted at a given node:

O(MN
2
) [9]

4)
For any G(V, E):

M=O(N
2
)

[8] D. R.
Karger

and C. Stein. A new approach to the minimum cut problem.

Journal of the ACM, 43:601

640, 1996.

[9] W. Liang and Y. Liu. On
-
line data gathering for
maximinizing

network lifetime in sensor networks.

IEEE Trans. on Mobile Computing, 6:2

11,
2007.

18

3. Performance Evaluation

Assumptions

Parameters

Values

#
-
Sensors

100


300

Tx

Range

100 m

Initial Energy Cap

200 J

Energy Consumption
Param

CC2420 radio for IEEE 802.15.4 radio

MO6012 radio for 3G radio

NAND flash memory

Data Generation Rate (
r
s
)

1 bit / s

Data Delivery Latency

(D)

1 hr

Network Throughput


Threshold (
α
)

0.7

19

3. Performance Evaluation

Residual Energy over Time


N = 100, m = 5,
τ

= 2 hr


25τ

50τ

175τ

186τ


75τ

100τ

125τ

150τ

20

3. Performance Evaluation

Residual Energy over Time


N = 100, m = 5,
τ

= 2 hr



In the first 75 rounds:

For all, E > 0.5E
init


In the 175
th

round:

44 nodes, E < 0.2E
init


In the last round:

37 nodes, E = 0

Throughput
req

isn’t met

25τ

50τ

175τ

186τ


75τ

100τ

125τ

150τ

21

3. Performance Evaluation

Lifetime over Constraint Parameters


Network throughput

α
:
steadily decreases, then
rapidly falls


#
-
nodes

N
: decreases
more with smaller
α


Duration of round

τ
: first
increases, then decreases


#
-
gateways

m
: first
increases, then decreases


Delivery delay

D
:
increases







Vary
α

from 0.3 to 1



N = 100, m = 5,
τ

= 2 hr



Vary N from 100 to 300



N = 100, m = 5,
τ

= 2 hr



Vary
τ

from 1 hr to 10 hr



m = 5



Vary m from 2 to 20



τ

= 2 hr



Vary D to 10, 20, 30, 60 and 120
min



m = 5,
τ

= 2 hr

22

3. Performance Evaluation

Three Algorithms

Algorithm

Gateway Selection

Time

Gateway Selection

Criteria

StaticAlg

LEACH [7]




DynamicAlg

Algorithm

Gateway Selection

Time

Gateway Selection

Criteria

StaticAlg

Once in lifetime

LEACH [7]

Each round




DynamicAlg

Each round

Algorithm

Gateway Selection

Time

Gateway Selection

Criteria

StaticAlg

Once in lifetime

Random

LEACH [7]

Each round

Random &

Time spent as gateways

in previous rounds

DynamicAlg

Each round

Residual energy

23

[7] W. R.
Heinzelman
, A.
Chandrakasan
, and H.
Balakrishnan
. Energy efficient communication protocol for wireless
microsensor

networks.

Proc. of HICSS. IEEE, 2000.

3. Performance Evaluation

Three Algorithms and Lifetime Delivered


In general, Dynamic >
LEACH > Static


Superiority of Dynamic
and LEACH over Static


More balanced energy
consumption


Advantages of Dynamic
over LEACH


More efficient gateway
identification


More advanced routing
forest establishment


0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
1,400,000
1,600,000
τ=1
hr

τ=2
hr

τ=3
hr

τ=4
hr

τ=5
hr

Axis Title

Network lifetime to
τ

DynamicAlg
LEACH
StaticAlg
0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
1,400,000
1,600,000
1,800,000
m=4
m=6
m=8
m=10
m=12
Network Lifetime

Network Lifetime to m

DynamicAlg
LEACH
StaticAlg
24

4. Related Work

[6] J. Gummeson, D. Ganesan, M. D. Corner, and P. Shenoy. An adaptive link layer for heterogeneous multi
-
radio mobile sensor net
works.

IEEE Journal on Selected Areas in Communications, 28:1094

1104, 2010.

[10] D. Lymberopoulos, N. B. Priyantha, M. Goraczko, and F. Zhao. Towards efficient design of multi
-
radio platforms for wireless

sensor
networks.
Proc. of IPSN. IEEE, 2008.

[12] C. Sengul, M. Bakht, A. F. Harris, T. Abdelzaher, and R. Kravets. Improving energy conservation using bulk transmission
ove
r high
-
power
radios in sensor networks.
Proc. of ICDCS. IEEE, 2008.

[13] T. Stathopoulos, M. Lukac, D. Mclntire, J. Heidemann, D. Estrin, and W. J. Kaiser. End
-
to
-
end routing for dual
-
radio sensor

networks.

Proc. of INFOCOM. IEEE, 2007.

Hierarchical Power

Management

Ours

Assumptions



Methods




Examples

Hierarchical Power

Management

Ours

Assumptions



Use both high

low BW radios



Optimize their use



Sensornet

within the network



3G only for remote data

Methods




Examples

Hierarchical Power

Management

Ours

Assumptions



Use both high

low BW radios



Optimize their use



Sensornet

within the network



3G only for remote data

Methods



䵩湩M楺攠瑩t攠f潲o桩杨 B圠
r慤楯a睨楬e

汯眠w圠r慤楯a楳
慬aay猠潮s



Sch敤u汥⁧ateways

t漠och楥v攠
浡砠汩fe瑩浥

Ex慭灬敳

Hierarchical Power

Management

Ours

Assumptions



Use both high

low BW radios



Optimize their use



Sensornet

within the network



3G only for remote data

Methods



䵩湩M楺攠瑩t攠f潲o桩杨 B圠
r慤楯a睨楬e

汯眠w圠r慤楯a楳
慬aay猠潮s



Sch敤u汥⁧ateways

t漠och楥v攠
浡砠汩fe瑩浥

Ex慭灬敳

嬶Ⱐ㄰ⰠㄲⰠㄳ]

25

4. Related Work

[7] W. R.
Heinzelman
, A.
Chandrakasan
, and H.
Balakrishnan
. Energy efficient communication protocol for wireless
microsensor

networks.

Proc. of HICSS. IEEE, 2000.

LEACH

Ours

Goals




Gateway

Selection

Energy
-
aware

Routing




LEACH

Ours

Goals



Maximizes lifetime

by
perio
-
dically

changing the set of GWs



Maximizes lifetime

by
perio
-
dically

changing the set of GWs

Gateway

Selection

Energy
-
aware

Routing




LEACH

Ours

Goals



Maximizes lifetime

by
perio
-
dically

changing the set of GWs



Maximizes lifetime

by
perio
-
dically

changing the set of GWs

Gateway

Selection



剡R摯洠☠呩浥 獰敮琠慳⁇W猠
楮⁰ie癩潵猠r潵o摳



R敳e摵慬a敮敲杹.

E湥n杹
-
aw慲e

Routing






LEACH

Ours

Goals



Maximizes lifetime

by
perio
-
dically

changing the set of GWs



Maximizes lifetime

by
perio
-
dically

changing the set of GWs

Gateway

Selection



剡R摯洠☠呩浥 獰敮琠慳⁇W猠
楮⁰ie癩潵猠r潵o摳



R敳e摵慬a敮敲杹.

E湥n杹
-
aw慲e

Routing



Total distance minimization in
LEACH does not necessarily lead
to the min energy cons.



䑡t愠a潵瑩湧 楳⁤敳e杮敤et漠
balance the energy cons.

26

Conclusion


Maximizing lifetime of a dual
-
radio sensor
network.


Proposed a model for energy cons and lifetime.


Proposed heuristics that maximizes lifetime


Identifies gateways


Finds the data routing structure


Experiment Results


Our heuristic outperforms other cluster methods.


Future works


Distributed algorithm


Experiments with real energy consumption

27

Backup Slides

28

3. Performance Evaluation

Lifetime over Throughput Threshold
α


Lifetime (L) steadily falls
down before
α

= 0.6,
rapidly falls after that.


For
α

<= 0.6


Additional nodes are
used for connected
components.


For
α

> 0.6


Additional nodes
increase required active
nodes and energy cons.







Vary
α

from 0.3 to 1



N = 100, m = 5,
τ

= 2 hr

29

3. Performance Evaluation

Lifetime of #
-
Nodes N


L starts to drop from a
smaller
α

as N gets larger.


With higher node density,


Smaller # of required extra
nodes for m
-
component


The more distinct impact
of
α
on lifetime.


Higher traffic cause shorter
lifetime.




Vary N from 100 to 300



N = 100, m = 5,
τ

= 2 hr

30

3. Performance Evaluation

Lifetime over Duration of Period
τ


Generally, the larger
τ
,
the shorter L.


Frequent identification
balances energy better.


From 1 to 2
-
3hr, L
slightly increases as
τ
increases.


Too frequent routing


With fixed
τ
, L gets
smaller as N increases.




Vary
τ

from 1 hr to 10 hr



m = 5

31

3. Performance Evaluation

Lifetime over #
-
gateways m


Lifetime first increases
and then decreases.


Before a turning point:

Better energy balancing


After passing the point:

More energy cons on 3G




Vary m from 2 to 20



τ

= 2 hr

32

3. Performance Evaluation

Lifetime over Delivery Delay D


A smaller value of D leads to
a shorter L


Frequent on
-
and
-
off
switching of the 3G radios
results in energy overheads




Vary D to 10, 20, 30, 60 and 120
min



m = 5,
τ

= 2 hr

33