www.studymafia.

com

A Seminar report on

Computational intelligence in wireless sensor networks

Submitted in partial fulfillment of the requirement for the award of degree Of Computer Science

SUBMITTED

TO:

SUBMITTED

BY:

www.studymafia.com

www.studymafia.com

www.studymafia.com

Preface

I have made this report file on the topic Computational intelligence in wireless sensor networks, I have tried my best to elucidate all the relevant detail to the topic to be included in the report. While in the beginning I have tried to give a general view about this topic. My efforts and wholehearted co corporation of each and everyone has ended on a successful note. I e!press my sincere gratitude to """"..who assisting me throughout the prepration of this topic. I than# him for providing me the reinforcement$ confidence and most importantly the trac# for the topic whenever I needed it.

www.studymafia.com
ABSTRACT ABSTRACT Wireless sensor networks (WSNs) are networks of distributed autonomous devices that can sense or monitor physical or environmental conditions cooperatively. WSNs face many challenges, mainly caused by communication failures, storage and computational constraints and limited power supply. Paradigms of omputational !ntelligence ( !) have been successfully used in recent years to address various challenges such as optimal deployment, data aggregation and fusion, energy aware routing, task scheduling, security, and locali"ation. ! provides adaptive mechanisms that e#hibit intelligent behavior in comple# and dynamic environments like WSNs. ! brings about fle#ibility, autonomous behavior, and robustness against topology changes, communication failures and scenario changes. $owever, WSN developers can make use of potential ! algorithms to overcome the challenges in Wireless Sensor Network. %he seminar includes some of the WSN challenges and their solutions using ! paradigms. &eywords ' lustering, computational intelligence, data aggregation, deployment, design, locali"ation, routing, scheduling, security, wireless sensor networks omputational

www.studymafia.com

ontent
I%&'O()C&IO% .....................................................................................................................* +. C,A--.%/.S I% WI'.-.SS S.%SO' %.&WO'0S ........................................................1 +.2 3'O3.'&I.S O4 WS% (.3-O5M.%&S.....................................................................1 +.+ MA6O' C,A--.%/.S..................................................................................................7 *. 3A'A(I/MS O4 COM3)&A&IO%A- I%&.--I/.%C.....................................................8 *.2 %.)'A- %.&WO'0S ..................................................................................................8 *.+. 4)995 -O/IC...............................................................................................................: *.*. .;O-)&IO%A'5 A-/O'I&,MS..............................................................................: *.1 SWA'M I%&.--I/.%C..............................................................................................< 1. CI =AS.( SO-)&IO%S 4O' WS% C,A--.%/.S...........................................................22 1.2 .;O-)&IO%A'5 A-/O'I&,M 4O' %.&WO'0 (.SI/%..................................22 1.+ 4)995 -O/IC 4O' (.3-O5M.%&.........................................................................2+ 1.* SWA'M I%&.--I/.%C. 4O' -OCA-I9A&IO%....................................................2* 1.1 %.)'A- %.&WO'0S 4O' S.C)'I&5...................................................................21 1.7 .%.'/5 AWA'. 'O)&I%/ A%( C-)S&.'I%/ .................................................27 7. /)I(. &O CI M.&,O(S 4O' WS%s .................................................................................2: >. CO%C-)SIO%........................................................................................................................2< 7. REFERENCES ........................................................................................................................20

www.studymafia.com
INTRODUCTION
A Wireless sensor networ# is a networ# of distributed autonomous devices that can sense or monitor physical or environmental conditions cooperatively. WS%s are used in numerous applications such as environmental monitoring$ habitat monitoring$ prediction and detection of natural calamities$ medical monitoring and structural health monitoring. WS%s consist of a large number of small$ ine!pensive$ disposable and autonomous sensor nodes that are generally deployed in an ad hoc manner in vast geographical areas for remote operations. Sensor nodes are severely constrained in terms of storage resources$ computational capabilities$ communication bandwidth and power supply. &ypically$ sensor nodes are grouped in clusters$ and each cluster has a node that acts as the cluster head. All nodes forward their sensor data to the cluster head$ which in turn routes it to a speciali?ed node called sin# node @or base stationA through a multi hop wireless communication. ,owever$ very often the sensor networ# is rather small and consists of a single cluster with a single base station .Other scenarios such as multiple base stations or mobile nodes are also possible. 'esource constraints and dynamic topology pose technical challenges in networ# discovery$ networ# control and routing$ collaborative information processing$ querying$ and tas#ing . CI combines elements of learning$ adaptation$ evolution and fu??y logic to create intelligent machines. In addition to paradigms li#e neuro computing$ reinforcement learning$ evolutionary computing and fu??y computing$ CI encompasses techniques that use swarm intelligence$ artificial immune systems and hybrids of two or more of the above. 3aradigms of CI have found practical applications in areas such as product design$ robotics$ intelligent control$ biometrics and sensor networ#s. 'esearchers have successfully used CI techniques to address many challenges in WS%s. ,owever$ various research communities are developing these applications concurrently$ and a single overview thereof does not e!ist. &heir aim is to bridge the gap between CI approaches and applications$ which provide the WS% researchers with new ideas and incentives. A discussion on yet une!plored challenges in WS%s$ and a proBection on potential CI applications in WS% are presented with an obBective of encouraging researchers to use CI techniques in WS% applications.

CHALLENGES IN WIRELESS SENSOR NETWORKS

www.studymafia.com
'eal deployments of Wireless Sensor %etwor#s @WS%A usually implement one of the three general applicationsC D 3eriodic reporting D .vent detection D (atabase li#e storage. 3eriodic reporting is by far the most used and the simplest application scenario$ in which at regular intervals the sensors sample their environment$ store the sensory data$ and send it to the base station@sA. Actuators such as automatic irrigation systems and alarm systems are often connected with such WS%s. &his scenario is used in most monitoring applications in agriculture$ microclimate and habitat surveillance$ military operations $ and disaster relief . &he main property of periodic reporting applications is the predictability of the data traffic and volume. In .vent detection applications nodes sense the environment and evaluate the data immediately for its usefulness. If useful data @an eventA is detected$ the data is transmitted to the base station@sA. &he data traffic can hardly be predictedC events usually occur randomly and the resulting data traffic is sporadic. ,owever$ a small amount of data has to be e!changed for route management and aliveness chec#s even when no events are detected. (atabase li#e storage systems are similar to event based systems. All sensory data @regular sampling or eventsA is stored locally on the nodes. =ase stations search for interesting data and retrieve it from the nodes directly. &he main challenge in these applications is to store the data in a smart way for fast search and retrieval. 2.1 PROPERTIES OF WS !EP"O#$E TS

www.studymafia.com
&he properties of WS% can be summari?ed as followsC Wireless a% &oc natureC A fi!ed communication infrastructure does not e!ist. &he shared wireless medium places additional restrictions on the communication between the nodes and poses new problems li#e unreliable and asymmetric lin#s. =ut$ it provides the broadcast advantageC A pac#et transmitted by a node to another is received by all neighbors of the transmitting node. $o'ilit( an% topolog( c&angesC WS%s may involve dynamic scenarios. %ew nodes may Boin the networ#$ and the e!isting nodes may either move through the networ# or out of it. %odes may cease to function$ and surviving nodes may go in or out of transmission radii of other nodes. WS% applications have to be robust against node failure and dynamic topology. Energ( limitationsC %odes in most WS%s have limited energy. &he basic scenario includes a topology of sensor nodes$ and a limited number of more powerful base stations. Maintenance or recharging of the batteries on sensor nodes is not possible after deployment. Communication tas#s consume ma!imum power available to sensor nodes$ and in order to ensure sustained long term sensing operation$ communication tas#s need to be e!ercised frugally. P&(sical %istri'utionC .ach node in a WS% is an autonomous computational unit that communicates with its neighbours via messages. (ata is distributed throughout the nodes in the networ# and can be gathered at a central station only with high communication costs. Consequently$ algorithms that require global information from the entire networ# become very e!pensive. &hus$ reticent distributed algorithms are highly desirable. 2.2 $)*OR C+)""E ,ES A brief description of the maBor WS% challenges addressed by CI techniques is presented in the following subsectionsC ). !esign an% !eplo(ment WS%s are used in vastly diversified applications ranging from monitoring a biological system through tissue implanted sensors to monitoring forest fire through air dropped sensors. In some applications$ the sensor nodes need to be placed accurately at predetermined locations$ whereas in others$ such positioning is unnecessary or impractical. Sensor networ# design aims at determining the type$ amount and location of sensor nodes to be placed in an environment in order to get a complete #nowledge of its functioning conditions. -. "ocali.ation %ode locali?ation refers to creating location awareness in all deployed sensor nodes. -ocation information is used to detect and record events$ or to route pac#ets using geometric aware routing. =esides$ location itself is often the data that needs to be sensed. -ocali?ation methods that use time of arrival of signals from multiple base stations are commonly used in WS%s. C. !ata )ggregation an% Sensor Fusion Sensor fusion is the process of combining of the data derived from multiple sources such that either the resulting information is in some sense better than would be possible with individual sources$ or the communication overhead of sending individual sensor readings to the base station is reduced. (ue to large scale deployment of sensors$ voluminous data is generated$ efficient collection of which is a critical issue. Most widely used non CI methods for sensor fusion include 0alman filter$ =ayesian networ#s and (empster Shafer method. !. Energ( )ware Routing an% Clustering

www.studymafia.com
.conomic usage of energy is important in WS%s because replacing or recharging the batteries on the nodes may be impractical$ e!pensive or dangerous. In many applications$ networ# life e!pectancy of a few months or years is desired. 'outing refers to determining a path for a message from a source node to a destination node. In proactive routing methods$ routing tables are created and stored regardless of when the routes are used. In reactive routing methods$ routes are computed as necessary. In densely deployed networ#s$ routing tables ta#e a huge amount of memory$ and therefore$ hybrids of proactive and reactive methods are suitable for such networ#s. Another possible solution is to cluster the networ# into hierarchies. E. Sc&e%uling In order to conserve energy$ typical sensor nodes remain in sleep mode most of the time$ and go into active mode periodically in order to acquire and transmit sensory data. A strict schedule needs to be followed regarding when a node should wa#e up$ sense$ transmit @or perform locomotionA$ ensuring ma!imum networ# lifetime. Causing the WS% nodes to ta#e right actions at right time is the maBor obBective of WS% scheduling. F. Securit( Wireless lin#s in WS%s are susceptible to eavesdropping$ impersonating$ message distorting etc. 3oorly protected nodes that move into hostile environments can be easily compromised. Administration becomes more difficult due to dynamic topology. ,. /ualit( o0 Ser1ice $anagement oS is a! overused term t"at "as various mea!i!#s a!d perspectives. oS #e!erally refers to t"e $uality as perceived %y t"e user&applicatio!' w"ile i! t"e !etwor(i!# commu!ity' oS is accepted as a measure of t"e service $uality t"at t"e !etwor( offers to t"e applicatio!s&users. oS refers to a! assura!ce %y t"e !etwor( to provide a set of measura%le service attri%utes to t"e e!d)to)e!d users&applicatio!s i! terms of fair!ess' delay' *itter' availa%le %a!dwidt"' a!d pac(et loss. A !etwor( "as to provide t"e oS w"ile ma+imi,i!# !etwor( resource utili,atio!. To ac"ieve t"is #oal' t"e !etwor( is re$uired to a!aly,e t"e applicatio! re$uireme!ts a!d deploy various !etwor( oS mec"a!isms.

PARADIGMS OF COMPUTATIONAL INTELLIGENCE

www.studymafia.com
Computational Intelligence @CIA is the study of adaptive mechanisms that enable or facilitate intelligent behaviour in comple! and changing environments. &hese mechanisms include paradigms that e!hibit an ability to learn or adapt to new situations$ to generali?e$ abstract$ discover and associate. CI is also defined as the computational models and tools of Intelligence capable of inputting raw numerical sensory data directly$ processing them by e!ploiting the representational parallelism and pipelining the problem$ generating reliable and timely responses and withstanding high fault tolerance. 3aradigms of CI are designed to model the aspects of biological intelligence. CI encompasses paradigms such as neural networ#s$ reinforcement learning$ swarm intelligence$ evolutionary algorithms$ fu??y logic and artificial immune systems. &hese paradigms are briefly introduced in the following subsections. 2.1 E3R)" ETWOR4S &he human brain$ which possesses an e!traordinary ability to learn$ memori?e and generali?e$ is a dense networ# of over 2E billion neurons$ each connected on average to about 2E$EEE other neurons. &he three basic components of an artificial neuron In 4igure areC 2A &he lin#s that provide weights WBi$ to the n inputs of Bth neuron !i$ i F 2$ . . . $ nG +A An aggregation function that sums the weighted inputs to compute the input to the activation function uB F$ where B is the bias$ which is a numerical value associated with the neuron. *A An activation function that maps uB to vB F @uBA$the output value of the neuron. Some e!amples of the activation functions areC step$ sigmoid$ tan hyperbolic and /aussian function.

www.studymafia.com
2.2. F355# "O,IC Classical set theory allows elements to be either included in a set or not. &his is in contrast with human reasoning$ which includes a measure of imprecision or uncertainty$ which is mar#ed by the use of linguistic variables such as most$ many$ frequently$ seldom etc. &his appro!imate reasoning is modelled by fu??y logic$ which is a multivalued logic that allows intermediate values to be defined between conventional threshold values. 4u??y systems allow the use of fu??y sets to draw conclusions and to ma#e decisions. 4u??y sets differ from classical sets in that they allow an obBect to be a partial member of a set. 4or e!ample$ a person may be a member of the set tall to a degree of E.: . In fu??y systems$ the dynamic behaviour of a system is characteri?ed by a set of linguistic fu??y rules based on the #nowledge of a human e!pert. 4u??y rules are of the general formC if antecedent@sA then consequent@sA$ where antecedents and consequents are propositions containing linguistic variables. Antecedents of a fu??y rule form a combination of fu??y sets through the use of logic operations. &hus$ fu??y sets and fu??y rules together form the #nowledge base of a rule based inference system as shown in 4igure. Figure 3.2 Fuzzy Inference Engine 2.2. E6O"3TIO )R# )",ORIT+$S

www.studymafia.com
.volutionary algorithms model the natural evolution$ which is the process of adaptation with the aim of improving survival capabilities through processes such as natural selection$ survival of the fittest$ reproduction$ mutation$ competition and symbiosis. .C encompasses a variety of .As that share a common underlying idea of survival of the fittest. .As use a population of solution candidates called chromosomes. Chromosomes are composed of genes$ which represent a distinct characteristic. A fitness function$ which the .A see#s to ma!imi?e over the generations$ quantifies the fitness of an individual chromosome. 3rocess of reproduction is used to mi! characteristics of two or more chromosomes @called parentsA into the new ones @called offspringA. Offspring chromosomes are mutated through small$ random genetic changes in order to increase diversity. Some fittest chromosomes are selected to go into the ne!t generation$ and the rest are eliminated. &he process is repeated generation after generation until either a fit enough solution is found or a previously set computational limit is reached. 4ollowing are the maBor classes of .As. D /enetic algorithms @/AA$ which model genetic evolution D /enetic programming whose individual chromosomes are computer programs D .volutionary programming which model adaptive behaviour in evolution D .volutionary strategies which model strategy parameters that control variation in evolution D (ifferential evolution which is identical to /A e!cept for the reproduction mechanism D Cultural evolution which models the evolution of culture of a population and cultureHs influence on genetic and adaptive evolution of individuals D Co evolution in which initially IdumbJ individuals evolve through cooperation or competition and become fit enough to survive Successful applications of .A include planning$ design$ control$ classification and clustering$ time series modelling$ music composing etc. 2.7 SW)R$ I TE""I,E CE Swarm Intelligence @SIA originated from the study of collective behavior of societies of biological species such as floc#s of birds$ shoals of fish and colonies of ants. SI is the property of a system whereby collective behaviors of unsophisticated agents interacting locally with their environment cause coherent functional global patterns to emerge. While graceful but unpredictable bird floc# choreography inspired the development of particle swarm optimi?ation $ impressive ability of a colony of ants to find shortest path to their food inspired the development of ant colony optimi?ation . &he honey bee algorithm mimics foraging behavior of swarms of honey bees . Particle Swarm Optimi.ation8 &he basic 3SO consists of a population @or a swarmA of s particles$ each of which represents a candidate solution. &he particles e!plore an n dimensional space in search of the global solution$ where n represents the number of parameters to be optimi?ed. .ach particle i occupies position !id and moves with a velocity vid$ 2 K i K s and 2 K d K n. 3articles are initially assigned random positions and velocities within fi!ed boundaries$ i.e.$ ! min K !id K !ma! and vmin K vid K vma! @in most cases vmin F Lvma!A. 4itness of a particle is determined from its position. &he fitness is defined in such a way that a particle closer to the solution has higher fitness value than a particle that is far away. In each iteration$ velocities and positions of all particles are updated to persuade them to achieve better fitness. &he process of updating is repeated iteratively either until a particle reaches the global solution within permissible tolerance limits$ or until a sufficiently large number of iterations is reached. Magnitude and direction of movement of a particle is influenced by its previous velocity$ its e!perience and the #nowledge it acquires from the swarm through social interaction.Computational Intelligence in WSNs Department of Computer Science 10 C S!"

www.studymafia.com
In the gbest version of 3SO$ each particle has a memory to store pbest id$ the position where it had the highest fitness. =esides$ each particle can access the position gbest d$ the position of the particle having the ma!imum fitness. &he gbest particle represents the best solution found as yet. At each iteration #$ 3SO adds velocity vid to each position !id and steers the particles towards its pbest id and gbestd using @2A and @+A. vid@# M 2A F w N vid@#A M c2 N rand2 N @pbestid L !idA Mc+ N rand+ N @gbestd L !idA @2A !id@# M 2A F !id@#A M vid@# M 2A .........@+A ,ere$ rand2 and rand+ are random numbers having uniform distribution in the range @E$ 2A. &he velocity update equation @2A shows that a particle is influenced by * components of acceleration. &he first term involves the inertia coefficient w$ E.+ O w O 2.+$ and it denotes the inertia of the particle . &he second term involves the cognitive acceleration constant c2. &his component propels the particle towards the position where it had the highest fitness. &he third term involves the social acceleration constant c+. &his component steers the particle towards the particle that currently has the highest fitness. &he velocity of a particle is bounded between properly chosen limits v ma! and vmin. Similarly$ the position of a particle is restricted between properly chosen constants ! ma! and !min. Several variants of 3SO have been devised and applied to optimi?ation problems in power systems$ stoc# mar#ets$ antenna control and WS%s. )nt Colon( Optimi.ationC ACO was introduced as a metaheuristic for solving combinatorial optimi?ation problems. 4oraging ants initially e!plore surroundings of their nest in a random manner. When an ant finds a source of food$ it evaluates quantity and quality of the food and carries some food to the nest. While returning to the nest$ the ant deposits a trail of chemical pheromone$ which guides other ants to the food source. &his characteristic of ant colonies is e!ploited in artificial ant colonies to solve combinatorial optimi?ation problems. Consider that two paths A and = e!ist between a nest and a food source$ and nA@tA and n=@tA number of ants use them at time step t respectively$ then the probability of ant choosing path A at the time step t M 2 is given by w"ere c is t"e de#ree of attractio! of a! u!e+plored %ra!c"' a!d - is t"e %ias to usi!# p"eromo!e deposits i! t"e decisio! process. A! a!t c"ooses %etwee! t"e pat" A or pat" B usi!# t"e decisio! rule. if /00' 12 3 4A0t512 t"e! c"oose pat" A ot"erwise c"oose pat" B. T"e mai! idea of t"e AC6 meta"euristic is to model t"e pro%lem as a searc" for t"e %est pat" i! a 7co!structio! #rap"8 t"at represe!ts t"e states of t"e pro%lem. Artificial a!ts wal( t"rou#" t"is #rap"' loo(i!# for #ood pat"s. T"ey commu!icate %y layi!# p"eromo!e trails o! ed#es of t"e #rap"' a!d t"ey c"oose t"eir pat" wit" respect to pro%a%ilities t"at depe!d o! t"e amou!t of p"eromo!e previously left.

www.studymafia.com

CI BASED SOLUTIONS FOR WSN CHALLENGES

In this section we can see how Computational Intelligence techniques can be used as solutions for wireless sensor networ# challenges. CI techniques can be very helpful in the process of designing and planning the deployments of sensor networ#s and there have been many efforts to apply them in this conte!t. 7.1 E6O"3TIO )R# )",ORIT+$ FOR ETWOR4 !ESI, A decision support system based on /enetic Algorithm is proposed. &he decision support system is meant for the use of a process engineer who interacts with it to determine optimal sensor networ# design as outlined in 4igure. Figure #.1 $ole of %! in Decision Support System &he process engineer introduces information about the process under stationary operating conditions through a rigorous module called model generator$ which involves linear functionalities and non linear equations. &hen the engineer chooses the initial sensor networ# from the candidates. &he chosen configuration undergoes observability analysis which determines which of the unmeasured variables are observable. =ased on the results$ the engineer decides whether the information achieved from the configuration is satisfactory or not. If not$ the sensor networ# needs to be improved by adding more instruments before the analysis is repeated. &raditionally$ the analysis begins from an initial sensor networ# with a few instruments chosen by the process engineer based on hisPher s#ills and e!perience. &hen$ an iterative process ta#es placeG the process engineer runs the tools in the decision support system and refines the configuration until satisfactory instrumentation is achieved. In the case of comple! industrial processes$ a complete iteration involves a great deal of e!pert e!amination and several e!ecutions of the analysis software tools. &herefore$ a good initial configuration assures lesser number of iterations before the final configuration is reached. .ach chromosome used in /A is a sequence of length l which represents number of variables in the mathematical model of the networ#$ li#e in 2EEE22EE. ,ere$ a 2 represents the presence of a sensor to measure the variable at that position in the mathematical model$ and l is the number of variables in the mathematical model. /A uses binary tournament to select individuals to go to ne!t generation maintaining the best up to the moment individual with the elitist approach. &he fitness function$ which /A see#s to ma!imi?e$ is defined as where %'@iA$ %obs@iA and %C@iA are the normali?ed values corresponding to the reliability$ observability and cost terms$ respectively. &he proposed scheme is tested on a ammonia synthesis plant$ where all the compositions and critical variables were set as measured. &he performance of an initial sensor configuration suggested by the /A reported to be more cost effective and more reliable than the initial configuration determined by the process engineer. &he initial configuration suggested by /A is reported to have achieved a cost saving of >EQ$ and a smaller amount of time to complete the whole observability analysis. 7.2 F355# "O,IC FOR !EP"O#$E T &his technique assumes that the area to be monitored by a sensor networ# is divided into a square grid of subareas$ each having its own terrain profile and a level of required surveillance @therefore$ its own path loss model and required path loss thresholdA. &he proposed technique uses fu??y logic to determine the number of sensors n@iA necessary to be scattered in a subarea i. 4or a sub area i$ path loss 3- @iA and threshold path loss 3-&, are normali?ed on a scale E to 2E$ then divided into overlapping membership functions low$ medium and high. &his requires * + F < rules. &he system computes an output weight @iA$ from which the number of sensors is determined as Membership 4unctions are defined to form three 4u??y sets -ow$ Medium and ,igh depending on the values of 3- and 3-&,

www.studymafia.com
7.2 SW)R$ I TE""I,E CE FOR "OC)"I5)TIO Most WS% locali?ation algorithms share a common feature that they estimate the location of nodes using the a priori #nowledge of the positions of special nodes called beacon nodes or anchor nodes. &he most intuitive solution to the locali?ation problem is to load each node with a global positioning system. =ut it is not an attractive solution because of cost$ si?e and power constraints. &ypically$ the nodes estimate their locations using signal propagation time or received signal strength. Signal propagation time is estimated through measurement of time of arrival$ round trip time of flight or time difference of arrival of two signals. &he locali?ation is formulated as a multidimensional optimi?ation problem$ and tac#led with population based CI methods such as /A and 3SO. WS% locali?ation is a two phase process which includes 'anging phase and 3osition estimation phase. Whatever be the ranging method$ there will be measurement errors in practical locali?ation systems that result in noisy range estimations.,ere proposed an optimi?ation approach that minimi?es the error in locating the coordinates of the target %odes based on Swarm Intelligence called 3article Swarm optimi?ation @3SOA. PSO 'ase% "ocali.ation sc&eme &he 3osition estimation of the coordinates of the target nodes can be formulated as an optimi?ation problem$ involving the minimi?ation of an obBective function representing the error in locating the target nodes. (istance between target node and anchor node is calculated by Initial value of position is estimated $And then base station runs 3SO to minimi?e locali?ation error with the obBective function$ where @!$ yA is the node location that needs to be determined$ M R * is the number of anchors @! i$ yiA is the location of anchor i. ,ere is the measured value of distance di between a node and the anchor i$ obtained under noise. MA&-A= simulation results shows that locali?ation error in 3SO is more than halved in all e!perimental setups. 7.7 E3R)" ETWOR4S FOR SEC3RIT# Many types of (oS attac#s on WS%s have been devised. In collision attac#s$ attac#ers transmit pac#ets regardless of status of the medium. &hese pac#ets collide with data or control pac#ets from the legitimate sensors. In unfairness attac#s$ adversaries transmit as any pac#ets as possible after sensing that the medium is free. &his prevents the legitimate sensors from transmitting their own pac#ets. In e!haustion attac#s$ adversaries transmit abnormally large number of ready to send pac#ets to normal sensor nodes$ thereby e!hausting their energy quic#ly. Multilayered perceptron and generali?ed neuron based @/%A distributed secure MAC protocols are proposed in which %%s onboard WS% nodes monitor traffic variations$ and compute a suspicion that an attac# is underway. When the suspicion grows beyond a preset threshold$ the WS% node is switched off temporarily. A node detects an attac# by monitoring abnormally large variations in sensitive parametersC collision rate 'c @number of collisions observed by a node per secondA$ average waiting time &w @waiting time of a pac#et in MAC buffer before transmissionA$ and arrival rate @' '&SA$ rate of '&S pac#ets received by a node successfully per second. &he strategy used in both the articles is the same$ but the nonlinear mapping between traffic pattern variation and suspicion factor is implemented by different types of %%s in these articles. M-3 and generali?ed neuron are trained using 3SO. &hese distributed methods ensure that only the part of the WS% under attac# is shut down$ which is an advantage. =esides$ these methods have been reported to be quite effective against (oS attac#s$ but the effectiveness depends on the value of the threshold suspicion. /% is a neural networ# model that is more compact and fle!ible. A /% uses both S @sumA and T @productA aggregation functions

www.studymafia.com
WS% nodes monitor traffic variations$ and compute a suspicion that an attac# is underway. When the suspicion grows beyond a preset threshold$ the WS% node is switched off temporarily. A node detects an attac# by monitoring abnormally large variations in sensitive parametersC Collision rate @'cA number of collisions observed by a node per second. Average waiting time @&wA waiting time of a pac#et in MAC buffer before transmission. Arrival rate @'rA rate of '&S pac#ets received by a node successfully per second , -ase% Secure $)C Protocol &he /% senses the rise in critical parameters 'c$ 'r and &w and produces the output$ ie used as a measure of suspicion of an attac# that shuts down the node. When the suspicion of an attac# e!ceeds a predefined threshold suspicion level $the MAC and physical layer of the node is switched off. &his results in saving of power that would have been wasted in retransmission of collided pac#ets. 7.9 E ER,# )W)RE RO3TI , ) ! C"3STERI , &he energy consumption can be reduced by allowing only some nodes to communicate with the base station. &hese nodes called cluster heads collect the data sent by each node in that cluster compressing it and then transmitting the aggregated data to the base station. Appropriate cluster head selection can significantly reduce energy consumption and enhance the lifetime of the WS%. A fu??y logic approach to cluster head election is proposed here. &he base station has the global #nowledge about the networ#. Moreover$ base stations are many times more powerful than the sensor nodes$ having sufficient memory$ power and storage. (igure ).* +rchitecture of luster Fuzzy Clus er!"ea# Elec $%& Sc"e'e 9udicious cluster "ead electio! ca! reduce t"e e!er#y co!sumptio! a!d e+te!d t"e lifetime of t"e !etwor(. T"e operatio! of t"e sc"eme is divided i!to two rou!ds eac" co!sisti!# of a setup a!d steady state p"ase. Se u( ("ase! cluster)"eads are determi!ed %y usi!# fu,,y (!owled#e processi!# a!d t"e! t"e cluster is or#a!i,ed. S ea#y P"ase! cluster)"eads collect t"e a##re#ated data a!d is t"e! se!t to t"e %ase statio!. A fu,,y lo#ic approac" %ased o! e!er#y' co!ce!tratio! a!d ce!trality is proposed for cluster "ead electio!. : N%#e E&er)y ) e!er#y level availa%le i! eac" !ode' desi#!ated %y t"e fu,,y varia%le ;e!er#y<. : N%#e C%&ce& ra $%& ) !um%er of !odes prese!t i! t"e vici!ity' desi#!ated %y t"e fu,,y varia%le ;co!ce!tratio!<. : N%#e Ce& ral$ y ) a value w"ic" classifies t"e !odes %ased o! "ow ce!tral t"e !ode is to t"e cluster' desi#!ated %y t"e fu,,y varia%le ;ce!trality<. T"e study uses a !etwor( model i! w"ic" all se!sor !odes tra!smit t"e i!formatio! a%out t"eir locatio! a!d availa%le e!er#y to t"e %ase statio!. T"e %ase statio! ta(es i!to accou!t t"e e!er#y eac" !ode "as' t"e !um%er of !odes i! t"e vici!ity a!d a !ode<s dista!ce from ot"er !odes i!to accou!t a!d determi!es w"ic" !odes s"ould wor( as t"e cluster "eads. T"e %ase statio! fu,,ifies t"e varia%les !ode e!er#y a!d !ode co!ce!tratio! i!to t"ree levels. low' medium a!d "i#"' a!d t"e varia%le !ode ce!trality i!to close' ade$uate a!d far. T"erefore' == 0>272 rules are used for t"e fu,,y rule %ase. T"e fu,,y outcome t"at represe!ts t"e pro%a%ility of a !ode %ei!# c"ose! as a cluster "ead' is divided i!to seve! levels. *ery s'all+ s'all+ ra "er s'all+ 'e#$u'+ ra "er lar)e+ lar)e+ a&# *ery lar)e . Tria!#ular mem%ers"ip fu!ctio!s are used to represe!t t"e fu,,y sets medium a!d ade$uate a!d trape,oidal mem%ers"ip fu!ctio!s to represe!t low' "i#"' close a!d far fu,,y sets. All t"e !odes are compared o! t"e %asis of c"a!ces a!d t"e !ode wit" t"e ma+imum c"a!ce is t"e! elected as t"e cluster)"ead.

www.studymafia.com
,3I!E TO CI $ET+O!S FOR WS s
Many CI methods have outperformed or complimented conventional methods under uncertain environments and severe limitations in power supply$ communication bandwidth$ and computational capabilities. ,owever$ all wor#s presented here are not the best possible solutions and many have not been compared to traditional or to other CI approaches. Additionally$ only a few researchers have evaluated their algorithms under real WS% environments li#e test bed or deployments. 4indings have been summari?ed in 4igure. &he columns of the table represent the application areas in WS%s considered in this survey$ while the rows represent the main CI techniques. &he si?e of the blac# circles represents the number of articles surveyed in this paper for the particular combination of WS% problem and CI approach. In contrast$ the shadowing of the bo! represents an evaluation of the applicability and suitability of the CI method for the particular problem. Of course$ this evaluation is not always trueC It depends highly on the e!act CI algorithm$ its parameters and the e!act formulation of the problem. ,owever$ this overview gives a good insight about which CI method to e!plore first$ when trying to solve a WS% problem.

www.studymafia.com
CONCLUSION

?e "ave discussed t"e i!!ovative use of C@ tec"!i$ues to address ?SN issues suc" as desi#! a!d deployme!t' locali,atio!' security' optimal routi!# a!d clusteri!# etc. Rece!t impleme!tatio!s of C@ met"ods i! various dy!amical a!d "etero#e!eous !etwor(s are prese!ted "ere. C@ paradi#ms a!d !umerous c"alle!#es i! se!sor !etwor(s are %riefly i!troduced' a!d t"e C@ approac"es used %y researc"ers to address t"e c"alle!#es are %riefly e+plai!ed. @! additio!' a #e!eral evaluatio! of C@ al#orit"ms is prese!ted' w"ic" will serve as a #uide for usi!# C@ al#orit"ms for ?SNs. Future researc" is li(ely to focus o! developi!# a well fou!ded a!alytical approac" to distri%uted multi)se!sor estimatio! pro%lem w"ere t"ere are time varyi!# commu!icatio! %a!dwidt" co!strai!ts. Cross)layer desi#! a!d parameter lear!i!# is e!visio!ed to %e a! i!teresti!# !ew researc" area for C@ i! ?SNs. Ri#"t !ow' a ma*ority of t"e solutio!s prese!ted "ere apply C@ to o!e limited pro%lem i! areas li(e multicast routi!#' li!( $uality' optimal clusteri!# or placeme!t. Aowever' most issues arise from cross)layer i!compati%ility a!d t"e "i#" "uma! i!terve!tio! !eeded for parameter setti!# a!d ad*ustme!t. Bear!i!# platforms a!d paradi#ms are !eeded rat"er t"a! speciali,ed solutio!s. @! spite of a multitude of successful C@ applicatio!s i! ?SNs' t"e mai! co!cer! is t"at t"e most of t"ese al#orit"ms or protocols are still i! developme!t sta#e' a!d t"ey may remai! forever i! !o!)fi!ali,ed state. Cery few protocols "ave #row! out of t"e simulatio! e!viro!me!t. Dost of t"em do !ot eve! co!sider u!relia%le or asymmetric li!(s' !ode failure a!d mo%ility. Besides' a commo! pro%lem is t"e lac( of compariso! to co!ve!tio!al state)of)t"e)art protocols to clearly ide!tify t"e adva!ta#es of i!troduci!# C@. T"us' t"e #oal of C@ researc" commu!ity for t"e future of C@ i! ?SNs is to improve already e+isti!# solutio!s' refi!e t"em a!d defi!e well)performi!# real) world protocols

www.studymafia.com

REFERENCES
2. U2V. 'aghavendra ;. 0ul#arni$ Anna 4Wrster and /anesh 0umar ;enayagamoorthy$ IComputational Intelligence in Wireless Sensor %etwor#sCA SurveyJ$ IEEE Communications Surveys X &utorials$ ;O-. 2*$ %O. 2$pp >* <>$ 4irst Yuarter +E22. +. U+V. %. =. =. 6essica A. Carballido$ Ignacio 3on?oni$ IC/( /AC A graph based genetic algorithm for sensor networ# design.J Inf. Sci.$ vol. 288$ no. ++$ pp. 7E<2Z72E+$ +EE8. *. U*V. -. 9hao and Y. -iang$ I4u??y deployment for wireless sensor networ#s$J in 3roc. I... Int. Conf. Computational Intelligence ,omeland Security 3ersonal Safety CI,S3S$ Y. -iang$ .d.$ +EE7$ pp.8<Z:*. 1. U1V. A. /opa#umar and -. 6acob$ I-ocali?ation in wireless sensor networ#s using particle swarm optimi?ation$J in 3roc. I.& Int. Conf. Wireless$ Mobile Multimedia %etw.$ +EE:$ pp. ++8Z+*E. 7. U7V. '. ;. 0ul#arni$ /. 0. ;enayagamoorthy$ A. ;. &ha#ur$ and S. 0. Madria$ I/enerali?ed neuron based secure media access control protocol for wireless sensor networ#s$J in 3roc. I... Symp.Computational Intelligence Multi Criteria (ecision Ma#ing MC(M$ Mar. *E Apr. + +EE<$ pp. 2>Z++. >. U>V.I. /upta$ (. 'iordan$ and S. Sampalli$ ICluster head election using fu??y logic for wireless sensor networ#s$J in 3roc. *rd Annual Commun. %etw. Services 'esearch Conf.$ (. 'iordan$ .d.$ +EE7$ pp.+77Z+>E.