You are on page 1of 18

matthias.handy@etechnik.uni-rostock.

de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Low Energy Adaptive Clustering Hierarchy
with Deterministic Cluster-Head Selection
M. J. Handy, M. Haase, D. Timmermann
Institute of Applied Microelectronics
and Computer Science
University of Rostock
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Outline
Introduction / Motivation
sensor networks, lifetime, communication models
Problem Formulation
cluster-head selection, LEACH algorithm
Contribution
improved CH-selection algorithm, definition of sensor
network lifetime
Simulations
simulation tool, simulation set-up, results
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
- Only the sandbags know
- Useful application of wireless
microsensor networks
- Equip each sandbag with a moisture
sensor
- Collect and evaluate data
How do sensors collaborate efficiently?
Introduction
Where is the spot of leakage?
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Efficient collaboration of sensors means:
- Ensure connectivity
- Efficient role assignment
- Collect only significant data
- Decrease latency
- Save energy
Our Goal: Extend network lifetime
Introduction
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Introduction
How to increase sensor lifetime?
Reduce energy consumption
- Hardware issue
(e.g. circuit design)
- Software issue
‡ Applications / OS
‡ Algorithms
‡ Protocols
Increase energy supply
- Energy density is the problem
- Battery capacity increases only
by 30-50 % in 5 years
- Compare with Mooreµs Law
- Micro-sensors vs. macro-
batteries?
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
§
!
!
N
i
i i amp i elec dt
d k k E E
i i
1
P
I
- Direct transmission
- Multihop transmission
- Clustering
¹
º
¸
©
ª
¨
! !
CH non CH cluster ct
E
l
N
E l lE E
¸ )
toBS Tx Tx
H
i
Tx Rx mt
E E E E E
g g
_
1
2
1
!
§

!
Communication Models
[1]
[1]
[1] Heinzelman, Chandrakasan `01
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Cluster-Based Communication
A Simple Algorithm
The problem: Select j cluster-heads of N nodes without
communication among the nodes
The simplest solution:
- Each node determines a random number x between 0 and 1
- If x < j / N node becomes cluster-head
...itµs good, but:
Cluster-heads dissipate much more energy than non cluster-heads!
How to distribute energy consumption?
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
LEACH Communication Protocol
Low-Energy Adaptive Clustering Hierarchy
- Cluster-based communication
protocol for sensor networks,
developed at MIT
- Adaptive, self-configuring
cluster formation
- The operation of LEACH is divided into rounds
- During each round a different set of nodes are cluster-heads
- Each node n determines a random number x between 0 and 1
- If x < T(n) node becomes cluster-head for current round
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Cluster-Head Selection
LEACH Algorithm
¸ ) G n
P
r P
P
n T V
¹
º
¸
©
ª
¨

!
1
mod 1
¸ ) G n n T ‘ V ! 0
P = cluster-head
probability (j/N)
r = number of the
current round
G = set of nodes not
been cluster-heads
in the last 1/P rounds
Every node becomes cluster-head exactly once within 1/P rounds
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Cluster-Head Selection
LEACH Algorithm
¸ ) G n
P
r P
P
n T V
¹
º
¸
©
ª
¨

!
1
mod 1
¸ ) G n n T ‘ V ! 0
P = cluster-head
probability (j/N)
r = number of the
current round
G = set of nodes not
been cluster-heads
in the last 1/P rounds
Every node becomes cluster-head exactly once within 1/P rounds
Drawback: Selection of cluster-heads is completely
stochastic!
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Cluster-Head Selection, Our Approach I
¸ )
x ma n
current n
E
E
P
r P
P
n T
_
_
1
mod 1
¹
º
¸
©
ª
¨

!
E
n_current
= current energy of node n
E
n_max
= initial energy of node n
Simulations showed:
+ longer network lifetime
- After a certain number of rounds the network is stuck, although there
are still nodes alive
- The reason: T(n) is too low since the remaining nodes have very low
energy level
Basic Idea: Include the remaining energy level
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Idea: Increase T(n) when network is stuck
¸ )
¦
¦
|
¦

¸

¹
¹
º
¸
©
©
ª
¨

¹
º
¸
©
ª
¨

¹
º
¸
©
ª
¨

!
max _
_
_
_
1
1
1
mod 1
n
current n
s
x ma n
current n
E
E
P
div r
E
E
P
r P
P
n T
r
s
= number of rounds a node has not been cluster-head
(reset to 0 when a node becomes cluster-head)
- T(n) is increased when the network is stuck
- Possible deadlock of the network is solved
Significant longer network lifetime
Cluster-Head Selection, Our Approach II
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Lifetime of Microsensor Networks
Introducing 3 New Metrics
First Node Dies (FND)
- Network quality decreases considerably
as soon as one node dies
Half of the Nodes Alive (HNA)
- The loss of a single or few nodes does
not diminish the QOS of the network
Last Node Dies (LND)
- Estimated value for overall lifetime of the
network
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Simulations
Simulation Tool
- YANASim (Yet Another Network
Analyzing and Simulation Tool)
- Simulates energy consumption of
microsensor networks
- Uses Clustering, Multihop and Direct
Transmission
- Visualisation of simulation results
- Platform independent (Java)
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Simulations
Energy Model
Transmit
Electronics
Receive
Electronics
Tx Amplifier
k bit packet
k bit packet
E
Tx
(d)
E
Rx
E
elec
* k e
amp
* k * d
2
E
elec
* k
d
¸ )
P
I kd k E d k E
amp elec Tx
! ,
¸ ) k E k E
elec Rx
!
Transmit:
Receive:
k = message length
d = distance
Ȝ = path-loss index
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Simulations
Simulation Results (1)
574
746
1104
1337
0
500
1000
1500
LEACH Improved CHS
L
i
f
e
t
i
m
e

(
R
o
u
n
d
s
)
FND
HNA
Simulation Setup:
Nodes: 200
Area: 200m*200m
Base Station Pos.: (100,300)m
Initial Energy / Node: 1 J
Message Length: 200 bit
CH-Probability: 0.05
Path-Loss (intra-cluster): 2
Path-Loss (to BS): 2.5
30 % longer lifetime for
FND, 20 % for HNA
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Simulations
Simulation Results (2)
107
134
251
297
0
100
200
300
400
LEACH Improved CHS
L
i
f
e
t
i
m
e

(
R
o
u
n
d
s
)
FND
HNA
Simulation Setup:
Nodes: 200
Area: 200m*200m
Base Station Pos.: (100,500)m
Initial Energy / Node: 1 J
Message Length: 200 bit
CH-Probability: 0.05
Path-Loss (intra-cluster): 2
Path-Loss (to BS): 2.5
25 % longer lifetime for
FND, 18 % for HNA
matthias.handy@etechnik.uni-rostock.de
http://www-md.e-technik.uni-rostock.de/
University of Rostock
Applied Microelectronics and Computer Science
Dept. of Electrical Engineering and Information Technology
Contribution / Conclusions
- Improvement of LEACHµs cluster-head selection algorithm
- 30 % increase of lifetime of sensor networks
- Only local information is necessary for cluster-head selection
- Communication with the base station or an arbiter node is not
necessary
- Three new lifetime metrics FNA, HNA, and LND
- Use of metrics depends on application.

de http://www-md. lifetime.uni-rostock.uni-rostock. simulation set-up.Outline  Introduction / Motivation  sensor networks. results University of Rostock Applied Microelectronics and Computer Science Dept.de/ .handy@etechnik. LEACH algorithm  Contribution  improved CH-selection algorithm. definition of sensor network lifetime  Simulations  simulation tool.e-technik. communication models  Problem Formulation  cluster-head selection. of Electrical Engineering and Information Technology matthias.

Useful application of wireless microsensor networks .Introduction Where is the spot of leakage? .uni-rostock.handy@etechnik.de http://www-md.de/ . of Electrical Engineering and Information Technology matthias.Equip each sandbag with a moisture sensor .uni-rostock.e-technik.Collect and evaluate data How do sensors collaborate efficiently? University of Rostock Applied Microelectronics and Computer Science Dept.Only the sandbags know .

de/ .handy@etechnik.uni-rostock.uni-rostock.Introduction Efficient collaboration of sensors means: .de http://www-md.Decrease latency .Collect only significant data .e-technik.Save energy Our Goal: Extend network lifetime University of Rostock Applied Microelectronics and Computer Science Dept. of Electrical Engineering and Information Technology matthias.Efficient role assignment .Ensure connectivity .

e-technik.Micro-sensors vs.Software issue ‡ Applications / OS ‡ Algorithms ‡ Protocols University of Rostock Applied Microelectronics and Computer Science Dept.Introduction How to increase sensor lifetime? Increase energy supply .Energy density is the problem . circuit design) .handy@etechnik.de http://www-md.Hardware issue (e. macrobatteries? Reduce energy consumption .uni-rostock.Battery capacity increases only by 30-50 % in 5 years .uni-rostock.g.Compare with Mooreµs Law . of Electrical Engineering and Information Technology matthias.de/ .

Multihop transmission H 1 i !2 Emt ! § E Rxg  ETxg  ETx1  ETx _ toBS .Clustering .Communication Models .Direct transmission Edt ! § Eeleci ki  I ampi ki d i i !1 N P [1] .

uni-rostock.uni-rostock.e-technik.de/ . of Electrical Engineering and Information Technology matthias. Ect ! lEcluster N ¸ ¨ ! l © ECH  Enon CH ¹ l º ª [1] [1] Heinzelman. Chandrakasan `01 University of Rostock Applied Microelectronics and Computer Science Dept.de http://www-md.handy@etechnik.

.de/ node becomes cluster-head .itµs good. of Electrical Engineering and Information Technology matthias.e-technik.handy@etechnik.Each node determines a random number x between 0 and 1 .uni-rostock.de http://www-md.If x < j / N .Cluster-Based Communication A Simple Algorithm The problem: Select j cluster-heads of N nodes without communication among the nodes The simplest solution: .uni-rostock.. but: Cluster-heads dissipate much more energy than non cluster-heads! How to distribute energy consumption? University of Rostock Applied Microelectronics and Computer Science Dept.

During each round a different set of nodes are cluster-heads .de http://www-md.uni-rostock.de/ University of Rostock Applied Microelectronics and Computer Science Dept.e-technik.The operation of LEACH is divided into rounds . developed at MIT .uni-rostock.If x < T(n) node becomes cluster-head for current round matthias.Adaptive.handy@etechnik. self-configuring cluster formation . of Electrical Engineering and Information Technology .Cluster-based communication protocol for sensor networks.Each node n determines a random number x between 0 and 1 .LEACH Communication Protocol Low-Energy Adaptive Clustering Hierarchy .

Cluster-Head Selection LEACH Algorithm T .

n ! 0 P 1¸ ¨ 1  P© r mod ¹ Pº ª n ‘ G P= r= cluster-head probability (j/N) number of the current round set of nodes not been cluster-heads in the last 1/P rounds T .

handy@etechnik.uni-rostock.e-technik.uni-rostock.de/ .n ! n  G G= Every node becomes cluster-head exactly once within 1/P rounds University of Rostock Applied Microelectronics and Computer Science Dept. of Electrical Engineering and Information Technology matthias.de http://www-md.

Cluster-Head Selection LEACH Algorithm T .

n ! 0 P 1¸ ¨ 1  P© r mod ¹ Pº ª n ‘ G P= r= cluster-head probability (j/N) number of the current round set of nodes not been cluster-heads in the last 1/P rounds T .

n ! n  G G= Every node becomes cluster-head exactly once within 1/P rounds Drawback: Selection of cluster-heads is completely stochastic! University of Rostock Applied Microelectronics and Computer Science Dept.e-technik.de/ .handy@etechnik.uni-rostock.uni-rostock.de http://www-md. of Electrical Engineering and Information Technology matthias.

Cluster-Head Selection. Our Approach I Basic Idea: Include the remaining energy level T .

handy@etechnik.de/ .n ! P En _ current 1¸ E ¨ 1  P© r mod ¹ n _ ma x Pº ª En_current = current energy of node n En_max = initial energy of node n Simulations showed: + longer network lifetime .The reason: T(n) is too low since the remaining nodes have very low energy level University of Rostock Applied Microelectronics and Computer Science Dept.de http://www-md.uni-rostock.After a certain number of rounds the network is stuck. of Electrical Engineering and Information Technology matthias.e-technik.uni-rostock. although there are still nodes alive .

Cluster-Head Selection. Our Approach II Idea: Increase T(n) when network is stuck « En _ current ¨ 1 ¸¨ En _ current  © rs div ¹©1  T .

T(n) is increased when the network is stuck .uni-rostock. of Electrical Engineering and Information Technology matthias.uni-rostock.de http://www-md.handy@etechnik.n ! ¬ 1 ¸ ¬ En _ ma x ª P º© En _ max ¨ ª 1  P© r mod ¹ ­ Pº ª P ¸» ¹¼ ¹¼ º½ rs = number of rounds a node has not been cluster-head (reset to 0 when a node becomes cluster-head) .de/ .Possible deadlock of the network is solved Significant longer network lifetime University of Rostock Applied Microelectronics and Computer Science Dept.e-technik.

of Electrical Engineering and Information Technology matthias.Lifetime of Microsensor Networks Introducing 3 New Metrics First Node Dies (FND) Network quality decreases considerably as soon as one node dies Half of the Nodes Alive (HNA) The loss of a single or few nodes does not diminish the QOS of the network Last Node Dies (LND) Estimated value for overall lifetime of the network University of Rostock Applied Microelectronics and Computer Science Dept.e-technik.uni-rostock.de http://www-md.handy@etechnik.de/ .uni-rostock.

handy@etechnik.uni-rostock.Platform independent (Java) University of Rostock Applied Microelectronics and Computer Science Dept.de/ .YANASim (Yet Another Network Analyzing and Simulation Tool) .Uses Clustering.Simulates energy consumption of microsensor networks . of Electrical Engineering and Information Technology matthias.e-technik. Multihop and Direct Transmission .de http://www-md.Visualisation of simulation results .uni-rostock.Simulations Simulation Tool .

Simulations Energy Model Transmit: ETx .

k . d ! Eelec k  I amp kd Receive: P k bit packet ETx(d) Transmit Electronics Eelec* k Tx Amplifier eamp* k * d2 ERx k bit packet Receive Electronics Eelec* k d E Rx .

e-technik.k ! Eelec k k = message length d = distance = path-loss index University of Rostock Applied Microelectronics and Computer Science Dept.handy@etechnik.uni-rostock.de http://www-md. of Electrical Engineering and Information Technology matthias.de/ .uni-rostock.

: Initial Energy / Node: Message Length: CH-Probability: 200 200m*200m (100.Simulations Simulation Results (1) Simulation Setup: 1500 Lifetime (Rounds) Nodes: Area: Base Station Pos. of Electrical Engineering and Information Technology 2.uni-rostock.300)m 1J 200 bit 1337 1104 1000 746 574 500 FND HNA 0 0. 20 % for HNA matthias.de/ .e-technik.uni-rostock.handy@etechnik.5 30 % longer lifetime for FND.05 LEACH Improved CHS Path-Loss (intra-cluster): 2 Path-Loss (to BS): University of Rostock Applied Microelectronics and Computer Science Dept.de http://www-md.

Simulations Simulation Results (2) Simulation Setup: Nodes: Area: Base Station Pos.: Initial Energy / Node: Message Length: CH-Probability: 200 Lifetime (Rounds) 300 251 200 134 107 100 FND HNA 400 297 200m*200m (100.e-technik.uni-rostock.500)m 1J 200 bit 0 0.handy@etechnik.de/ .de http://www-md.05 LEACH Improved CHS Path-Loss (intra-cluster): 2 Path-Loss (to BS): University of Rostock Applied Microelectronics and Computer Science Dept.5 25 % longer lifetime for FND. of Electrical Engineering and Information Technology 2.uni-rostock. 18 % for HNA matthias.

of Electrical Engineering and Information Technology matthias.Contribution / Conclusions .30 % increase of lifetime of sensor networks .handy@etechnik.Communication with the base station or an arbiter node is not necessary . University of Rostock Applied Microelectronics and Computer Science Dept.uni-rostock.de/ .e-technik.Use of metrics depends on application.Only local information is necessary for cluster-head selection .uni-rostock.de http://www-md. and LND . HNA.Improvement of LEACHµs cluster-head selection algorithm .Three new lifetime metrics FNA.