Sensors, in the sense of devices that perform the measurement of certain quantities
andtransformthemintoacomputerreadabledigitalformat,havebeenaroundforat
least a few decades. These sensors were either connected to a data collection device
or connected directly to computers, using traditional wired communication such as
a serial interface. The development of highly integrated computer devices, wireless
radios,andminiaturizationallowedthedevelopmentofwirelesssensornodes,which
are miniature computers with integrated sensing and wireless communication capa-
bilities[1,2].Thecontinuingminiaturizationeffortallowedthedevelopmentofnodes
with a physical size of several millimeters. Although many of these devices can act
as a general-purpose computer, with the ability to perform computation as well as
sensing, there are obvious limitations. First, the limited memory and computational
powerdoesnotallowustorunafull-featuredoperatingsystem.Mostofthetime,the
networking stack needs to be simplified as well. The wireless transmission range is
limited by the small size of the antennas. The power resources of the sensor nodes
are limited by the physical size of the batteries; moreover, some of the proposed
deploymentmodels donotallowthesensornodestorecharge theirbatteries.
AlgorithmsandProtocolsforWireless SensorNetworks, EditedbyAzzedineBoukerche
Copyright©2009byJohnWiley&SonsInc.
129
130 A TAXONOMY OF ROUTING PROTOCOLS IN SENSOR NETWORKS
Letusnowconsiderthedeploymentmodelsofthewirelesssensornodes.Classical
sensors are typically used in pre-engineered deployment, being placed at a carefully
chosenlocations.Forinstance,thespaceshuttleusesseveraldozentemperaturesen-
sors, carefully positioned in the structure of the spacecraft and reporting through
a wired connection to a central computer. Naturally, wireless sensor nodes can be
deployed similarly, simply by replacing the wired connection with a wireless one.
However, we can also choose a radically different deployment method: Blanket the
desired area with a large number of nodes. Instead of careful positioning, we need
to worry only about making sure that every area of interest is covered by one node
(or preferably, several). The deployed nodes might not have the transmission range
to reach the central computer, but they can transmit the collected information on a
hop-by-hop basis to collection points called sinks. We call the resulting structure a
“wireless sensor network (WSN).” In our example application, a WSN can provide
severaladvantages:Duetothelargenumberofsensors,itcancollectmoredatathan
sensorswithpre-engineereddeployment.Asseveralsensorscoverthesamearea,they
providefaulttolerance.Inaddition,havingcomputationalaswellassensingcapabil-
ities,thewirelesssensornetworkcanprovidepreliminaryprocessingofthecollected
data concomitantly with the sensingand forwarding.
Letusnowinvestigatethepropertiesofawirelesssensornetworkfromthenetwork-
ingpointofview.Firstofall,thereisnoinfrastructureavailable.Becausethesinksare
accessibleonlytoalimitedsubsetofnodes,thesensornodesneedtoparticipateinthe
forwarding of the packets. Due to the random deployment, the routing architecture
cannotbepre-established;thenetworkneedstobesetupthroughself-configuration.
Intheserespects,wirelesssensor networks aresimilar to adhocwirelessnetworks.
There are, however, several important differences. The sensor nodes have signif-
icantly lower communication and computation capabilities than do the full-featured
computersparticipatinginadhocnetworks.Theproblemofenergyresourcesisespe-
ciallydifficult.Duetotheirdeploymentmodel,theenergysourceofthesensornode
is considered nonrenewable (although some sensor nodes might be able to scavenge
resources from their environment). Routing protocols deployed in sensor networks
needtoconsider theproblemof efficient useof power resources.
Anadditionaldifferencebetweenadhocandsensornetworksreferstotheunique-
ness of the nodes. Ad hoc nodes have a hard-wired unique MAC address, which
forms the basis of node identification on the higher levels of the networking stack.
The cheap, disposable sensor nodes usually come without any pre-wired identifiers;
theyacquireauniqueidentityonlyafterdeployment,byvirtueoftheirpositioninthe
environment.
In addition to these, several other factors such as the large number of nodes in
sensor networks, the high failure rates of the sensor nodes, and the frequent use of
broadcasting in sensor networks as opposed to the typically unicast communication
in ad hoc networks [3] require new types of MAC [4, 5] and routing protocols,
specifically targetedtowardtherequirements ofWSNs.
Inthischapter,wesuccintlypresentthemajorapplicationsoftheWSNs,describe
some of the design issues associated with routing algorithms for WSN, and finally
presentasurveyofthe stateof the art inWSN routing protocols.
6.2 APPLICATIONS
Sensor networks can be deployed in a wide variety of applications. One of the main
classification criteria is whether the sensor nodes are mobile or immobile. The data
collection might be either continuous or periodic; the latter can lead to bursty traffic
patterns.Naturally,everyapplicationrequiresaspecificsetofsensortypes.Someof
themostpopularsensortypesare:light,sound,magneticfield,accelerator,tempera-
ture, humidity, chemical composition such as soil makeup, mechanical stress levels
onanobject,andmanyothers[6].Someoftheprimaryapplicationdomainsforsensor
networksare the following:
Environmental. Environmentalsensorscanbeusedtodetectandtracknaturaldis-
asterssuchasforestfiresorfloods.Theycanalsobeusedtotrackthemovement
ofbirdsandotheranimals.
Military. The sensor networks will be an integral part of the future C4ISRT sys-
tems (command, control, communications, computing, intelligence, surveil-
lance, reconnaissance, and targeting.) They can, for instance, be used to track
themovementoftheenemyinthebattlefield.ThemainadvantageofWSNsis
that they can be deployed and operated remotely, without putting human lives
at risk. Naturally, military deployments bring their own challenges of security
andconfidentiality.
Health. Sensornetworkscanbeusedinhospitalsandclinicsforpatientmonitoring
andtrackingofvarioussystemsandhumans.Sensorscanbealsousedtotrack
andmonitorthedrugdosesprescribedtopatientsandpreventsituationswhere
the drugs are administered to the wrong patient. Sensors can be deployed for
telemonitoring of patients, a promising new direction for at-home monitoring
andcarefortheelderly.
Home. The various home appliances can be sensor enabled and interconnected
witheachotherandacentralcontrolsystemofthehome.Thesesensor-enabled
sensor homes might not only offer additional conveniences, but will also be
safer andmore energy-efficient.
DESIGN ISSUES
The challenges posed by the deployment of sensor networks is a superset of those
found in wireless ad hoc networks. Sensor nodes communicate over wireless, lossy
lineswithnoinfrastructure.Anadditionalchallengeisrelatedtothelimited,usually
nonrenewableenergysupplyofthesensornodes.Inordertomaximizethelifetimeof
thenetwork,theprotocolsneedtobedesignedfromthebeginningwiththeobjective
ofefficientmanagementoftheenergyresources[3].Letusnowdiscusstheindividual
design issuesingreater detail.
Fault Tolerance. Sensor nodes are vulnerable and frequently deployed in danger-
ousenvironment.Nodescanfailduetohardwareproblemsorphysicaldamage
or by exhausting their energy supply. We expect the node failures to be much
higherthantheonenormallyconsideredinwiredorinfrastructure-basedwire-
less networks. The protocols deployed in a sensor network should be able to
detectthesefailuresassoonaspossibleandberobustenoughtohandlearela-
tivelylargenumberoffailureswhilemaintainingtheoverallfunctionalityofthe
network.Thisisespeciallyrelevanttotheroutingprotocoldesign,whichhasto
ensure that alternate paths are available for rerouting of the packets. Different
deploymentenvironments posedifferent fault tolerancerequirements.
Scalability. Sensor networks vary in scale from several nodes to potentially
severalhundredthousand.Inaddition,thedeploymentdensityisalsovariable.
For collecting high-resolution data, the node density might reach the level
where a node has several thousand neighbors in their transmission range. The
protocols deployed in sensor networks need to be scalable to these levels and
beableto maintainadequate performance.
Production Costs. Because many deployment models consider the sensor nodes
to be disposable devices, sensor networks can compete with traditional
information gathering approaches only if the individual sensor nodes can be
produced very cheaply. The target price envisioned for a sensor node should
ideallybe less than$1.
Hardware Constraints. At minimum, every sensor node needs to have a sensing
unit, a processing unit, a transmission unit, and a power supply. Optionally,
the nodes may have several built-in sensors or additional devices such as
a localization system to enable location-aware routing. However, every
additional functionality comes with additional cost and increases the power
consumption and physical size of the node. Thus, additional functionality
needstobe always balancedagainstcostandlow-powerrequirements.
Transmission Media. The communication between the nodes is normally imple-
mented using radio communication over the popular ISM bands. However,
some sensor networks use optical or infrared communication, with the latter
havingtheadvantageof being robustandvirtually interferencefree.
Power Consumption. As we have already seen, many of the challenges of
sensor networks revolve around the limited power resources. The size of the
nodes limits the size of the battery. The software and hardware design needs
to carefully consider the issues of efficient energy use. For instance, data
compression might reduce the amount of energy used for radio transmission,
but uses additional energy for computation and/or filtering. The energy policy
alsodependsontheapplication;insomeapplications,itmightbeacceptableto
turn off a subset of nodes in order to conserve energy while other applications
requireall nodes operatingsimultaneously.
andtransformthemintoacomputerreadabledigitalformat,havebeenaroundforat
least a few decades. These sensors were either connected to a data collection device
or connected directly to computers, using traditional wired communication such as
a serial interface. The development of highly integrated computer devices, wireless
radios,andminiaturizationallowedthedevelopmentofwirelesssensornodes,which
are miniature computers with integrated sensing and wireless communication capa-
bilities[1,2].Thecontinuingminiaturizationeffortallowedthedevelopmentofnodes
with a physical size of several millimeters. Although many of these devices can act
as a general-purpose computer, with the ability to perform computation as well as
sensing, there are obvious limitations. First, the limited memory and computational
powerdoesnotallowustorunafull-featuredoperatingsystem.Mostofthetime,the
networking stack needs to be simplified as well. The wireless transmission range is
limited by the small size of the antennas. The power resources of the sensor nodes
are limited by the physical size of the batteries; moreover, some of the proposed
deploymentmodels donotallowthesensornodestorecharge theirbatteries.
AlgorithmsandProtocolsforWireless SensorNetworks, EditedbyAzzedineBoukerche
Copyright©2009byJohnWiley&SonsInc.
129
130 A TAXONOMY OF ROUTING PROTOCOLS IN SENSOR NETWORKS
Letusnowconsiderthedeploymentmodelsofthewirelesssensornodes.Classical
sensors are typically used in pre-engineered deployment, being placed at a carefully
chosenlocations.Forinstance,thespaceshuttleusesseveraldozentemperaturesen-
sors, carefully positioned in the structure of the spacecraft and reporting through
a wired connection to a central computer. Naturally, wireless sensor nodes can be
deployed similarly, simply by replacing the wired connection with a wireless one.
However, we can also choose a radically different deployment method: Blanket the
desired area with a large number of nodes. Instead of careful positioning, we need
to worry only about making sure that every area of interest is covered by one node
(or preferably, several). The deployed nodes might not have the transmission range
to reach the central computer, but they can transmit the collected information on a
hop-by-hop basis to collection points called sinks. We call the resulting structure a
“wireless sensor network (WSN).” In our example application, a WSN can provide
severaladvantages:Duetothelargenumberofsensors,itcancollectmoredatathan
sensorswithpre-engineereddeployment.Asseveralsensorscoverthesamearea,they
providefaulttolerance.Inaddition,havingcomputationalaswellassensingcapabil-
ities,thewirelesssensornetworkcanprovidepreliminaryprocessingofthecollected
data concomitantly with the sensingand forwarding.
Letusnowinvestigatethepropertiesofawirelesssensornetworkfromthenetwork-
ingpointofview.Firstofall,thereisnoinfrastructureavailable.Becausethesinksare
accessibleonlytoalimitedsubsetofnodes,thesensornodesneedtoparticipateinthe
forwarding of the packets. Due to the random deployment, the routing architecture
cannotbepre-established;thenetworkneedstobesetupthroughself-configuration.
Intheserespects,wirelesssensor networks aresimilar to adhocwirelessnetworks.
There are, however, several important differences. The sensor nodes have signif-
icantly lower communication and computation capabilities than do the full-featured
computersparticipatinginadhocnetworks.Theproblemofenergyresourcesisespe-
ciallydifficult.Duetotheirdeploymentmodel,theenergysourceofthesensornode
is considered nonrenewable (although some sensor nodes might be able to scavenge
resources from their environment). Routing protocols deployed in sensor networks
needtoconsider theproblemof efficient useof power resources.
Anadditionaldifferencebetweenadhocandsensornetworksreferstotheunique-
ness of the nodes. Ad hoc nodes have a hard-wired unique MAC address, which
forms the basis of node identification on the higher levels of the networking stack.
The cheap, disposable sensor nodes usually come without any pre-wired identifiers;
theyacquireauniqueidentityonlyafterdeployment,byvirtueoftheirpositioninthe
environment.
In addition to these, several other factors such as the large number of nodes in
sensor networks, the high failure rates of the sensor nodes, and the frequent use of
broadcasting in sensor networks as opposed to the typically unicast communication
in ad hoc networks [3] require new types of MAC [4, 5] and routing protocols,
specifically targetedtowardtherequirements ofWSNs.
Inthischapter,wesuccintlypresentthemajorapplicationsoftheWSNs,describe
some of the design issues associated with routing algorithms for WSN, and finally
presentasurveyofthe stateof the art inWSN routing protocols.
6.2 APPLICATIONS
Sensor networks can be deployed in a wide variety of applications. One of the main
classification criteria is whether the sensor nodes are mobile or immobile. The data
collection might be either continuous or periodic; the latter can lead to bursty traffic
patterns.Naturally,everyapplicationrequiresaspecificsetofsensortypes.Someof
themostpopularsensortypesare:light,sound,magneticfield,accelerator,tempera-
ture, humidity, chemical composition such as soil makeup, mechanical stress levels
onanobject,andmanyothers[6].Someoftheprimaryapplicationdomainsforsensor
networksare the following:
Environmental. Environmentalsensorscanbeusedtodetectandtracknaturaldis-
asterssuchasforestfiresorfloods.Theycanalsobeusedtotrackthemovement
ofbirdsandotheranimals.
Military. The sensor networks will be an integral part of the future C4ISRT sys-
tems (command, control, communications, computing, intelligence, surveil-
lance, reconnaissance, and targeting.) They can, for instance, be used to track
themovementoftheenemyinthebattlefield.ThemainadvantageofWSNsis
that they can be deployed and operated remotely, without putting human lives
at risk. Naturally, military deployments bring their own challenges of security
andconfidentiality.
Health. Sensornetworkscanbeusedinhospitalsandclinicsforpatientmonitoring
andtrackingofvarioussystemsandhumans.Sensorscanbealsousedtotrack
andmonitorthedrugdosesprescribedtopatientsandpreventsituationswhere
the drugs are administered to the wrong patient. Sensors can be deployed for
telemonitoring of patients, a promising new direction for at-home monitoring
andcarefortheelderly.
Home. The various home appliances can be sensor enabled and interconnected
witheachotherandacentralcontrolsystemofthehome.Thesesensor-enabled
sensor homes might not only offer additional conveniences, but will also be
safer andmore energy-efficient.
DESIGN ISSUES
The challenges posed by the deployment of sensor networks is a superset of those
found in wireless ad hoc networks. Sensor nodes communicate over wireless, lossy
lineswithnoinfrastructure.Anadditionalchallengeisrelatedtothelimited,usually
nonrenewableenergysupplyofthesensornodes.Inordertomaximizethelifetimeof
thenetwork,theprotocolsneedtobedesignedfromthebeginningwiththeobjective
ofefficientmanagementoftheenergyresources[3].Letusnowdiscusstheindividual
design issuesingreater detail.
Fault Tolerance. Sensor nodes are vulnerable and frequently deployed in danger-
ousenvironment.Nodescanfailduetohardwareproblemsorphysicaldamage
or by exhausting their energy supply. We expect the node failures to be much
higherthantheonenormallyconsideredinwiredorinfrastructure-basedwire-
less networks. The protocols deployed in a sensor network should be able to
detectthesefailuresassoonaspossibleandberobustenoughtohandlearela-
tivelylargenumberoffailureswhilemaintainingtheoverallfunctionalityofthe
network.Thisisespeciallyrelevanttotheroutingprotocoldesign,whichhasto
ensure that alternate paths are available for rerouting of the packets. Different
deploymentenvironments posedifferent fault tolerancerequirements.
Scalability. Sensor networks vary in scale from several nodes to potentially
severalhundredthousand.Inaddition,thedeploymentdensityisalsovariable.
For collecting high-resolution data, the node density might reach the level
where a node has several thousand neighbors in their transmission range. The
protocols deployed in sensor networks need to be scalable to these levels and
beableto maintainadequate performance.
Production Costs. Because many deployment models consider the sensor nodes
to be disposable devices, sensor networks can compete with traditional
information gathering approaches only if the individual sensor nodes can be
produced very cheaply. The target price envisioned for a sensor node should
ideallybe less than$1.
Hardware Constraints. At minimum, every sensor node needs to have a sensing
unit, a processing unit, a transmission unit, and a power supply. Optionally,
the nodes may have several built-in sensors or additional devices such as
a localization system to enable location-aware routing. However, every
additional functionality comes with additional cost and increases the power
consumption and physical size of the node. Thus, additional functionality
needstobe always balancedagainstcostandlow-powerrequirements.
Transmission Media. The communication between the nodes is normally imple-
mented using radio communication over the popular ISM bands. However,
some sensor networks use optical or infrared communication, with the latter
havingtheadvantageof being robustandvirtually interferencefree.
Power Consumption. As we have already seen, many of the challenges of
sensor networks revolve around the limited power resources. The size of the
nodes limits the size of the battery. The software and hardware design needs
to carefully consider the issues of efficient energy use. For instance, data
compression might reduce the amount of energy used for radio transmission,
but uses additional energy for computation and/or filtering. The energy policy
alsodependsontheapplication;insomeapplications,itmightbeacceptableto
turn off a subset of nodes in order to conserve energy while other applications
requireall nodes operatingsimultaneously.
ليست هناك تعليقات:
إرسال تعليق