Wireless Sensor Networks (WSNs)cgi.di.uoa.gr/~istavrak/courses/WSN-2005-notes.pdf · wireless sensor network ... sensor network and therefore they must be taken into consideration

Slide 1

Wireless Sensor Wireless Sensor Networks (WSNs)Networks (WSNs)

Slide 2

2005 Prof. I Stavrakakis 2

Technological Revolution

1. Computer NetworkingLANInternet

2. Wireless CommunicationsGSM/UMTSWLAN

3. Wireless Sensing TechnologiesMEMS TechnologyWSNs

1990

2000

2010

During the decade of the 90’s research on computer networking has

evolved to a big technological breakthrough with many consequences

on the social and economical worlds, which can be summarized under

the term ‘Internet’.

During the first decade of the 21st century it became apparent, that

wireless connectivity was a major concern for the costumers. The

demand for wireless telephony was enormous and thus, a new market

boomed suddenly(GSM/UMTS). However, the demand for mobility of

connected users is intense also for data communications, thus new

standards for wireless networking were defined (802.11, HIPERLAN,

etc.)

Nowadays, technological evolution is speeding up due to the increased

use of computers in research and development. The field of micro-

electro-mechanical-systems (MEMS) has shown great advances

recently and, combined together with wireless communications and

digital electronics, has enabled the development of low-cost, low-

power, multifunctional sensor nodes that are small in size and

communicate untethered in short distances. The possibility of these

sensor nodes to be networked together over a wireless medium, and to

provide, through collaborative effort, an overall result of their sensing

functionality is raising a whole, new field of research for networking

engineers and researchers (WSNs).

Slide 3


Applications for Wireless Sensor Networksoo Military ApplicationsMilitary Applications

(monitoring friendly forces, monitoring equipment, battlefield surveillance, reconnaissance of opposing forces and terrain)

oo Environmental MonitoringEnvironmental Monitoring(flood/forest fire detection, space exploration, biological attack detection))

oo Commercial ApplicationsCommercial Applications((home/office smart environments, health applications. environmental control in buildings)

oo TrackingTracking(targeting in intelligent ammunition, tracking of doctors and patients inside a hospital)

These are some of the possible uses of sensor networks. Generally, it

is assumed that sensor networks will be ubiquitous in the future,

because they will provide new possibilities for the interaction of

humans with their physical world. Sensor networks are going to be

inside houses, offices, hospitals and in the military. Furthermore,

space exploration is a field, to which WSNs are in position to

contribute, because they can be sent there, where people themselves

can not go. It is generally accepted, that WSN technology will be the

groundbreaking technology of the next decade.

Slide 4


Application Examples

Military applications include hostile environment monitoring,

friendly or hostile forces tracking and surveillance applications. The

main concept is that the sensor network is dropped from an aeroplane

or other flying object to the field of interest, and there the sensors

organize themselves as appropriate in order to fulfil the assigned

tasks. The user of the sensor network is in a remote place. Ideally, the

network would provide the possibility of task re-assignment from the

user.

Health applications for WSNs are also very important, because they

can revolutionize the way patients are treated. For example, the

organization of a big hospital may change completely, if the doctors

carry little sensors on them to provide tracking of personnel. Also,

patients may be provided with new technologies for monitoring.

Environmental monitoring is a major application driver for wireless

sensor networks. Well studied examples in literature include animal

target tracking, forest detection and flood detection. Many different

scenarios are possible.

Commercial applications are expected to emerge as soon as sensor

networks are fully functional. Seismic activity monitoring and smart

environment applications are two important examples. The first one

will introduce new methods of research for geologists, and one can

only hope that they will help for more accurate prediction of seismic

activity. Smart environments are expected to be user-interactive,

integrated and ubiquitous.

Finally, Control and Automation are a field, which sensors have

already penetrated. However, the use of sensor networks will improve

their overall performance and provide new methodologies for

production.

Slide 5


WSN Model Terminology1.1. SensorsSensors

Make discrete, local samples (measurements) of the phenomenon Communicate over wireless medium, forming a wireless sensor networkDisseminate information about the phenomenon to the observer

2.2. ObserverObserverIs interested in measuring/ monitoring the behaviour of a phenomenonAccepts measurements under specific performance requirements (accuracy or delay)

3.3. PhenomenonPhenomenonEntity of interest to the observer

A model for WSN is made of the above three entities (Individual

sensors forming a sensor network, the observer and the phenomenon).

Each entity of the model interacts with the other entities in ways,

which are listed as bullets:

Sensors are the devices that implement the physical sensing of

environmental phenomena and the reporting of measurements

through wireless communication. For reporting their measurements,

individual sensors organize themselves into a sensor network,

exchange sensor readings and disseminate information as needed to

the observer. The measurements taken by the sensors are discrete

samples of the physical phenomenon subject to individual sensor

measurement accuracy as well as location with respect to the

phenomenon

The observer is the end user interested in information about the

phenomenon. The observer may indicate interests (or queries) to the

network and receive responses to these queries. Multiple observers

may exist in a sensor network.

The phenomenon is being sensed and potentially analyzed/ filtered by

the sensor network. Multiple phenomena may be under observation

concurrently in the same network.

Slide 6


System ArchitectureCheap, lowCheap, low--power, tiny power, tiny sensors used in sensors used in thousandsthousandsCommunication with Communication with the use of miniaturized the use of miniaturized wirelesswireless transceiverstransceiversData aggregationData aggregationduring data during data propagation or at the propagation or at the sinksinkUnattendedUnattended operation operation of the sensor networkof the sensor networkSink transmits data to Sink transmits data to the endthe end--user at the user at the other endother end of the worldof the world

Internet,Satellite,etc.

SINK

SINK

USER

WSN

WSN

The sensor field is situated far from the user. The sensor network is

supposed to be self-organizing, meaning that there is no need for a

network engineer to set it up. During network operation, many sensor

nodes will die, due to lack of energy or due to other reasons. However,

human intervention is not possible at the sight, so the network must

be able to re-configure itself, so that the operation as a whole will

continue.

In the above figure we can see the nodes being connected via an

aggregation tree. The aggregation tree is representing the information

gathering and data fusion procedures. These functionalities are

essential for any WSN, because they contribute vastly to the overall

energy savings. They are an inherit part of the routing protocol of the

sensor network and therefore they must be taken into consideration

during the network-layer protocol design.

The sinks may be base stations, located near the sensor field, or well-

equipped nodes of the sensor network, which can connect via satellite

or other means to the Internet. The user can collect through the sinks

the requested data about the phenomenon taking place at the sensor

field. In particular, the user obtains an aggregated view of the sensor

field, where the aggregation points are usually the sinks.

Preferably the user should be able to interact with the sensor network

at the field, that means there should be established a bi-directional

link between sinks and user. That is essential for the re-assignment of

tasks or other instructions.

Slide 7


Sensors Hardware Platform

Node Node characteristicscharacteristics

•• Tiny sizeTiny size•• Low powerLow power•• Low bit rateLow bit rate•• High densitiesHigh densities•• Low cost Low cost

(dispensable)(dispensable)•• AutonomousAutonomous•• AdaptiveAdaptive

Power Unit

Sensor,A/D

Converter

CPU,Memory

DigitalTransceiver

PowerGenerator

Location Finding System Mobilizer

Real world data To user

Sensors are made up of four basic components: a sensing unit, a

processing unit, a transceiver unit and a power unit. They may also

have application dependent additional components such as a location

finding system, a power generator and a mobilizer.

Sensing units are usually composed of two subunits: sensors and

analog to digital converters (ADCs). The analog signals produced by

the sensors based on the observed phenomenon are converted to

digital signals by the ADC, and then fed into the processing unit.

The processing unit manages the procedures that make the sensor

node collaborate with the other nodes to carry out the assigned

sensing tasks.

A transceiver unit connects the node to the network.

One of the most important components of the sensor node is the

power unit. Power units may be supported by a power scavenging unit

such as solar cells.

It is common that a sensor node has a location finding system,

because most of the sensor network routing techniques and sensing

tasks require the knowledge of location with high accuracy.

A mobilizer may sometimes be needed to move sensor nodes when it is

required to carry out the assigned tasks.

Slide 8


Communication Architecture•• CrossCross--layer designlayer design of of

protocol stackprotocol stack•• IntegrationIntegration of routing of routing

functionality and power functionality and power awareness (energyawareness (energy--aware aware routing)routing)

•• IntegrationIntegration of routing of routing functionality and data functionality and data transport (aggregation)transport (aggregation)

•• InclusionInclusion of mobility as a of mobility as a network control primitivenetwork control primitive

•• Promotes cooperative Promotes cooperative efforts (task management efforts (task management plane)plane)

Application LayerApplication Layer

Transport LayerTransport Layer

Network LayerNetwork Layer

Data Link LayerData Link Layer

Physical LayerPhysical Layer

Power M

anagement Plane

Power M

anagement Plane

Mobility M

anagement Plane

Mobility M

anagement Plane

Task Managem

ent PlaneTask M

anagement Plane

Depending on the sensing tasks, different types of application

software can be built and used on the application layer. The transport

layer helps to maintain the flow of data if the sensor networks

application requires it. The network layer takes care of routing the

data supplied by the transport layer. Since the environment is noisy

and sensor nodes can be mobile, the MAC protocol must be power

aware and able to minimize collision with neighbours’ broadcast. The

physical layer addresses the needs of a simple but robust modulation,

transmission and receiving techniques.

The power management plane manages how a sensor node uses its

power. For example, the sensor node may turn off its receiver after

receiving a message from one of its neighbours. This is to avoid

duplicated messages. Also, when the power level of the node is low,

the sensor node broadcasts to its neighbours that it is low in power

and can not participate in routing messages. The remaining power is

reserved for sensing. The mobility management plane detects and

registers the movement of sensor nodes. So, a route back to the user

is always maintained, and the sensor nodes can keep track of who are

their neighbour sensor nodes. Thus, the sensor nodes can balance

their power and task usage. The task management plane balances and

schedules the sensing tasks given to a specific region. Not all sensor

nodes in that region are required to perform the sensing task at the

same time. As a result, some sensor nodes perform the task more

than the others depending on their power level.

The management planes are needed, so that sensor nodes can work

together in a power efficient way, route data in a mobile sensor

network, and share resources between sensor nodes. From the whole

sensor network standpoint, it is more efficient if sensor nodes can

collaborate with each other, so the lifetime of the sensor networks can

be prolonged.

Slide 9


WSNs vs. MANETs

SimilaritiesSimilaritiesData communication over wireless mediumAd-hoc network topologyPower and bandwidth are scarce resources

WSNs and MANETs are equivalent networks build for different purposes!

Nodes in sensor networks are resource constrained. They have limited

energy and computing power. Among the existing network models

MANETs are the closest to sensor networks. MANETS and sensor

networks share the following characteristics:

Node are connected to each other by wireless communication links

Network topology is not fixed (ad-hoc)

Power and bandwidth is an expensive resource

Slide 10


WSNs vs. MANETsDifferencesDifferences

WSNs are deployed and owned by a single userSensor nodes are extremely cheap, tinydevices, not like ad-hoc network nodes (PDAs, laptops, etc.)No general purpose communication network, but a data-gathering, surveillance networkNumber of nodes several orders of magnitudehigher than MANETsEnergy and bandwidth conservation is a primary concern in WSN protocol design

Sensor networks are mainly used to collect information while MANETs

are designed for distribute computing rather than information

gathering.

Usually sensor networks are deployed by one owner while MANETs

could be run by several units without any relationship

The number of sensor nodes in sensor networks can be several orders

of magnitude higher than that of the nodes in MANETs

Unlike a node in an ad-hoc network, a node in a sensor network may

not have a unique ID.

Sensor nodes are much cheaper than nodes in an ad-hoc network and

are usually deployed in thousands.

Power resource of sensor nodes could be very limited because of their

cost and un-attendedness during their lifetime. However, nodes in a

MANET can be re-charged somehow.

Usually, the data in sensor networks are bound either down-stream to

the nodes from a sink or up-stream to a sink from nodes, while in a

MANET the data flows are irregular.

Usually, sensors are deployed once in their lifetime although they may

be re-tasked or moved to other places for various reasons. But nodes

in MANET move really in an ad hoc manner

Sensor nodes are much more limited in their computation and

communication capabilities than their MANET counterparts due to

their low cost.

Slide 11


WSNs vs. MANETsComparison Summary

YesYesNoNoLowLow--cost nodes of tiny sizecost nodes of tiny size

YesYesYesYesRobust to node failuresRobust to node failures(self(self--healing)healing)

YesYesNoNoExtreme power constraints Extreme power constraints for nodes operationfor nodes operation

YesYesYesYesAdAd--hoc deploymenthoc deployment(unattended operation)(unattended operation)

YesYesYesYesMultiMulti--hop routing protocols hop routing protocols applicableapplicable

WSNWSNMANETMANETFeaturesFeatures

Slide 12


WSNs vs. MANETsComparison Summary

YesYesNoNoInIn--network data network data processingprocessing

WSNWSNMANETMANETFeaturesFeatures

NoNoYesYesUnique global IP addressesUnique global IP addresses

YesYesYesYesMobility of nodesMobility of nodes

<1000 <1000 <100<100Node densityNode density

NoNoYesYesGeneral purpose General purpose communication networkcommunication network

Slide 13


Sensor Network Protocols Design Challenges

Energy depletionEnergy depletion is the is the main resource main resource bottleneckbottleneck

Reduce each sensor’s Reduce each sensor’s active duty cycleactive duty cycleMinimize data communicationMinimize data communication over over wireless channelwireless channel

Use computation to reduce data size (data aggregation)Communicate only network state summaries instead of actual data

Maximize total network lifetimeMaximize total network lifetimeMinimum energy routing

Extra slide to focus and make clear on the new, important challenges

that sensor networks impose on the network-layer protocol design

Slide 14


Sensor Network Protocols Design Challenges

RobustnessRobustness to dynamic environmentto dynamic environmentNetwork should be self-configuringNetwork should be self-healingNetwork should be adaptive (measure and act)

Scalable to thousandsScalable to thousands of nodesof nodesOrganize network in a Organize network in a hierarchicalhierarchical manner manner (possibly with the use of clustering)(possibly with the use of clustering)Use only Use only localizedlocalized algorithmsalgorithms; with localized ; with localized interactions between nodes interactions between nodes

Extra slide to focus and make clear on the new, important challenges

that sensor networks impose on the network-layer protocol design

Slide 15


Sensor Network Protocols Design Characteristics

Data-centric operation• Focus on application data, not

individual nodes: information gathering is the purpose of sensor networks

Traditional networks: : “What is the temperature “What is the temperature at sensor #27at sensor #27 ?? ””

Sensor Networks: : ““Where areWhere are thethe nodesnodes whose temperatureswhose temperatures

recently exceeded 30 degrees?recently exceeded 30 degrees? ””

Unlike traditional networks, a sensor node doesn’t need an identity

(e.g. address). That is, applications are unlikely to to ask the question:

‘What is the temperature at sensor #27?’ Rather, applications focus

on the data generated by sensors. Data is named by attributes and

applications request data matching certain attribute values. So, the

communication primitive in this system is a request: Where are nodes

whose temperatures recently exceeded 30 degrees? This approach

decouples data from the sensor that produced it. This allows for more

robust application design: even if sensor #27 dies, the data it

generates can be cached in other (possibly neighbouring) sensors for

later retrieval.

Slide 16


Sensor Network Protocols Design Characteristics

ApplicationApplication--specific designspecific design• WSN networks can be tailored to the

sensing task at hand• Intermediate nodes can perform

application-specific data aggregationand caching

Low energy expenditure at nodesLow energy expenditure at nodes• Use of low duty-cycled sensors• Coordinate groups of sensors to fall to

the sleep stated

Traditional networks are designed to accommodate a wide variety of

applications. On the contrary WSN networks can be tailored to the

sensing task at hand. This means that intermediate nodes can

perform application specific data aggregation and caching or informed

forwarding of requests for data. This is in contrast to routers that

facilitate node to node packet switching n traditional networks.

Traditional networks provide large bandwidths, wall power and

powerful compute elements. Sensor nodes will often be limited in one

or all of these dimensions.

Slide 17


Classification of Routing Protocols

According to route discoveryAccording to route discovery1.1. ProactiveProactive2.2. ReactiveReactive3.3. HybridHybrid

According to location awarenessAccording to location awareness1.1. Location aware routingLocation aware routing2.2. LocationLocation--less routingless routing

Proactive protocols are too expensive for WSNs in terms of storage and

bandwidth consumption. Reactive or hybrid protocols are preferred

Location awareness is too expensive for a sensor network. However,

geographic protocols are scalable. So geographic protocols without

location information are needed for WSNs

Slide 18


Classification of Routing Protocols (cont’d)

According to nodes’ participating styleAccording to nodes’ participating style1.1. Direct communicationDirect communication2.2. Flat routingFlat routing3.3. Clustering routing protocolsClustering routing protocols

SINKSINK

SINK

A.Direct communication is out of the question, since the energy

requirements grow with the diameter of the sensor network.

B. In flat routing, simple, multi-hop communication is employed for

information dissemination. Since the nodes near the sinks relay all

the traffic to the sink, their power is depleted very fast.

C. Clustered routing protocols are the most appropriate for sensor

networks, since they have many advantages:

1. Nodes need to store info about the clusterhead only!It is

scalable!

2 . Routes are easily discovered and maintained

3. Energy efficient, since data is collected and processed at the

clusterheads.

Slide 19


Sensor Network Communication Protocols

Proposed Sensor Network Proposed Sensor Network Performance Performance MetricsMetrics

Energy efficiency/system lifetimeLatencyAccuracyFault-toleranceScalability

Energy efficiency/system lifetime. As sensor nodes are battery-operated, protocols

must be energy-efficient to maximize system lifetime. System lifetime can be

measured by generic parameters such as the time until half of the nodes die or by

application-directed metrics, such as when the network stops providing the

application with the desired information about the phenomena.

Latency. The observer is interested in knowing about the phenomena within a given

delay. The precise semantics of latency are application dependent.

Accuracy. Obtaining accurate information is the primary objective of the observer,

where accuracy is determined by the given application. There is a trade-off between

accuracy, latency and energy efficiency. The given infrastructure should be adaptive

so that the application obtains the desired accuracy and delay with minimal energy

expenditure. For example, the application can either request more frequent data

dissemination from the same sensor nodes or it can direct data dissemination from

more sensor nodes with the same frequency.

Fault-tolerance: Sensors may fail due to surrounding physical conditions or when

their energy runs out. It may be difficult to replace existing sensors; the network must

be fault-tolerant such that non-catastrophic failures are hidden from the application.

Scalability: Scalability for sensor networks is also a critical factor. For large-scale

networks, it is likely that localizing interactions through hierarchy and aggregation

will be critical for ensuring scalability.

Slide 20


SPANProblemProblem: Need to : Need to minimize the minimize the

energy consumptionenergy consumption of wireless of wireless nodes in a wireless ad hoc nodes in a wireless ad hoc network!network!

IDEA:IDEA:Leverage the time the network Leverage the time the network interface of a node remains interface of a node remains idleidleto to powerpower--downdown the radio of the the radio of the node.node.

Reducing energy consumption in a wireless ad hoc network is the

primary goal of this protocol. It aims also at being completely inter-

operable with different routing protocols running in the ad hoc

network. The approach adopted here is: since the network interface

remains lots of time idle, why not power down the radio of the node

for this time. BUT: it is not always straightforward, when to turn the

radio off, since a node does not only send/receive packets when it

wants to communicate with the rest of the network, but it also

participates in the network as a relaying node.

Slide 21


SPANDesired CharacteristicsDesired Characteristics

1.1. As many nodes as possibleAs many nodes as possible should should be be in sleep modein sleep mode

2.2. Forwarding ofForwarding of packets packets should occur should occur with with minimal minimal additionaladditional delaysdelays

3.3. Awake nodesAwake nodes should provide should provide as as much total capacitymuch total capacity as original as original networknetwork

4.4. Distributed algorithmDistributed algorithm for so that for so that nodes make nodes make locallocal decisionsdecisions

SPAN is a power saving technique for multi-hop ad hoc wireless

networks that reduces energy consumption without significantly

diminishing the capacity or latency characteristics of the network.

A good power-saving coordination technique for wireless ad-hoc

networks ought to have certain characteristics:

Allow as many nodes as possible to turn their radio receivers off

most of the time, since even an idle receive circuit can consume

almost as much energy as an active transmitter.

It should forward packets between any source and destination with

minimally more delay than if all nodes were awake.

The backbone formed by the awake nodes should provide about as

much total capacity as the original network, since otherwise

congestion may increase.

The algorithm for picking this backbone should be distributed,

requiring each node to make a local decision (localised algorithm).

Slide 22


SPAN

Span is a powerSpan is a power--saving protocol that saving protocol that operates operates betweenbetween the routing layer and the routing layer and the MAC layer.the MAC layer.

802.11, H802.11, HIIPERLAN/2PERLAN/2

SpanSpan

DSRDSRAODVAODVGPSRGPSRRouting layer

MAC/Phy layer

As shown in the figure above, Span runs above the link and MAC

layers and interacts with the routing protocol. This structuring allows

Span to take advantage of power-saving features of the link layer

protocol, while still being able to affect the routing process. For

example, non-coordinator nodes can periodically turn on their radios

and listen or poll for their packets. Span leverages a feature of modern

power-saving MAC layers, in which if a node has been asleep for a

while, packets destined for it are not lost but are buffered at a

neighbor. When the node awakens, it can retrieve these packets from

the buffering node, typically a coordinator. Span also requires a

modification to the route lookup process at each node – at any time,

only those entries in a node’s routing table that correspond to

currently active coordinators can be used as valid next-hops (unless

the next hop is the destination itself).

Slide 23


SPANOperation of SPANOperation of SPAN

Certain nodes are elected as Certain nodes are elected as ‘coordinators’‘coordinators’to participate in the backbone network. to participate in the backbone network. Coordinators stay Coordinators stay alwaysalways--onon to provide to provide global connectivity of the network. The rest global connectivity of the network. The rest of nodes remain in of nodes remain in powerpower--save modesave mode and and periodically check to change statusperiodically check to change statusCoordinators are rotated among nodesCoordinators are rotated among nodesAttempt to minimize the number of Attempt to minimize the number of coordinatorscoordinatorsDistributed coordinators election processDistributed coordinators election process

Span adaptively elects “coordinators” from all nodes in the network.

Span coordinators stay awake continuously and perform multi-hop

packet routing within the ad hoc network, while other nodes remain in

power-saving mode and periodically check if they should wake up and

become a coordinator.

Span achieves four goals.

First, it ensures that enough coordinators are elected so that every

node is in radio range of at least one coordinator.

Second, it rotates the coordinators in order to ensure that all nodes

share the task of providing global connectivity roughly equally.

Third, it attempts to minimize the number of nodes elected as

coordinators, thereby increasing network lifetime, but without

suffering a significant loss of capacity or an increase in latency.

Fourth, it elects coordinators using only local information in a

decentralized manner – each node only consults state stored in local

routing tables during the election process.

Slide 24


SPANSpan is Span is proactiveproactive: each node : each node periodicallyperiodicallybroadcastsbroadcasts HELLOHELLO messages:messages:

1.1. the node’s statusthe node’s status2.2. its current coordinatorsits current coordinators3.3. its current neighborsits current neighbors

From the HELLO messages each node From the HELLO messages each node buildsbuilds

1.1. a list of own neighbors and a list of own neighbors and coordinatorscoordinators

2.2. for each neighbor: a list of its for each neighbor: a list of its neighbors and coordinatorsneighbors and coordinators

Span is proactive: each node periodically broadcasts HELLO messages

that contain the node’s status (i.e., whether or not the node is a

coordinator), its current coordinators, and its current neighbors. From

these HELLO messages, each node constructs a list of the node’s

neighbors and coordinators, and for each neighbor, a list of its

neighbors and coordinators.

Slide 25


SPANCoordinator announcementCoordinator announcement

Regular nodes Regular nodes periodicallyperiodically wake up and wake up and decide to become decide to become coordinatorscoordinators or not based on or not based on a a coordinator eligibility rulecoordinator eligibility rule

Coordinator eligibility ruleCoordinator eligibility ruleA nonA non--coordinator node should become a coordinator if coordinator node should become a coordinator if it discovers, using only information gathered from local it discovers, using only information gathered from local broadcast messages, that two of its neighbors cannot broadcast messages, that two of its neighbors cannot reach each other either directly or via one or two reach each other either directly or via one or two coordinatorscoordinators

Coordinator announcement

Periodically, a non-coordinator node determines if it should become a

coordinator or not. The following coordinator eligibility rule in Span

ensures that the entire network is covered with enough coordinators:

Coordinator eligibility rule. A non-coordinator node should become

a coordinator if it discovers, using only information gathered from

local broadcast messages, that two of its neighbors cannot reach each

other either directly or via one or two coordinators.

This election algorithm does not yield the minimum number of

coordinators required to merely maintain connectedness. However, it

roughly ensures that every populated radio range in the entire

network contains at least one coordinator. Because packets are routed

through coordinators, the resulting coordinator topology should yield

good capacity.

This election algorithm does not yield the minimum number of

coordinators required to merely maintain connectedness. However, it

roughly ensures that every populated radio range in the entire

network contains at least one coordinator. Because packets are routed

through coordinators, the resulting coordinator topology should yield

good capacity.

Slide 26


SPANContention resolutionContention resolution

What happens if two nodes decide to become What happens if two nodes decide to become coordinators at the same time?coordinators at the same time?

Introduce a Introduce a randomized randomized backoffbackoff delay delay at each at each node, based onnode, based on

Nodes with Nodes with roughly equal remaining energyroughly equal remaining energyNNii: number of : number of neighborsneighbors at node iat node iCCii: number of additional pairs of nodes to be : number of additional pairs of nodes to be connected if i became a coordinatorconnected if i became a coordinator

0 ≤ Ci ≤ (Ni ov. 2)Define as utilityutility of a node i: of a node i: CCii / (N/ (Nii ovov. 2). 2)

Announcement contention occurs when multiple nodes discover the

lack of a coordinator at the same time, and all decide to become a

coordinator. Span resolves contention by delaying coordinator

announcements with a randomized backoff delay. Each node chooses

a delay value, and delays the HELLO message that announces the

node’s volunteering as a coordinator for that amount of time. At the

end of the delay,

the node re-evaluates its eligibility based on HELLO messages recently

received, and makes its announcement if and only if the eligibility rule

still holds.

We consider a variety of factors in our derivation of the backoff delay.

Consider first the case when all nodes have roughly equal energy,

which implies that only topology should play a role in deciding which

nodes become coordinators.

Let Ni be the number of neighbors for node i and let Ci be the number

of additional pairs of nodes among these neighbors that would be

connected if i were to become a coordinator and forward packets.

We call Ci/Ni the utility of node i.

Slide 27



Nodes with Nodes with higher higher CCii should volunteer should volunteer more quicklymore quicklythan ones with smaller than ones with smaller CCii

the delay for each node is randomly chosen over an interval proportional to Ni x T

R picked uniformly at random from interval (0,1]

If nodes with high Ci become coordinators, fewer coordinators in total

may be needed in order to make sure every node can talk to a

coordinator; thus a node with high Ci should volunteer more quickly

than one with smaller Ci .

If there are multiple nodes within radio range that all have the same

utility, Span prevents too many of them becoming coordinators. This

is because such coordinators would be redundant – they would not

increase system capacity, but simply drain energy. If the potential

coordinators make their decisions simultaneously, they may all decide

to become coordinators. If, on the other hand, they decide one at a

time, only the first few will become coordinators, and the rest will

notice that there are already enough coordinators and go back to

sleep. To handle this, we use a randomized “slotting-and damping”

method reminiscent of techniques to avoid multiple retransmissions of

lost packets by multicast protocols, such as XTP, IGMP and SRM: the

delay for each node is randomly chosen over an interval proportional

to Ni Χ T , where T is the round-trip delay for a small packet over the

wireless link.

Slide 28



Nodes with Nodes with unequal remaining energyunequal remaining energyEErr: amount of remaining energy at a node: amount of remaining energy at a nodeEEmm: maximum amount of energy available: maximum amount of energy available

Fairness ruleFairness ruleA node with A node with larger larger EErr/E/Emm should become should become

coordinator coordinator more quickly more quickly

Consider the case when nodes may have unequal energy left in their

batteries. We observe that what matters in a heterogeneous network is

not necessarily the absolute amount of energy available at the node,

but the amount of energy scaled to the maximum amount of energy

that the node can have. Let Er denote the amount of energy at a node

that still remains, and Em be the maximum amount of energy

available

at the same node. A reasonable (but not the only) notion of fairness

can be achieved by ensuring that a node with a larger value of Er/Em

is more likely to volunteer to become a coordinator more quickly than

one with a smaller ratio. Thus, we need to add a decreasing function

of Er/Em that reflects this, to equation (1). There are an infinite

number of such functions, from which we choose a simple linear one:

1−Er/Em.

Observe that the first term does not have a random component; thus

if a node is running low on energy, its propensity to become a

volunteer is guaranteed to diminish relative to other nodes in the

neighborhood with similar neighbors.

In a network with uniform density and energy, our election algorithm

rotates coordinators among all nodes of the network. It achieves

fairness because the likelihood of becoming a coordinator falls as a

coordinator uses up its battery. In practice, however, ad hoc networks

are rarely uniform. Our announcement rule adapts to non-uniform

topology: a node that connects network partitions together will always

be elected

a coordinator. This property preserves capacity over the lifetime of the

network. Because of Span’s emphasis on capacitypreservation to the

extent possible, such critical nodes will unavoidably die before other

less-critical ones. However, in a mobile Span network, a given node is

rarely stuck in such a position, and this improves fairness

dramatically.

Slide 29


SPANCoordinators withdrawalCoordinators withdrawal

Each coordinator Each coordinator periodicallyperiodically checks if it should checks if it should withdraw as a coordinatorwithdraw as a coordinatorRule to withdrawRule to withdraw: every pair of its : every pair of its neighborsneighborsshould be able to reach each other either should be able to reach each other either directlydirectlyor via or via one or twoone or two other coordinatorsother coordinatorsTo rotate coordinators among all nodes fairly: use To rotate coordinators among all nodes fairly: use of of tentativetentative coordinatorscoordinatorsTentative coordinators: Tentative coordinators: provide the chance for provide the chance for nonnon--coordinators to become coordinatorscoordinators to become coordinatorsCoordinators Coordinators stay tentativestay tentative for Wfor WTT amount of timeamount of time

WWTT= 3 x N= 3 x Nii x T (max. delay for cont. resolution)x T (max. delay for cont. resolution)After WAfter WT T , the tentative bit is removed, the tentative bit is removed

Coordinator withdrawal

Each coordinator periodically checks if it should withdraw as a

coordinator. A node should withdraw if every pair of its neighbors can

reach each other either directly or via one or two other coordinators.

In order to also rotate the coordinators among all nodes fairly, after a

node has been a coordinator for some period of time, it marks itself as

a tentative coordinator if every pair of neighbor nodes can reach each

other via one or two other neighbors, even if those neighbors are not

currently coordinators. A tentative coordinator can still be used to

forward packets. However, the coordinator announcement algorithm

described above treats a tentative coordinator as a non-coordinator.

Thus, by marking itself as tentative, a coordinator gives its neighbors

a chance to become coordinators.

A coordinator stays tentative for WT amount of time, where WT is the

maximum value of equation (2). That is, WT = 3 Χ Ni Χ T. (3) If a

coordinator has not withdrawn afterWT , it clears its tentative bit. To

prevent an unlucky low energy node from draining all of its energy

once it becomes a coordinator, the amount of time a node stays as a

coordinator before turning on its tentative bit is proportional to the

amount of energy it has (Er/Em).

Slide 30


SPANIllustration of SPAN Illustration of SPAN algalg. at some arbitrary . at some arbitrary momentmoment

+: non-coordinator nodes

*: coordinator nodes

Solid lines: connect neighboring coordinators

Figure shows the result of our election algorithmat a random point in

time on a network of 100 nodes in a 1000 m Χ 1000 m area, where

each radio has an isotropic circular range with a 250 m radius. Solid

lines connect coordinators that are within radio range of each other.

A scenario with 100 nodes, 19 coordinators, and a radio range of 250

m. The nodes marked “�” are coordinators; the nodes marked “+” are

non-coordinator nodes. Solid lines connect coordinators that are

within radio range of each other.

Slide 31


SPANEnergy consumption characteristicsEnergy consumption characteristics

per-node power usage in networksrunning Span, 802.11 PSM, and 802.11

This section evaluates Span’s ability to save energy. The potential for

savings depends on node density, since the fraction of sleeping nodes

depends on the number of nodes per radio coverage area. The energy

savings also depend on a radio’s power consumption in sleep mode

and the amount of time that sleeping nodes must turn on their

receivers to listen for 802.11 beacons and Span HELLO messages.

Figure shows the per-node power usage in networks running Span,

802.11 PSM, and 802.11. These numbers are calculated from the

initial energy and the energy remaining at each of the 100 mobile

nodes over 500 s. Each value is an average over 5 mobile simulations.

From these results, we find that Span provides a considerable amount

of energy savings over 802.11, while 802.11 PSM saves essentially no

power. This is because geographic forwarding needs to send broadcast

messages. With 802.11 PSM, each time a node receives a broadcast

advertisement, it must stay up for the entire beacon period. This

prevents non-coordinators from going back to sleep. When the node

density is low, the number of broadcast messages in a radio range

decreases, and 802.11 PSM yields a small amount of energy savings.

Slide 32


SPANProsPros

Achieves high energyAchieves high energy--savings, even with savings, even with regular ad hoc routing protocolsregular ad hoc routing protocolsSlow increase of energy savings with higher Slow increase of energy savings with higher network densities due to periodicitynetwork densities due to periodicityLow latency, low throughput degradationLow latency, low throughput degradation

ConsConsCan not be applied to sensor networks, Can not be applied to sensor networks, because sensing nodes may not be powered because sensing nodes may not be powered up or downup or downHigh communication overheadHigh communication overhead

Span leverages a feature of modern power-saving MAC layers, in

which if a node has been asleep for a while, packets destined for it are

not lost but are buffered at a neighbor.

Performance evaluation shows that Span not only preserves network

connectivity, it also preserves capacity, decreases latency and provides

significant energy savings. E.g. for a practical range of node densities

and a practical energy model, the system lifetime with span is more

than a factor of two better than without span!

Slide 33


LEACH

LLow ow EEnergy nergy AAdaptive daptive CClustering lustering HHierarchyierarchyA clustering-based protocol utilizing randomized rotation of local cluster base stations (cluster-heads) to evenly distribute the energy load among the sensors in the networkLEACH makes the following assumptions:

1. The base station is fixed and located far from the sensors

2. All nodes in the network are homogeneous and energy-constrained

Sensor networks can contain hundreds or thousands of sensing

nodes. It is desirable to make these nodes as cheap and energy-

efficient as possible and rely on their large numbers to obtain high

quality results. Network protocols must be designed to achieve fault

tolerance in the presence of individual node failure while minimizing

energy consumption. In addition, since the limited wireless channel

bandwidth must be shared among all the sensors in the network,

routing protocols for these networks should be able to perform local

collaboration to reduce bandwidth requirements.

Eventually, the data being sensed by the nodes in the network must

be transmitted to a control center or base station, where the end-user

can access the data.

Communication between the sensor nodes and the base station is

expensive, and there are no “high-energy” nodes through which

communication can proceed.

Slide 34


LEACH

Key features of LEACHKey features of LEACH: : Localized coordination and control for cluster set-up and operationRandomized rotation of the cluster “base stations” or “cluster-heads” and the corresponding clustersLocal compression to reduce global communication

By analyzing the advantages and disadvantages of conventional

routing protocols using our model of sensor networks, we have

developed LEACH (Low-Energy Adaptive Clustering Hierarchy), a

clustering-based protocol that minimizes energy dissipation in sensor

networks.

The use of clusters for transmitting data to the base station leverages

the advantages of small transmit distances for most nodes, requiring

only a few nodes to transmit far distances to the base station.

However, LEACH outperforms classical clustering algorithms by using

adaptive clusters and rotating cluster-heads, allowing the energy

requirements of the system to be distributed among all the sensors. In

addition, LEACH is able to perform local computation in each cluster

to reduce the amount of data that must be transmitted to the base

station. This achieves a large reduction in the energy dissipation, as

computation is much cheaper than communication.

Sensor networks contain too much data for an end-user to process.

Therefore, automated methods of combining or aggregating the data

into a small set of meaningful information is required. In addition to

helping avoid information overload, data aggregation, also known as

data fusion, can combine several unreliable data measurements to

produce a more accurate signal by enhancing the common signal and

reducing the uncorrelated noise. The classification performed on the

aggregated data might be performed by a human operator or

automatically. Both the method of performing data aggregation and

the classification algorithm are application-specific.

Slide 35


LEACHProtocol descriptionProtocol description

Nodes organize themselves into local clusters, with one node acting as local base station or “cluster-head”Randomized rotation of high-energy cluster-head position so as not to ‘drain’ the energy of a single nodeElection of clusterheads at any given time with a certain probabilitySensors choose their preferred clusterhead to belong to, based on the minimum required energy to communicate withClusterheads create schedules for the nodes in their cluster, so that plain nodes can power-down when they are not scheduled to transmitClusterheads aggregate data from sensors in cluster and transmit compressed data to the base station

LEACH is a self-organizing, adaptive clustering protocol that uses

randomization to distribute the energy load evenly among the sensors

in the network. In LEACH, the nodes organize themselves into local

clusters, with one node acting as the local base station or cluster-

head. If the clusterheads were chosen a priori and fixed throughout

the system lifetime, as in conventional clustering algorithms, it is easy

to see that the unlucky sensors chosen to be cluster-heads would die

quickly, ending the useful lifetime of all nodes belonging to those

clusters. Thus LEACH includes randomized rotation of the high-

energy cluster-head position such that it rotates among the various

sensors in order to not drain

the battery of a single sensor.

Sensors elect themselves to be local cluster-heads at any given time with a certain

probability. These clusterhead nodes broadcast their status to the other sensors in the

network. Each sensor node determines to which cluster it wants to belong by choosing

the cluster-head that requires the minimum communication energy. Once all the nodes

are organized into clusters, each cluster-head creates a schedule for the nodes in its

cluster. This allows the radio components of each non-cluster-head node to be turned

off at all times except during its transmit time, thus minimizing the energy dissipated

in the individual sensors. Once the cluster-head has all the data from the nodes in its

cluster, the cluster-head node aggregates the data and then transmit the compressed

data to the base station. Since the base station is far away in the scenario we are

examining, this is a high energy transmission. However, since there are only a few

cluster-heads, this only affects a small number of nodes.

As discussed previously, being a cluster-head drains the battery of that node. In order

to spread this energy usage over multiple nodes, the cluster-head nodes are not fixed;

rather, this position is self-elected at different time intervals. Thus a set C of nodes

might elect themselves cluster-heads at time t, but at time t d a new set C of nodes

elect themselves as cluster-heads. The decision to become a cluster-head depends on

the amount of energy left at the node. In this way, nodes with more energy remaining

will perform the energy-intensive functions of the network. Each node makes its

decision about whether to be a cluster-head independently of the other nodes in the

Network and thus no extra negotiation is required to determine the

clusterheads . The system can determine, a priori, the optimal

number of clusters to have in the system. This will depend on several

parameters, such as the network topology and the relative costs of

computation versus communication.

Slide 36


LEACHLEACH operates in LEACH operates in consecutive roundsconsecutive roundsClusterheadsClusterheads areare elected newelected new at at each round of each round of operationoperation

C: set of clusterheads

at time t0

For even energy dissipation

Slide 37


LEACH

NewNew set of clusterheads C`set of clusterheads C` for the next for the next round round

C`: set of clusterheads

at time t0 + δ0

Slide 38


LEACHPhases of operationPhases of operation

1. Advertisement Phase• Clusterheads are elected in this phase• Election is based on P (percentage of clusterheads for

the network) and the number of times the node has been a clusterhead so far

• Node n chooses a random number between 0 and 1 and if this number is less than a threshold T(n), the node becomes clusterhead in this round

• Clusterheads broadcast advertisement messages using CSMA MAC protocol using the same energy

• Receiving nodes decide which clusterehad to belong to based on the received advertisement signal strength

2. Cluster Set-up Phase• Nodes inform the clusterheads that they want to join

their cluster• Again a CSMA MAC protocol is used

Advertisement phase

Initially, when clusters are being created, each node decides whether

or not to become a cluster-head for the current round. This decision is

based on the suggested percentage of cluster heads for the network

(determined a priori) and the number of times the node has been a

cluster-head so far. This decision is made by the node n choosing a

random number between 0 and 1. If the number is less than a

threshold Tn, the node becomes a cluster-head for the current round.

Each node that has elected itself a cluster-head for the current round broadcasts an

advertisement message to the rest of the nodes. For this “cluster-head-advertisement”

phase, the cluster-heads use a CSMA MAC protocol, and all cluster-heads transmit

their advertisement using the same transmit energy. The non-cluster-head nodes must

keep their receivers on during this phase of set-up to hear the advertisements of all the

cluster-head nodes. After this phase is complete, each non-cluster-head node decides

the cluster to which it will belong for this round. This decision is based on the

received signal strength of the advertisement. Assuming symmetric propagation

channels, the cluster-head advertisement heard with the largest signal strength is the

cluster-head to whom the minimum amount of transmitted

Cluster set-up phase

After each node has decided to which cluster it belongs, it must inform the cluster-

head node that it will be a member of the cluster. Each node transmits this information

back to the cluster-head again using a CSMA MAC protocol. During this phase, all

cluster-head nodes must keep their receivers

Slide 39


LEACHPhases of operationPhases of operation

3. Schedule Creation Phase• Clusterheads receive all messages from nodes to be

included in cluster• Based on the number of nodes in the cluster,

clusterhead creates TDMA schedule• Schedule is broadcast to all cluster nodes

4. Data Transmission Phase• Assuming nodes have data to send, they wait for

their allocated time to send data to the clusterhead• The rest of the time they power down their radio to

conserve energy• Clusterhead performs data fusion so as to send

compressed data to the sink• This final transmission is a high-energy data

transmission

Schedule creation

The cluster-head node receives all the messages for nodes that would like to be

included in the cluster. Based on the number of nodes in the cluster, the cluster-head

node creates a TDMA schedule telling each node when it can transmit. This schedule

is broadcast back to the nodes in the cluster.

Data transmission

Once the clusters are created and the TDMA schedule is fixed, data transmission can

begin. Assuming nodes always have data to send, they send it during their allocated

transmission time to the cluster head. This transmission uses a minimal amount of

energy (chosen based on the received strength of the cluster-head advertisement). The

radio of each non-cluster-head node can be turned off until the node’s allocated

transmission time, thus minimizing energy dissipation in these nodes. The cluster-

head node must keep its receiver on to receive all the data from the nodes in the

cluster. When all the data has been received, the cluster head node performs signal

processing functions to compress the data into a single signal. For example, if the data

are audio or seismic signals, the cluster-head node can beamform the individual

signals to generate a composite signal. This composite signal is sent to the base

station. Since the base station is far away, this is a high-energy transmission.

This is the steady-state operation of LEACH networks. After a certain

time, which is determined a priori, the next round begins with each

node determining if it should be a cluster-head for this round and

advertising this information.

Slide 40


LEACHNormalized total system energy dissipated versus the percent of nodes that are cluster-heads.

Optimal point of LEACH operation

Over a factor of 7 for reduction in energy dissipation when optimal number of clusterheads

Normalized total system energy dissipated versus the percent of nodes

that are cluster-heads. Note that direct transmission is equivalent to 0

nodes being cluster-heads or all the nodes being cluster-heads.

The system can determine, a priori, the optimal number of clusters to

have in the system. This will depend on several parameters, such as

the network topology and the relative costs of computation versus

communication. Figure shows how the energy dissipation in the

system varies as the percent of nodes that are cluster-heads is

changed. Note that 0 cluster-heads and 100% clusterheads is the

same as direct communication.

From this plot, we find that there exists an optimal percent of nodes N

that should be cluster-heads. If there are fewer than N clusterheads,

some nodes in the network have to transmit their data very far to

reach the cluster-head, causing the global energy in the system to be

large. If there are more than N clusterheads, the distance nodes have

to transmit to reach the nearest cluster-head does not reduce

substantially, yet there are more cluster-heads that have to transmit

data the long-haul distances to the base station, and there is less

compression being performed locally. For our system parameters and

topology, N is roughly equal to 5%.

Figure also shows that LEACH can achieve over a factor of 7 reduction

in energy dissipation compared to direct communication with the base

station, when using the optimal number of cluster-heads. The main

energy savings of the LEACH protocol is due to combining lossy

compression

with the data routing. There is clearly a trade-off between the quality

of the output and the amount of compression achieved. In this case,

some data from the individual signals is lost, but this results in a

substantial reduction of the overall energy dissipation of the system.

Slide 41


LEACH

Up to 8x reduction in energy dissipation between LEACH and conventional routing protocols

Total system energy dissipated using direct communication, MTE and LEACH for a 100-node random network

Total system energy dissipated using direct communication, MTE and

LEACH for a 100-node random network.

Direct communication: All sensor nodes communicate directly with

base station

MTE: “Minimum Transmission Energy”, power-aware protocol, where

nodes route data destined ultimately for the base station through

intermediate nodes. Thus nodes act as routers for other nodes’ data in

addition to sensing the environment. These protocols differ in the way

the routes are chosen. In this case the intermediate nodes are chosen

such that the transmit amplifier energy is minimized.

The figure shows how these algorithms compare with LEACH using

energy dissipated per bit as Eelec=50nJ/bit. The plot shows that

LEACH achieves between 7x and 8x reduction in energy compared

with direct communication and between 4x and 8x reduction in

energy compared with MTE routing.

Slide 42


LEACH

LEACH’sLEACH’s strengthsstrengths•• Localised coordination of clustersLocalised coordination of clusters•• Randomized rotation of the Randomized rotation of the

clusterheadsclusterheads•• Scalable due to clustering hierarchyScalable due to clustering hierarchy•• EnergyEnergy--efficient due to the combination efficient due to the combination

of data compression and routingof data compression and routing

Slide 43


LEACH

LEACH’sLEACH’s weaknessesweaknesses•• Presence of a Presence of a hot spothot spot can deplete can deplete the power of nodes in its vicinity the power of nodes in its vicinity very quicklyvery quickly

•• Some sensors may not be able to Some sensors may not be able to power down due to their assigned power down due to their assigned taskstasks

Slide 44


SPINAdaptive Protocols for Information Adaptive Protocols for Information Dissemination in Wireless Sensor NetworksDissemination in Wireless Sensor Networks

Family of adaptive protocols called SPIN for efficient dissemination of information in energy-constrained wireless sensor network

SPIN characteristicsSPIN characteristicsIntroduction of high-level data descriptors (use of meta-data)Use of meta-data negotiation to eliminate transmission of redundant informationNodes base communication decisions upon application-specific knowledge and knowledge of the resources that are available to them

We present a family of adaptive protocols, called SPIN (Sensor

Protocols for Information via Negotiation), that efficiently disseminates

information among sensors in an energy-constrained wireless sensor

network. Nodes running a SPIN communication protocol name their

data using high-level data descriptors, called meta-data. They use

meta-data negotiations to eliminate the transmission of redundant

data throughout the network. In addition, SPIN nodes can base their

communication decisions both upon application-specific knowledge of

the data and upon knowledge of the resources that are available to

them. This allows the sensors to efficiently distribute data given a

limited energy supply. We simulate and analyze the performance of

two specific SPIN protocols, comparing them to other possible

approaches and a theoretically optimal protocol.

Slide 45


SPINAnalysis of problems characterizing Analysis of problems characterizing conventional protocols for data conventional protocols for data dissemination in a sensor network:dissemination in a sensor network:

1. Implosion2. Overlap3. Resource blindness

SPIN solutions:SPIN solutions:1. Negotiation2. Resource adaptation

The design of SPIN grew out of an analysis of the different strengths and limitations

of conventional classic flooding protocols for disseminating data in a sensor network.

In classic flooding, each node keeps a record containing a list of all the data that it has

sent to its neighbours. The protocol begins when a source node sends its data to all of

its neighbours. Upon receiving a piece of data, each node stores the data and checks

the record to see whether it has already forwarded the data to its neighbour. If not, it

forwards a copy of the data to all of its neighbours and updates the record. This is

therefore a straightforward protocol requiring only a small amount of protocol state at

any node, and it disseminates data quickly in a network where bandwidth is not scarce

and links are not loss-prone.

The SPIN family of protocols incorporates two key innovations that

overcome these deficiencies: negotiation and resource-adaptation.

To overcome the problems of implosion and overlap, SPIN nodes

negotiate with each other before transmitting data. Negotiation helps

ensure that only useful information will be transferred. To negotiate

successfully, however, nodes must be able to describe or name the

data they observe. We refer to the descriptors used in SPIN

negotiations as meta-data. In SPIN, nodes poll their resources before

data transmission. Each sensor node has its own resource manager

that keeps track of resource consumption; applications probe the

manager before transmitting or processing data. This allows sensors

to cut back on certain activities when energy is low, e.g., by being

more prudent in forwarding third-party data.

Together, these features overcome the three deficiencies of classic

flooding. The negotiation process that precedes actual data

transmission eliminates implosion because it eliminates transmission

of redundant data messages. The use of meta-data descriptors

eliminates the possibility of overlap because it allows nodes to name

the portion of the data that they are interested in obtaining. Being

aware of local energy resources allows sensors to cut back on

activities whenever their energy resources are low, thereby extending

longevity.

1.2. The solution

The SPIN family of protocols incorporates two key innovations that overcome these

deficiencies: negotiation and resource-adaptation. To overcome the problems of

implosion and overlap, SPIN nodes negotiate with each other before transmitting data.

Negotiation helps ensure that only useful information will be transferred. To negotiate

successfully, however, nodes must be able to describe or name the data they observe.

We refer to the descriptors used in SPIN negotiations as meta-data.

In SPIN, nodes poll their resources before data transmission. Each sensor node has its

own resource manager that keeps track of resource consumption; applications probe

the manager before transmitting or processing data. This allows sensors to cut back on

certain activities when energy is low, e.g., by being more prudent in forwarding third-

party data.

It also allows sensors to take resource tradeoffs into account when making decisions.

For example, a SPIN node may decide to send a piece of data unconditionally,

without any negotiation, if it believes that the associated costs of sending the data are

less than the costs of negotiating for it. Together, these features can help SPIN nodes

overcome the three deficiencies of classic flooding. The negotiation process that

precedes actual data transmission eliminates implosion because it eliminates

transmission of redundant data messages. The use of meta-data descriptors eliminates

the possibility of overlap because it allows nodes to name the portion of the data that

they are interested in obtaining. Being aware of local energy resources allows sensors

to make prudent decisions about using these resources, thereby extending longevity.

Exchanging sensor data may be an expensive network operation, but exchanging data

about sensor data need not be. Second, nodes in a network must monitor and adapt to

changes in their own energy resources to extend the operating lifetime of the system.

This section presents the individual features that make up the SPIN family of

protocols.

Slide 46


SPINImplosion problem Overlap problem

Figure 1 Figure 2

Figure 1. The implosion problem. In this graph, node A starts by flooding its data to

all of its neighbors. Two copies of the data eventually arrive at node D. The system

wastes energy and bandwidth in one unnecessary send and receive.

1. Implosion. In classic flooding, a node always sends data to its neighbours,

regardless of whether or not the neighbour has already received the data from another

source. This leads to the implosion problem.

Here, node A starts out by flooding data to its two neighbours, B and C. These nodes

store the data from A and send a copy of it on to their neighbour D. The protocol,

thus, wastes resources by sending two copies of the data to D. It is easy to see that

implosion is linear in the degree of any node.

Figure 2. The overlap problem. Two sensors cover an overlapping geographic region.

When these sensors flood their data to node C, C receives two copies of the data

marked r.

2. Overlap. Sensor nodes often cover overlapping geographic areas, and nodes often

gather overlapping pieces of sensor data. Figure 2 illustrates what happens when two

nodes (Aand B) gather such overlapping data and then flood the data to their common

neighbor (C). Again, the algorithm wastes energy and bandwidth sending two copies

of a piece of data to the same node. Overlap is a harder problem to solve than the

implosion problem – implosion is a function only of network topology, whereas

overlap is a function of both topology and the mapping of observed data to sensor

nodes.

3. Resource blindness. In classic flooding, nodes do not modify their activities based

on the amount of energy available to them at a given time. A network of embedded

sensors can be “resource-aware” and adapt its communication and computation to the

state of its energy resources.

Slide 47


SPIN: Sensor Protocol for Information via Negotiation

Two basic ideas:Two basic ideas:1. sensor applications need to communicate with

each other about the data that they already have and the data they still need to obtain

2. nodes in a network must monitor and adaptto changes in their own energy resources to extend the operating lifetime of the system

MetaMeta--data:data:If x is the meta-data descriptor for sensor data X, then size of x < size of X for SPIN to be efficient

The SPIN family of protocols rests upon two basic ideas. First, to operate efficiently and to

conserve energy, sensor applications need to communicate with each other about the data

that they already have and the data they still need to obtain. Exchanging sensor data may be

an expensive network operation, but exchanging data about sensor data need not be. Second,

nodes in a network must monitor and adapt to changes in their own energy resources to

extend the operating lifetime of the system.

Our design of the SPIN protocols is motivated in part by the principle of Application Level

Framing (ALF). With ALF, network protocols must choose transmission units that are

meaningful to applications, i.e., packetization is best done in terms of Application Data Units

(ADUs). One of the important components of ALF-based protocols is the common data naming

between the transmission protocol and application, which we follow in the design of our meta-

data. We take ALF-like ideas one step further by arguing that routing decisions are also best

made in application-controlled and application-specific ways, using knowledge of not just

network topology but application data layout and the state of resources at each node. We

believe that such integrated approaches to naming and routing are attractive to a large range

of network situations, especially in mobile and wireless networks of devices and sensors.

Sensors use meta-data to succinctly and completely describe the data that they collect. If x is

the meta-data descriptor for sensor data X, then the size of x in bytes must be shorter than

the size of X, for SPIN to be beneficial. If two pieces of actual data are distinguishable, then

their corresponding meta-data should be distinguishable. Likewise, two pieces of

indistinguishable data should share the same meta-data

representation.

SPIN does not specify a format for meta-data; this format is application-specific. Sensors that

cover disjoint geogTaphic regions may simply use their own unique IDS as meta-data. The

meta-data x would then stand for “all the data gathered by sensor x”. A camera sensor, in

contrast, might use (x, y, 4) as meta-data, where (z, y) is a geographic coordinate and C$ is an

orientation. Because each application’s

meta-data format may be different, SPIN relies on each application to interpret and synthesize

its own metadata. There are costs associated with the storage, retrieval, and general

management of meta-data, but the benefit of having a succinct representation for large data

messages in SPIN far outweighs these costs.

Slide 48


SPINSPIN messages:SPIN messages:

1. ADV: New Data Advertisement (meta-data)Nodes that have data to share send advertisement messages containing meta data

2. REQ: Request for Data (meta-data)Nodes wishing to receive some data, send request messages to inform the source node

3. DATA: Data message (data)This message type contains actual sensor data with a meta-data header

SPIN Messages

SPIN nodes use three types of messages to communicate:

ADV - new data advertisement. When a SPIN node has data to share,

it can advertise this fact by transmitting an ADV message containing

meta-data.

REQ - request for data. A SPIN node sends an REQ message when it

wishes to receive some actual data.

DATA - data message. DATA messages contain actual sensor data

with a meta-data header.

Because ADV and REQ messages contain only meta data, they are

smaller, and cheaper to send and receive, than their corresponding

DATA messages.

Slide 49


SPIN-1: A 3-stage Handshake Protocol

1. ADV stageNew Data AdCheck for DataData Request

2. REQ stageData TransmissionData FusionNew Data Ad

3. DATA stageData RequestData Transmission

SPIN-l: A 3-Stage Handshake Protocol

The SPIN-l protocol is a simple handshake protocol for disseminating

data through a lossless network. It works in three stages (ADV-REQ-

DATA), with each stage corresponding to one of the messages

described above. The protocol starts when a node obtains new data

that it is willing to disseminate. It does this by sending an ADV

message to its neighbors, naming the new data (ADV stage). Upon

receiving

an ADV, the neighboring node checks to see whether it has already

received or requested the advertised data. If not, it responds by

sending an REQ message for the missing data back to the sender

(REQ stage). The protocol completes when the initiator of the protocol

responds to the REQ with a DATA message, containing the missing

data (DATA stage).

There are several important things to note about this example. First, if

node B had its own data, it could aggregate this with the data of node

A and send advertisements of the aggregated data to all of its

neighbors (d). Second, nodes are not required to respond to every

message in the protocol. In this example, one neighbor does not send

an REQ packet back to node B (e). This would occur if that node

already possessed the data being advertised.

Though this protocol has been designed for lossless networks, it can

easily be adapted to work in lossy or mobile networks. Here, nodes

could compensate for lost ADV messages by re-advertising these

messages periodically. Nodes can compensate for lost REQ and DATA

messages by rerequesting data items that do not arrive within a fixed

time period. For mobile networks, changes in the local topology can

trigger updates to a node’s neighbor list. If a node notices that its

neighbor list has changed, it can spontaneously re-advertise all of its

data.

This protocol’s strength is its simplicity. Each node in the network

performs little decision making when it receives new data, and

therefore wastes little energy in computation. Furthermore, each node

only needs to know about its single-hop network neighbors. The fact

that no other topology information is required to run the algorithm

has some important consequences. First, SPIN-l can be run in a

completely unconfigured network with a small, startup cost to

determine nearest neighbors. Second, if the topology of the network

changes frequently, these changes only have to travel one hop before

the nodes can continue running the algorithm.

Slide 50


SPIN: Limited-energy simulations

DetermineDetermine how effectively how effectively each protocol uses its each protocol uses its available energyavailable energy

SPIN-1 distributes 68%SPIN-2 is able to distribute 73%the ideal protocol distributes 85%flooding distributes 53%gossiping distributes only 38%

For this experiment, we limited the total energy in the system to 1.6

Joules to determine how effectively each protocol uses its available

energy. Figure shows the data acquisition rate for the SPIN-l, SPIN-2,

flooding, gossiping, and ideal protocols. This figure shows that SPIN-2

puts its available energy to best use and comes close to distributing

the same amount of data as the ideal protocol. SPIN-2 is able to

distribute 73% of the total data as compared with the ideal protocol

which distributes 85%. We note that SPIN-1 distributes 68%, flooding

distributes 53%, and gossiping distributes only 38%.

Slide 51


SPINOverall assessmentOverall assessment

Focus on efficient dissemination of sensor data to data sinks and energy conservation at the sensorsEmploys two key innovations: negotiation and resource-adaptationIntroduces meta-data as descriptors for negotiationsEach sensor has a resource manager for monitoring resourcesExchanging meta-data is more efficient than exchanging dataPolling the resource manager allows for extensive energy savings of sensors

Slide 52


Directed Diffusion for WSNMotivationMotivation for algorithm designfor algorithm design

1. Robustness of communication2. Scaling for high nubmers of nodes3. Energy efficienct network operation

•• Example of operation:Example of operation:• “How many pedestrians do you observe in the

geographical region X?”

• “In what direction is that vehicle in region Y moving?”

Advances in processor, memory, and radio technology will enable

small and cheap nodes capable of sensing, communication, and

computation. Networks of such nodes can coordinate to perform

distributed sensing of environmental phenomena. In this paper, we

explore the directed-diffusion paradigm for such coordination. Directed

diffusion is data-centric in that all communication is for named data.

All nodes in a directed-diffusion- based network are application aware.

This enables diffusion to achieve energy savings by selecting

empirically good paths and by caching and processing data in-

network (e.g., data aggregation). We explore and evaluate the use of

directed diffusion for a simple remote-surveillance sensor network

analytically and experimentally. Our evaluation indicates that directed

diffusion can achieve significant energy savings and can outperform

idealized traditional schemes (e.g., omniscient multicast) under the

investigated scenarios.

Motivated by robustness, scaling and energy efficiency requirements we will

examine a new data dissemination paradigm for sensor networks. This

paradigm, directed diffusion, is data-centric. Data generated by sensor nodes

is named by attribute-value pairs. A node requests data by sending interests

for named data. Data matching the interest is then ‘drawn’ down towards that

node. Intermediate nodes can cache or transform data and may direct

interests based on previously cached data.

To motivate, consider this simplified model of how such a sensor

network will work. One or more human operators pose, to any node in

the network, questions of the form: “How many pedestrians do you

observe in the geographical region X?” or “In what direction is that

vehicle in region Y moving?” These queries result in sensors within the

specified region being tasked to start collecting information. Once

individual nodes detect pedestrians or vehicle movements, they might

collaborate with neighboring nodes to disambiguate pedestrian

location or vehicle movement direction. One of these nodes might then

report the result back to the human operator.

Slide 53


Example of operation:Example of operation:The operator’s query will be transformed into an interest that is diffused toward nodes in regions X or Y (broadcast, geographical routing)Nodes activate their sensors which begin collecting information about pedestriansInformation returns along the reverse path of interest propagationIntermediate nodes might aggregate the data

Directed Diffusion for WSN

Using this communication paradigm, our example might be

implemented as follows. The human operator’s query would be

transformed into an interest that is diffused (e.g., broadcasted,

geographically routed) toward nodes in regions X or Y. When a node in

that region receives an interest, it activates its sensors which begin

collecting information about pedestrians. When the sensors report the

presence of pedestrians, this information returns along the reverse

path of interest propagation. Intermediate nodes might aggregate the

data, e.g., more accurately pinpoint

the pedestrian’s location by combining reports from several sensors.

An important feature of directed diffusion is that interest and data

propagation and aggregation are determined by localized interactions

(message exchanges between neighbors or nodes within some vicinity).

Slide 54


Directed Diffusion for WSNDirected Diffusion elements:

Algorithm based on InterestsInterestsData messagesData messagesGradientsGradientsReinforcementsReinforcements

Sinks request data by sending interest messagesinterest messagesEach interest contains a description of a sensing a description of a sensing tasktask for acquiring dataData is a collection of eventscollection of events or processed processed informationinformation of a physical phenomenon

Directed diffusion consists of several elements: interests, data

messages, gradients, and reinforcements. An interest message is a

query or an interrogation which specifies what a user wants. Each

interest contains a description of a sensing task that is supported by a

sensor network for acquiring data. Typically, data in sensor networks

is the collected or processed information of a physical phenomenon.

Such data can be an event, which is a short description of the sensed

phenomenon.

Slide 55


Directed Diffusion elementsDirected Diffusion elements:Data is named using attributeattribute--value pairsvalue pairsThe interest dissemination sets up gradients gradients within the networkwithin the network designed to “draw” eventsA gradient direction stateA gradient direction state is created in each node that receives an interestEvents start flowingstart flowing towardtoward the originators of interests along multiple gradient pathsThe sensor network reinforces onereinforces one or a small a small numbernumber of these paths


In directed diffusion, data is named using attribute-value pairs. A

sensing task (or a subtask thereof) is disseminated throughout the

sensor network as an interest for named data. This dissemination sets

up gradients within the network designed to “draw” events (i.e., data

matching the interest). Specifically, a gradient direction state is

created in each node that receives an interest. The gradient direction

is set toward the neighboring node from which the interest is received.

Events start flowing toward the originators of interests along multiple

gradient paths. The sensor network reinforces one or a small number

of these paths.

Slide 56



Key Key featuresfeatures1.1. Interests Interests

disseminationdissemination2.2. Gradients setupGradients setup3.3. Reinforcement of Reinforcement of

one or more one or more gradient pathsgradient paths

2. Reinforcement

1. Low data rate

3. High data rate

In Directed Diffusion, a query would be transformed into an interest that is diffused or

flooded towards nodes in the interested region. When a sensor node in that region

receives the interest, it activates its sensors and begins to monitor the interested event.

The sensed data is then returned in the reverse path(s) of the interest propagation. The

intermediate nodes might aggregate the data based on their name and attribute-value

pairs. The propagation and aggregation procedures in Directed Diffusion are all based

on local information gained by localized interactions.

Directed Diffusion employs reinforcement to choose a particular path to sending

events. To do that, the interest is initially diffused with a longer interval (low-rate

events). Then low-rate events might reach the sink via multiple paths. After the sink

receives these low rate events, it reinforces one particular neighbor in order to “draw

down” real data. To do that, the sink re-sends the original interest message but with a

smaller interval. When the neighboring node receives the interest, it notices its

gradient to this neighbor and the new interval in the interest. In this case, if the new

interval is smaller than any existing gradient, it must reinforce at least one neighbor.

As this process is repeated, a path will be eventually established between the source

and the sink.

Since Directed Diffusion networks are application aware, they can achieve energy

saving by selecting good paths empirically by caching and processing data in-

network.

Slide 57


Naming for a vehicle tracking exampleNaming for a vehicle tracking example


Interest Naming

{type = wheeled vehicle;

interval = 20 ms;

duration = 10 s;

rect = [-100, 100, 200, 400] }

Data Naming{type = wheeled vehicle;interval = truck;location = [125; 220];intensity = 0:6;confidence = 0:85;timestamp = 01 : 20 : 40}

In directed diffusion, task descriptions are named by, for example, a list of

attribute-value pairs that describe a task. A vehicle-tracking task might be described

as:

type = wheeled vehicle // detect vehicle location

interval = 20 ms // send events every 20 ms

duration = 10 s // for the next 10 s

rect = [-100, 100, 200, 400] // from sensors within rectangle

For ease of exposition, we choose the subregion representation to be a rectangle

defined on some coordinate system; in practice, this might be based on GPS

coordinates. Intuitively, the task description specifies an interest for data matching

the attributes. For this reason, such a task description is called an interest. The

data sent in response to interests are also named using a similar naming scheme.

Thus, for example, a sensor that detects a wheeled vehicle might generate the

following data:

type = wheeled vehicle // type of vehicle seen

interval = truck // instance of this type

location = [125; 220] // node location

intensity = 0:6 // signal amplitude measure

confidence = 0:85 // confidence in the match

timestamp = 01 : 20 : 40 // event generation timestamp

For our sensor network, we have chosen a simple attribute-value based interest and

data naming scheme. In general, each attribute has an associated value range. For

example, the range of the attribute is the set of codebook values representing mobile

(vehicles, animal, humans). The value of an attribute can be any subset of its range.

In our example, the value of the attribute in the interest is that corresponding to

wheeled vehicles.

Slide 58


An example of path ReinforcementAn example of path Reinforcementinitial interest: { type = wheeled vehicle; interval = 1 s; rect = [-100, 200, 200, 400]; timestamp = 01 : 20 : 40; expiresAt = 01 : 30 : 40}A possible rule: Reinforce any neighbor from which a node receives a previously unseen eventthe sink resends the original interest: { type = wheeled vehicles; interval = 10 ms; rect = [-100, 200, 200, 400]; timestamp = 01 : 22 : 35; expiresAt = 01 : 30 : 40}


An interest is usually injected into the network at some (possibly arbitrary) node in the network. We use

the term sink to denote this node.

Interest Propagation: Given our choice of naming scheme, we now describe how interests are diffused

through the sensor network. Suppose that a task, with a specified type and rect, a dureation of 10 min

and an interval of 10 ms, is instantiated at a particular node in the network. The interval parameter

specifies an event data rate; thus, in our example, the specified data rate is 100 events per second. This

sink node records the task; the task state is purged from the node after the time indicated by the

duration attribute. For each active task, the sink periodically broadcasts an interest message to each of

its neighbors. This initial interest contains the specified rect and duration attributes, but contains a

much larger interval attribute. Intuitively, this initial interest may be thought of as exploratory; it tries to

determine if there indeed are any sensor nodes that detect the wheeled vehicle. To do this, the initial

exploratory interest specifies a low data rate (in our example, one event per second)

Then, the initial interest takes the following form:

type = wheeled vehicle

interval = 1 s

rect = [-100, 200, 200, 400]

timestamp = 01 : 20 : 40

expiresAt = 01 : 30 : 40

When a node receives an interest, it checks to see if the interest exists in the cache. If no matching entry

exists (where a match is determined by the definition of distinct interests specified above), the node

creates an interest entry. The parameters of the interest entry are instantiated from the received interest.

This entry has a single gradient toward the neighbor from which the interest was received, with the

specified event data rate. In our example, a neighbor of the sink will set up an interest entry with a

gradient of one event per second toward the sink.

Gradient Establishment: Previous figures shows the gradients established in the case where interests are

flooded through a sensor field. Note that for our sensor network, a gradient specifies both a data rate and

a direction in which to send events. More generally, a gradient specifies a value and a direction. In

summary, interest propagation sets up state in the network (or parts thereof) to facilitate “pulling down”

data toward the sink. The interest propagation rules are local and bear some resemblance to join

propagation in some Internet multicast routing protocols.

Reinforcement for Path Establishment: In the scheme we have described so far, the sink initially and

repeatedly diffuses an interest for a low-rate event notification. We call these exploratory events, since

they are intended for path setup and repair. We call the gradients set up for exploratory

events exploratory gradients. Once a source detects a matching target, it sends exploratory events,

possibly along multiple paths, toward the sink. After the sink starts receiving these exploratory events, it

reinforces one particular neighbor in order to “draw down” real data (i.e., events at a higher data rate that

allow high quality tracking of targets). We call the gradients set up for receiving high-quality tracking

events data gradients.

In general, reinforcememnt feature of directed diffusion is achieved by data driven local rules. One

example of such a rule is to reinforce any neighbor from which a node receives a previously unseen event.

To reinforce this neighbor, the sink resends the original interest message but with a smaller interval

(higher data rate), as follows:

type = wheeled vehicles

interval = 10 ms

rect = [-100, 200, 200, 400]

timestamp = 01 : 22 : 35

expiresAt = 01 : 30 : 40

When the neighboring node receives this interest, it notices that it already has a gradient toward this

neighbor. Furthermore, it notices that the sender’s interest specifies a higher data rate than before. If this

new data rate is also higher than that of any existing gradient (intuitively, if the “outflow” from this node

has increased), the node must also reinforce at least one neighbor. The node uses its data cache for this

purpose. Again, the same local rule choices apply. For example, this node might choose that neighbor

from whom it first received the latest event matching the interest. Alternatively, it might choose all

neighbors from which new events were recently received. This implies that we reinforce that neighbor only

if it is sending exploratory events. Obviously, we do not need to reinforce neighbors that are already

sending traffic at the higher data rate.

Slide 59


Differences w.r.t. IPDifferences w.r.t. IP--based networksbased networksdiffusion is datadiffusion is data--centriccentricall communication in diffusion is neighborneighbor--toto--neighborneighbor (not end-to-end)sensor nodes do not need to have globally globally unique identifiersunique identifiers (no IP address required)every node can cachecache, aggregateaggregate, and more generally, process messagesprocess messages (no servers for performing such tasks)


Our description points out several key features of diffusion and how it

differs from traditional networking. First, diffusion is data-centric; all

communication in a diffusion-based sensor network uses interests to

specify named data. Second, all communication in diffusion is

neighbor-to-neighbor, unlike the end-to-end communication in

traditional data networks. In other words, every node is an “end” in a

sensor network. In the

sense, there are no “routers” in a sensor network. Each sensor node

can interpret data and interest messages. This design choice is

justified by the task specificity of sensor networks. Sensor networks

are not general-purpose communication networks. Third, sensor

nodes do not need to have globally unique identifiers or globally

unique addresses. Nodes, however, do need to distinguish among

neighbors. Finally, in an IP-based

sensor network, for example, sensor data collection and processing

might be performed by a collection of specialized servers which may,

in general, be far removed from the sensed phenomena. In our sensor

network, because every node can cache, aggregate, and more

generally, process messages, it is generally desirable to perform

coordinated sensing close to the sensed phenomena. Diffusion is

clearly related to traditional network data-routing algorithms. In some

sense, it is a reactive routing technique, since “routes” are established

on demand. However, it differs from other ad hoc reactive routing

techniques in several way. First, no attemptis made to find one loop-

free path between source and sink before data transmission

commences. Instead, constrained or directional flooding is used to set

up a multiplicity of paths and data messages are initially sent

redundantly along these paths. Second, soon thereafter, reinforcement

attempts to reduce this multiplicity of paths to a small number, based

on empirically observed path performance. Finally, a message cache is

used to perform loop avoidance. The interest and gradient setup

mechanisms themselves do not guarantee loop-free paths between

source and sink

Slide 60


Directed Diffusion for WSNDirected Diffusion characteristicsDirected Diffusion characteristics

All communication is for named dataData is named by attribute-value pairsIntermediate nodes may aggregate dataThus achieving significant energy-savingsPropagation and aggregation procedures are based on local information, gained by localized interactions

DD is capable of realizing robust, multi-path, energy-efficient data delivery in WSNs

Directed Diffusion can be used to realize robust multi-path delivery. It empirically

adapts to a small subset of network paths, and therefore achieves energy saving when

intermediate nodes aggregate responses to the previous queries.

Documents

Wireless Sensor Networks (WSNs)cgi.di.uoa.gr/~istavrak/courses/WSN-2005-notes.pdf · wireless sensor network ... sensor network and therefore they must be taken into consideration