SlideShare a Scribd company logo
1 of 41
Ehsan Ramezani
MohammadAmin Amjadi
Shahid bahonar university
May 2013
11/19/2015
 Energy conservation probably constitutes the most
important challenge in the design of wireless sensor
networks (WSNs).
 Another issue in WSN design is the connectivity of
the network according to some specific
communication protocol (Cluster-based architecture).
11/19/2015
 The purpose of the sensor network, which is the
collection and possibly the management of measured
data for some particular application, must not be
neglected.
 Most algorithms that lead to optimal topologies of
WSNs towards power conservation, do not take into
account the principles , characteristics and
requirements of application-specific WSNs.
11/19/2015
 One of the most powerful heuristics that could be
applied to our multi-objective optimization problem
is based on “Genetic Algorithms” (GAs).
 an integrated GA approach , both in the direction of
degrees of freedom of network characteristics and of
application-specific requirements represented in the
performance metric of the GA is considered.
11/19/2015
 optimization problem → minimization of the
energy-related parameters and the maximization of
sensing points’ uniformity, subject to the connectivity
constraints and the spatial density requirement.
 ↓
 single objective function
 ↓
 minimization
11/19/2015
 he original GA-based algorithm was applied
dynamically to obtain a dynamic sequence of
operating modes for each sensor.
 i.e. a sequence of WSN designs, which leads to
maximization of network lifetime in terms of number
of data collection (measuring) cycles.
11/19/2015
 Memetic algorithms introduce local search techniques
at specific parts of a GA optimization process, with a
goal to improve its performance.
 In this work, we develop and parameterize a
memetic algorithm (MA) which hybridizes the GA
system and the goal being to improve its performance
by guiding the population formulation of the GA
towards more intelligent decisions.
11/19/2015
 WSN modeling
 Design parameters
11/19/2015
 A cluster-based network architecture is used where
sensors are partitioned into several clusters.
 All sensors are identical and may be either active or
inactive.
 Sensores three supported signal ranges:
 1)CH sensor that allows the communication with
the remote base station (sink).
 2) HSR sensor
 3)LSR sensor
11/19/2015
 An evolutionary algorithm, requires the proper
definition of some optimization criteria.
 In order for the optimization to be possible, it is
required that these design parameters are explicitly
defined and expressed mathematically.
 Application specific parameters
 Connectivity parameters
 Energy related parameters
11/19/2015
 he satisfaction or not of the demand on uniformity of
measuring points has been taken into consideration
using two design parameters:
 (a) MRD : the spatial mean relative deviation of
sensing points, representing the uniformity of
those points .
 (b) the desired spatial density of measuring points.
11/19/2015
 MRD →minimized ≡ uniformity is maximized
 The spatial density of sensing points has to be as
close as possible to the desired value.
 The entire area of interest was divided into several
overlapping sub-areas.
 Sub-areas are defined by four factors : length and
width(size) , overlapping ratio (ratios in the two
directions).
11/19/2015
 MRD= 𝑖=1
𝑁
|𝑃𝑠𝑖 − 𝑃𝑠| 𝑁 . 𝑃𝑠
 Ps : spatial density of sensing points in total area
 Psi : spatial density of sensing points in sub-area Si
 N : number of overlapping sub-areas into which the
entire area was divided
 Low MRD → high uniformity

11/19/2015
 SDE =
𝑃𝑑 −𝑃𝑠
𝑃𝑑
𝑖𝑓 𝑃𝑠 < 𝑃𝑑
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
 Pd : set to 0.2
11/19/2015
 This set of design parameters includes two factors :
first : that each clusterhead does not have more than
a maximum predefined number of sensors in its
cluster (SCE)
second : that each sensor of the network can
communicate with its clusterhead (SORE)
11/19/2015
 SCE = 𝑖=1
𝑛 𝑓𝑢𝑙𝑙
𝑛𝑖 𝑛 𝑓𝑢𝑙𝑙 𝑖𝑓
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑛 𝑓𝑢𝑙𝑙 > 0
 𝑛 𝑓𝑢𝑙𝑙 : number of clusterheads that have more than 15
active sensors in clusters
 𝑛𝑖 : number of sensors in the ith of those clusters
11/19/2015
 SORE = 𝑛 𝑜𝑢𝑡 𝑛
 𝑛 𝑜𝑢𝑡 : number of active sensors that cannot
communicate with their clusterhead
 n : total number of active sensors in the network
11/19/2015
 This set of design parameters can be divided into two
sub-categories :
OE : operational energy consumption parameter
CE: communicational energy consumption parameter
11/19/2015
 OE = 20 . 𝑛 𝐶𝐻 𝑛 + 2 . 𝑛 𝐻𝑆 𝑛 + 𝑛 𝐿𝑆 𝑛
 𝑛 𝐶𝐻 : operation mode of the sensor is CH
 𝑛 𝐻𝑆 : operation mode of the sensor is HS
 𝑛 𝐿𝑆 : operation mode of the sensor is LS
 corresponding relevance factors for three active
operating modes of the sensors : 20:2:1
11/19/2015
 CE =
𝑖=1
𝑐
𝑗=1
𝑛 𝑖
𝜇. 𝑑𝑗𝑖
𝑘
 c : number of clusters in the network
 𝑛𝑖 : number of sensors in the ith cluster
 𝑑𝑗𝑖:Euclidean distance from sensor j to its clusterhead
 𝜇 & k : constants
11/19/2015
 The maximization of the life duration of the network
depends mainly on the remaining battery capacities
of the sensors.
 𝐵𝐶𝑃 𝑡
=
𝑖=1
𝑛 𝑔𝑟𝑖𝑑
𝑃𝐹𝑖
𝑡
. 1 𝐵 𝐶𝑡
𝑡
− 1 ,t=1,2,…
 𝐵𝐶𝑖
𝑡
= 𝐵𝐶𝑖
𝑡−1
− 𝐵𝑅𝑅𝑖
𝑡−1
11/19/2015
 𝐵𝐶𝑃 𝑡
: Battery Capacity Penalty of the WSN at
measuring cycle t
 𝑛 𝑔𝑟𝑖𝑑 : total number of available sensor nodes
 𝑃𝐹𝑖
𝑡
: Penalty Factor assigned to sensor i
 𝐵𝐶𝑖
𝑡
and 𝐵𝐶𝑖
𝑡−1
: are the Battery Capacities of
sensor i at measuring cycles t and t – 1
 𝐵𝑅𝑅𝑖
𝑡−1
: Battery Reduction Rate
11/19/2015
 Methodology of GA:
 i) encoding mechanism of the problems
phenotypes into genotypes
ii) formulation of an appropriate fitness function
iii) choice of the genetic operators and the
selection mechanism
11/19/2015
 The parameters of each WSN design that needs to be
encoded in the representation scheme of the GA are
the following:
 i) placement of the active sensors of the network
 ii) operation mode of each active sensor(clusterhead
or a ‘‘regular sensor”)
 iii) range of signal (high or low),in ‘‘regular sensor”
 X sensors in the WSN → each string in the GA
population has a length of 2X
11/19/2015
 F= 1 (𝛼1 . 𝑀𝑅𝐷 + 𝛼2. 𝑆𝐷𝐸 + 𝛼3. 𝑆𝐶𝐸 + 𝛼4. 𝑆𝑂𝑅𝐸 +
𝛼5. 𝑂𝐸 + 𝛼6. 𝐶𝐸 + 𝛼7. 𝐵𝐶𝑃)
 𝛼i : weighting coefficients
 The values of these coefficients were determined
based on experience about the importance of each
parameter.
11/19/2015
 The concept of the memetic algorithm is based on the
introduction of some battery level threshold values
for each operating mode of the sensors.
 at each measuring cycle to allow a sensor i to operate
at some specific mode if and only if its battery level
at the time is above the threshold value for that
operating mode.
11/19/2015
 The MA approach is materialized through two
separate processes:
 (i) the local search : battery level of each sensor
against the corresponding threshold value
 (ii) the appropriate update of the threshold values for
each operating mode, according to some specific
reduction scheme.

11/19/2015
 The main part of the MA is the local search that is
performed in the generation of the population of the
original GA.
 local search is performed in each individual of the
GA population .
11/19/2015
 Some initial threshold values of battery levels for
each of the three possible operating modes of the
sensors are defined.
 he intention of these threshold values is to put
specific constraints in the operation modes of each
sensor, throughout the dynamic design of the
network.
 Operating mode of each sensor is checked
11/19/2015
 battery level is below threshold
↓
 operating mode is changed to the lower mode
↓
 until its corresponding threshold value becomes
lower than (or equal to) its battery level
11/19/2015
 Three facts are evident about the proposed
modification:
 First, it is applied to the phenotype of the design
problem and not the genotype of the optimization
process of the GA
 Second, investigation on the appropriate level of
reduction of the operating mode of the modified
sensors resembles some kind of search operation.
 Third, modification always leads towards the
direction of a local improvement , without any
assurance that this would result into a generally better
solution.
11/19/2015
 Threshold values of battery levels for each of the
three possible operating modes of the sensors are
initialized.
 Three parameters of reduction scheme of each
threshold:
initial value
reduction formula
reduction rate
11/19/2015
 (i) geometric reduction:
𝑇 𝑡+1 = (1 − 𝑅𝑅). 𝑇 𝑡
T : any of the three types of threshold
t : some specific measuring cycle
RR : reduction rate parameter
 1t
T

11/19/2015
 (ii) linear reduction:
𝑇 𝑡+1 = 𝑇 𝑡 − 𝑅𝑅
(iii) no update, where thresholds are kept constant
over time.
11/19/2015
 he performance of the MA approach to the dynamic
optimization of WSN designs was compared to that
of the original GA-based system, during 15
consecutive measuring cycles of the WSN.
11/19/2015
11/19/2015
11/19/2015
 The MA system showed considerable improvement in
energy conservation of the network resources over
the already successful performance of the GA system
, while the application-specific characteristics of the
sensor networks were kept close to optimal values.
11/19/2015
 The satisfactory performance of the algorithm during
the dynamic network design process makes it a
valuable tool for design optimization towards
maximization of the life span of WSNs, especially in
cases where satisfaction of some application-specific
requirements is a necessity.
11/19/2015
 In addition, it was shown that appropriate
manipulation of the population of the GA (something
that was introduced by the proposed local search
scheme) can lead to performance improvement.
11/19/2015
 Thanks for listening!!!
11/19/2015

More Related Content

What's hot

Artificial bee colony (abc)
Artificial bee colony (abc)Artificial bee colony (abc)
Artificial bee colony (abc)
quadmemo
 

What's hot (20)

Artificial bee colony (abc)
Artificial bee colony (abc)Artificial bee colony (abc)
Artificial bee colony (abc)
 
Linear regression with gradient descent
Linear regression with gradient descentLinear regression with gradient descent
Linear regression with gradient descent
 
Fuzzy Genetic Algorithm
Fuzzy Genetic AlgorithmFuzzy Genetic Algorithm
Fuzzy Genetic Algorithm
 
Relational Algebra and MapReduce
Relational Algebra and MapReduceRelational Algebra and MapReduce
Relational Algebra and MapReduce
 
Local search algorithms
Local search algorithmsLocal search algorithms
Local search algorithms
 
Local search algorithm
Local search algorithmLocal search algorithm
Local search algorithm
 
Hill climbing algorithm
Hill climbing algorithmHill climbing algorithm
Hill climbing algorithm
 
MACHINE LEARNING - GENETIC ALGORITHM
MACHINE LEARNING - GENETIC ALGORITHMMACHINE LEARNING - GENETIC ALGORITHM
MACHINE LEARNING - GENETIC ALGORITHM
 
Artificial Bee Colony algorithm
Artificial Bee Colony algorithmArtificial Bee Colony algorithm
Artificial Bee Colony algorithm
 
Gradient descent method
Gradient descent methodGradient descent method
Gradient descent method
 
Flowchart of GA
Flowchart of GAFlowchart of GA
Flowchart of GA
 
Vanishing & Exploding Gradients
Vanishing & Exploding GradientsVanishing & Exploding Gradients
Vanishing & Exploding Gradients
 
Genetic programming
Genetic programmingGenetic programming
Genetic programming
 
Lecture 14 Heuristic Search-A star algorithm
Lecture 14 Heuristic Search-A star algorithmLecture 14 Heuristic Search-A star algorithm
Lecture 14 Heuristic Search-A star algorithm
 
Random forest
Random forestRandom forest
Random forest
 
K mean-clustering algorithm
K mean-clustering algorithmK mean-clustering algorithm
K mean-clustering algorithm
 
An introduction to reinforcement learning
An introduction to reinforcement learningAn introduction to reinforcement learning
An introduction to reinforcement learning
 
Differential Evolution Algorithm (DEA)
Differential Evolution Algorithm (DEA) Differential Evolution Algorithm (DEA)
Differential Evolution Algorithm (DEA)
 
Genetic algorithms in Data Mining
Genetic algorithms in Data MiningGenetic algorithms in Data Mining
Genetic algorithms in Data Mining
 
Machine Learning with Decision trees
Machine Learning with Decision treesMachine Learning with Decision trees
Machine Learning with Decision trees
 

Viewers also liked

Memetic search in differential evolution algorithm
Memetic search in differential evolution algorithmMemetic search in differential evolution algorithm
Memetic search in differential evolution algorithm
Dr Sandeep Kumar Poonia
 
Genetic algorithms
Genetic algorithmsGenetic algorithms
Genetic algorithms
zamakhan
 

Viewers also liked (12)

Memetic algorithm
Memetic algorithm Memetic algorithm
Memetic algorithm
 
An improved memetic search in artificial bee colony algorithm
An improved memetic search in artificial bee colony algorithmAn improved memetic search in artificial bee colony algorithm
An improved memetic search in artificial bee colony algorithm
 
An Unorthodox View on Memetic Algorithms
An Unorthodox View on Memetic AlgorithmsAn Unorthodox View on Memetic Algorithms
An Unorthodox View on Memetic Algorithms
 
Memetic search in differential evolution algorithm
Memetic search in differential evolution algorithmMemetic search in differential evolution algorithm
Memetic search in differential evolution algorithm
 
10 Rules of Memetic Marketing
10 Rules of Memetic Marketing10 Rules of Memetic Marketing
10 Rules of Memetic Marketing
 
Introduction to Evolutionary Algorithms
Introduction to Evolutionary AlgorithmsIntroduction to Evolutionary Algorithms
Introduction to Evolutionary Algorithms
 
Genetic Algorithms Made Easy
Genetic Algorithms Made EasyGenetic Algorithms Made Easy
Genetic Algorithms Made Easy
 
Genetic Algorithms
Genetic AlgorithmsGenetic Algorithms
Genetic Algorithms
 
Introduction to Genetic Algorithms
Introduction to Genetic AlgorithmsIntroduction to Genetic Algorithms
Introduction to Genetic Algorithms
 
Genetic algorithms
Genetic algorithmsGenetic algorithms
Genetic algorithms
 
Harmony search algorithm
Harmony search algorithmHarmony search algorithm
Harmony search algorithm
 
Deep Learning - The Past, Present and Future of Artificial Intelligence
Deep Learning - The Past, Present and Future of Artificial IntelligenceDeep Learning - The Past, Present and Future of Artificial Intelligence
Deep Learning - The Past, Present and Future of Artificial Intelligence
 

Similar to memetic algorithm

Overview of Sensors project
Overview of Sensors projectOverview of Sensors project
Overview of Sensors project
Shan Guan
 
Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*
Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*
Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*
Editor IJCATR
 
Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02
Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02
Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02
ratnmani mukesh
 

Similar to memetic algorithm (20)

A Survey on Data Aggregation Cluster based Technique in Wireless Sensor Netwo...
A Survey on Data Aggregation Cluster based Technique in Wireless Sensor Netwo...A Survey on Data Aggregation Cluster based Technique in Wireless Sensor Netwo...
A Survey on Data Aggregation Cluster based Technique in Wireless Sensor Netwo...
 
A Survey of Routing Protocols for Structural Health Monitoring
A Survey of Routing Protocols for Structural Health MonitoringA Survey of Routing Protocols for Structural Health Monitoring
A Survey of Routing Protocols for Structural Health Monitoring
 
Research Inventy : International Journal of Engineering and Science
Research Inventy : International Journal of Engineering and ScienceResearch Inventy : International Journal of Engineering and Science
Research Inventy : International Journal of Engineering and Science
 
COMPARISON OF ENERGY OPTIMIZATION CLUSTERING ALGORITHMS IN WIRELESS SENSOR NE...
COMPARISON OF ENERGY OPTIMIZATION CLUSTERING ALGORITHMS IN WIRELESS SENSOR NE...COMPARISON OF ENERGY OPTIMIZATION CLUSTERING ALGORITHMS IN WIRELESS SENSOR NE...
COMPARISON OF ENERGY OPTIMIZATION CLUSTERING ALGORITHMS IN WIRELESS SENSOR NE...
 
Overview of Sensors project
Overview of Sensors projectOverview of Sensors project
Overview of Sensors project
 
Power balancing optimal selective forwarding
Power balancing optimal selective forwardingPower balancing optimal selective forwarding
Power balancing optimal selective forwarding
 
Energy-Efficient Compressive Data Gathering Utilizing Virtual Multi-Input Mul...
Energy-Efficient Compressive Data Gathering Utilizing Virtual Multi-Input Mul...Energy-Efficient Compressive Data Gathering Utilizing Virtual Multi-Input Mul...
Energy-Efficient Compressive Data Gathering Utilizing Virtual Multi-Input Mul...
 
Capacity planning in cellular network
Capacity planning in cellular networkCapacity planning in cellular network
Capacity planning in cellular network
 
IRJET - Energy Efficient Enhanced K-Means Cluster-Based Routing Protocol for WSN
IRJET - Energy Efficient Enhanced K-Means Cluster-Based Routing Protocol for WSNIRJET - Energy Efficient Enhanced K-Means Cluster-Based Routing Protocol for WSN
IRJET - Energy Efficient Enhanced K-Means Cluster-Based Routing Protocol for WSN
 
A SURVEY ON DIFFERENT TYPES OF CLUSTERING BASED ROUTING PROTOCOLS IN WIRELESS...
A SURVEY ON DIFFERENT TYPES OF CLUSTERING BASED ROUTING PROTOCOLS IN WIRELESS...A SURVEY ON DIFFERENT TYPES OF CLUSTERING BASED ROUTING PROTOCOLS IN WIRELESS...
A SURVEY ON DIFFERENT TYPES OF CLUSTERING BASED ROUTING PROTOCOLS IN WIRELESS...
 
IRJET - A Review on Analysis of Location Management in Mobile Computing
IRJET -  	  A Review on Analysis of Location Management in Mobile ComputingIRJET -  	  A Review on Analysis of Location Management in Mobile Computing
IRJET - A Review on Analysis of Location Management in Mobile Computing
 
Iisrt divya nagaraj (networks)
Iisrt divya nagaraj (networks)Iisrt divya nagaraj (networks)
Iisrt divya nagaraj (networks)
 
IRJET- Voltage Stability, Loadability and Contingency Analysis with Optimal I...
IRJET- Voltage Stability, Loadability and Contingency Analysis with Optimal I...IRJET- Voltage Stability, Loadability and Contingency Analysis with Optimal I...
IRJET- Voltage Stability, Loadability and Contingency Analysis with Optimal I...
 
IRJET- Optimal Strategy for Extraction of Traffic Based on PLC
IRJET- Optimal Strategy for Extraction of Traffic Based on PLCIRJET- Optimal Strategy for Extraction of Traffic Based on PLC
IRJET- Optimal Strategy for Extraction of Traffic Based on PLC
 
Throughput in cooperative wireless networks
Throughput in cooperative wireless networksThroughput in cooperative wireless networks
Throughput in cooperative wireless networks
 
Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*
Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*
Energy-Aware Routing in Wireless Sensor Network Using Modified Bi-Directional A*
 
Development of Effective Crop Monitoring and Management System with Weather R...
Development of Effective Crop Monitoring and Management System with Weather R...Development of Effective Crop Monitoring and Management System with Weather R...
Development of Effective Crop Monitoring and Management System with Weather R...
 
Energy Efficient Clustering Algorithm based on Expectation Maximization for H...
Energy Efficient Clustering Algorithm based on Expectation Maximization for H...Energy Efficient Clustering Algorithm based on Expectation Maximization for H...
Energy Efficient Clustering Algorithm based on Expectation Maximization for H...
 
A Topology Control Algorithm Taking into Account Energy and Quality of Transm...
A Topology Control Algorithm Taking into Account Energy and Quality of Transm...A Topology Control Algorithm Taking into Account Energy and Quality of Transm...
A Topology Control Algorithm Taking into Account Energy and Quality of Transm...
 
Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02
Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02
Wirelesspowertheftmonitoringoriginalbyrupalipatra 121126043543-phpapp02
 

More from Mohammad Amin Amjadi (18)

Seminar-Parallel Processing
Seminar-Parallel ProcessingSeminar-Parallel Processing
Seminar-Parallel Processing
 
Seminar-Architecture
Seminar-ArchitectureSeminar-Architecture
Seminar-Architecture
 
Image Cryptography and Steganography
Image Cryptography and SteganographyImage Cryptography and Steganography
Image Cryptography and Steganography
 
Amjadi - Ebook 7 - Class - v1
Amjadi - Ebook 7 - Class - v1Amjadi - Ebook 7 - Class - v1
Amjadi - Ebook 7 - Class - v1
 
Amjadi - Ebook 6 - Ref,Out - v1
Amjadi - Ebook 6 - Ref,Out - v1Amjadi - Ebook 6 - Ref,Out - v1
Amjadi - Ebook 6 - Ref,Out - v1
 
Amjadi - Ebook 5 - Function - v1
Amjadi - Ebook 5 - Function - v1Amjadi - Ebook 5 - Function - v1
Amjadi - Ebook 5 - Function - v1
 
Az Micro
Az MicroAz Micro
Az Micro
 
my project
my projectmy project
my project
 
Rajabi
RajabiRajabi
Rajabi
 
Cloud Computing
Cloud ComputingCloud Computing
Cloud Computing
 
Amjadi
AmjadiAmjadi
Amjadi
 
rivercode.PDF
rivercode.PDFrivercode.PDF
rivercode.PDF
 
Documention
DocumentionDocumention
Documention
 
HotSpot
HotSpotHotSpot
HotSpot
 
GPGPU
GPGPUGPGPU
GPGPU
 
Lecture3
Lecture3Lecture3
Lecture3
 
Lecture2
Lecture2Lecture2
Lecture2
 
Lecture 1
Lecture 1Lecture 1
Lecture 1
 

memetic algorithm

  • 1. Ehsan Ramezani MohammadAmin Amjadi Shahid bahonar university May 2013 11/19/2015
  • 2.  Energy conservation probably constitutes the most important challenge in the design of wireless sensor networks (WSNs).  Another issue in WSN design is the connectivity of the network according to some specific communication protocol (Cluster-based architecture). 11/19/2015
  • 3.  The purpose of the sensor network, which is the collection and possibly the management of measured data for some particular application, must not be neglected.  Most algorithms that lead to optimal topologies of WSNs towards power conservation, do not take into account the principles , characteristics and requirements of application-specific WSNs. 11/19/2015
  • 4.  One of the most powerful heuristics that could be applied to our multi-objective optimization problem is based on “Genetic Algorithms” (GAs).  an integrated GA approach , both in the direction of degrees of freedom of network characteristics and of application-specific requirements represented in the performance metric of the GA is considered. 11/19/2015
  • 5.  optimization problem → minimization of the energy-related parameters and the maximization of sensing points’ uniformity, subject to the connectivity constraints and the spatial density requirement.  ↓  single objective function  ↓  minimization 11/19/2015
  • 6.  he original GA-based algorithm was applied dynamically to obtain a dynamic sequence of operating modes for each sensor.  i.e. a sequence of WSN designs, which leads to maximization of network lifetime in terms of number of data collection (measuring) cycles. 11/19/2015
  • 7.  Memetic algorithms introduce local search techniques at specific parts of a GA optimization process, with a goal to improve its performance.  In this work, we develop and parameterize a memetic algorithm (MA) which hybridizes the GA system and the goal being to improve its performance by guiding the population formulation of the GA towards more intelligent decisions. 11/19/2015
  • 8.  WSN modeling  Design parameters 11/19/2015
  • 9.  A cluster-based network architecture is used where sensors are partitioned into several clusters.  All sensors are identical and may be either active or inactive.  Sensores three supported signal ranges:  1)CH sensor that allows the communication with the remote base station (sink).  2) HSR sensor  3)LSR sensor 11/19/2015
  • 10.  An evolutionary algorithm, requires the proper definition of some optimization criteria.  In order for the optimization to be possible, it is required that these design parameters are explicitly defined and expressed mathematically.  Application specific parameters  Connectivity parameters  Energy related parameters 11/19/2015
  • 11.  he satisfaction or not of the demand on uniformity of measuring points has been taken into consideration using two design parameters:  (a) MRD : the spatial mean relative deviation of sensing points, representing the uniformity of those points .  (b) the desired spatial density of measuring points. 11/19/2015
  • 12.  MRD →minimized ≡ uniformity is maximized  The spatial density of sensing points has to be as close as possible to the desired value.  The entire area of interest was divided into several overlapping sub-areas.  Sub-areas are defined by four factors : length and width(size) , overlapping ratio (ratios in the two directions). 11/19/2015
  • 13.  MRD= 𝑖=1 𝑁 |𝑃𝑠𝑖 − 𝑃𝑠| 𝑁 . 𝑃𝑠  Ps : spatial density of sensing points in total area  Psi : spatial density of sensing points in sub-area Si  N : number of overlapping sub-areas into which the entire area was divided  Low MRD → high uniformity  11/19/2015
  • 14.  SDE = 𝑃𝑑 −𝑃𝑠 𝑃𝑑 𝑖𝑓 𝑃𝑠 < 𝑃𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒  Pd : set to 0.2 11/19/2015
  • 15.  This set of design parameters includes two factors : first : that each clusterhead does not have more than a maximum predefined number of sensors in its cluster (SCE) second : that each sensor of the network can communicate with its clusterhead (SORE) 11/19/2015
  • 16.  SCE = 𝑖=1 𝑛 𝑓𝑢𝑙𝑙 𝑛𝑖 𝑛 𝑓𝑢𝑙𝑙 𝑖𝑓 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 𝑛 𝑓𝑢𝑙𝑙 > 0  𝑛 𝑓𝑢𝑙𝑙 : number of clusterheads that have more than 15 active sensors in clusters  𝑛𝑖 : number of sensors in the ith of those clusters 11/19/2015
  • 17.  SORE = 𝑛 𝑜𝑢𝑡 𝑛  𝑛 𝑜𝑢𝑡 : number of active sensors that cannot communicate with their clusterhead  n : total number of active sensors in the network 11/19/2015
  • 18.  This set of design parameters can be divided into two sub-categories : OE : operational energy consumption parameter CE: communicational energy consumption parameter 11/19/2015
  • 19.  OE = 20 . 𝑛 𝐶𝐻 𝑛 + 2 . 𝑛 𝐻𝑆 𝑛 + 𝑛 𝐿𝑆 𝑛  𝑛 𝐶𝐻 : operation mode of the sensor is CH  𝑛 𝐻𝑆 : operation mode of the sensor is HS  𝑛 𝐿𝑆 : operation mode of the sensor is LS  corresponding relevance factors for three active operating modes of the sensors : 20:2:1 11/19/2015
  • 20.  CE = 𝑖=1 𝑐 𝑗=1 𝑛 𝑖 𝜇. 𝑑𝑗𝑖 𝑘  c : number of clusters in the network  𝑛𝑖 : number of sensors in the ith cluster  𝑑𝑗𝑖:Euclidean distance from sensor j to its clusterhead  𝜇 & k : constants 11/19/2015
  • 21.  The maximization of the life duration of the network depends mainly on the remaining battery capacities of the sensors.  𝐵𝐶𝑃 𝑡 = 𝑖=1 𝑛 𝑔𝑟𝑖𝑑 𝑃𝐹𝑖 𝑡 . 1 𝐵 𝐶𝑡 𝑡 − 1 ,t=1,2,…  𝐵𝐶𝑖 𝑡 = 𝐵𝐶𝑖 𝑡−1 − 𝐵𝑅𝑅𝑖 𝑡−1 11/19/2015
  • 22.  𝐵𝐶𝑃 𝑡 : Battery Capacity Penalty of the WSN at measuring cycle t  𝑛 𝑔𝑟𝑖𝑑 : total number of available sensor nodes  𝑃𝐹𝑖 𝑡 : Penalty Factor assigned to sensor i  𝐵𝐶𝑖 𝑡 and 𝐵𝐶𝑖 𝑡−1 : are the Battery Capacities of sensor i at measuring cycles t and t – 1  𝐵𝑅𝑅𝑖 𝑡−1 : Battery Reduction Rate 11/19/2015
  • 23.  Methodology of GA:  i) encoding mechanism of the problems phenotypes into genotypes ii) formulation of an appropriate fitness function iii) choice of the genetic operators and the selection mechanism 11/19/2015
  • 24.  The parameters of each WSN design that needs to be encoded in the representation scheme of the GA are the following:  i) placement of the active sensors of the network  ii) operation mode of each active sensor(clusterhead or a ‘‘regular sensor”)  iii) range of signal (high or low),in ‘‘regular sensor”  X sensors in the WSN → each string in the GA population has a length of 2X 11/19/2015
  • 25.  F= 1 (𝛼1 . 𝑀𝑅𝐷 + 𝛼2. 𝑆𝐷𝐸 + 𝛼3. 𝑆𝐶𝐸 + 𝛼4. 𝑆𝑂𝑅𝐸 + 𝛼5. 𝑂𝐸 + 𝛼6. 𝐶𝐸 + 𝛼7. 𝐵𝐶𝑃)  𝛼i : weighting coefficients  The values of these coefficients were determined based on experience about the importance of each parameter. 11/19/2015
  • 26.  The concept of the memetic algorithm is based on the introduction of some battery level threshold values for each operating mode of the sensors.  at each measuring cycle to allow a sensor i to operate at some specific mode if and only if its battery level at the time is above the threshold value for that operating mode. 11/19/2015
  • 27.  The MA approach is materialized through two separate processes:  (i) the local search : battery level of each sensor against the corresponding threshold value  (ii) the appropriate update of the threshold values for each operating mode, according to some specific reduction scheme.  11/19/2015
  • 28.  The main part of the MA is the local search that is performed in the generation of the population of the original GA.  local search is performed in each individual of the GA population . 11/19/2015
  • 29.  Some initial threshold values of battery levels for each of the three possible operating modes of the sensors are defined.  he intention of these threshold values is to put specific constraints in the operation modes of each sensor, throughout the dynamic design of the network.  Operating mode of each sensor is checked 11/19/2015
  • 30.  battery level is below threshold ↓  operating mode is changed to the lower mode ↓  until its corresponding threshold value becomes lower than (or equal to) its battery level 11/19/2015
  • 31.  Three facts are evident about the proposed modification:  First, it is applied to the phenotype of the design problem and not the genotype of the optimization process of the GA  Second, investigation on the appropriate level of reduction of the operating mode of the modified sensors resembles some kind of search operation.  Third, modification always leads towards the direction of a local improvement , without any assurance that this would result into a generally better solution. 11/19/2015
  • 32.  Threshold values of battery levels for each of the three possible operating modes of the sensors are initialized.  Three parameters of reduction scheme of each threshold: initial value reduction formula reduction rate 11/19/2015
  • 33.  (i) geometric reduction: 𝑇 𝑡+1 = (1 − 𝑅𝑅). 𝑇 𝑡 T : any of the three types of threshold t : some specific measuring cycle RR : reduction rate parameter  1t T  11/19/2015
  • 34.  (ii) linear reduction: 𝑇 𝑡+1 = 𝑇 𝑡 − 𝑅𝑅 (iii) no update, where thresholds are kept constant over time. 11/19/2015
  • 35.  he performance of the MA approach to the dynamic optimization of WSN designs was compared to that of the original GA-based system, during 15 consecutive measuring cycles of the WSN. 11/19/2015
  • 38.  The MA system showed considerable improvement in energy conservation of the network resources over the already successful performance of the GA system , while the application-specific characteristics of the sensor networks were kept close to optimal values. 11/19/2015
  • 39.  The satisfactory performance of the algorithm during the dynamic network design process makes it a valuable tool for design optimization towards maximization of the life span of WSNs, especially in cases where satisfaction of some application-specific requirements is a necessity. 11/19/2015
  • 40.  In addition, it was shown that appropriate manipulation of the population of the GA (something that was introduced by the proposed local search scheme) can lead to performance improvement. 11/19/2015
  • 41.  Thanks for listening!!! 11/19/2015