Wireless Sensor Placement for Reliable and Efficient Data Collection Edo Biagioni and Galen Sasaki University of Hawaii at Manoa.

Wireless Sensor Wireless Sensor PlacementPlacement

for Reliable and Efficient for Reliable and Efficient Data CollectionData Collection

Edo Biagioni and Galen SasakiEdo Biagioni and Galen Sasaki

University of Hawaii at ManoaUniversity of Hawaii at Manoa

OverviewOverview

• Wireless Sensor Networks

• Case study: an ecological wireless sensor network

• Design Considerations

• Regular Deployments

• Linear and other arrangements

Sensor Networks are UsefulSensor Networks are Useful

Ecological study: under what Ecological study: under what conditions does the endangered conditions does the endangered species thrive?species thrive?

Knowing the environment aids in Knowing the environment aids in setting goals or controlling processessetting goals or controlling processes

Many applications, including Many applications, including ecological, industrial, and militaryecological, industrial, and military

Ad-hoc Wireless NetworksAd-hoc Wireless Networks

Low-power operationLow-power operation Range-limited radiosRange-limited radios Ad-hoc networking: each node Ad-hoc networking: each node

forwards data for other nodesforwards data for other nodes Data may be combined Data may be combined en routeen route

Wireless Sensor Network Wireless Sensor Network DesignDesign

How densely must we sample the How densely must we sample the environment?environment?

What is the radio communications What is the radio communications range?range?

How much reliability do we have, and How much reliability do we have, and how does it improve if we add more how does it improve if we add more units?units?

How many units can we afford?How many units can we afford?

The PODS project at the The PODS project at the University of HawaiiUniversity of Hawaii

Ecological sensing of Ecological sensing of Rare Plant Rare Plant environmentenvironment

Temperature, Temperature, sunlight, rainfall, sunlight, rainfall, humidityhumidity

High-resolution High-resolution imagesimages

Kim Bridges, Brian Kim Bridges, Brian CheeChee

Pod placementPod placement

Intensive deployment Intensive deployment where the plant does where the plant does growgrow

Interested also in Interested also in where the plant does where the plant does notnot grow grow

Connection to the Connection to the internet is also a line internet is also a line of sensorsof sensors

Sub-region

Practical ConstraintsPractical Constraints

Higher radios have Higher radios have more rangemore range

CamouflageCamouflage Plant densities may Plant densities may

varyvary Different units may Different units may

have different have different sensorssensors

Ignored in this talkIgnored in this talk

Design Goals for DeploymentDesign Goals for Deployment

We are given a 2-dimensional square region We are given a 2-dimensional square region with total area Awith total area A

Minimize the maximum distance between Minimize the maximum distance between any point in A and the nearest sensorany point in A and the nearest sensor

Keep the distance between adjacent Keep the distance between adjacent sensors less than sensors less than rr

Measure point values, compute gradients Measure point values, compute gradients and significant thresholdsand significant thresholds

Design ConsiderationsDesign Considerations

Financial and other constraints often limit Financial and other constraints often limit the total number of nodes, the total number of nodes, NN

Failure of individual nodes should not Failure of individual nodes should not disable the entire networkdisable the entire network

Reducing the transmission range improves Reducing the transmission range improves the energy efficiencythe energy efficiency

Regular DeploymentsRegular Deployments

Square, triangular, or Square, triangular, or hexagonal tileshexagonal tiles

Nodes must be within Nodes must be within range range rr of their of their neighborsneighbors

Sampling distance Sampling distance δδ Degree 4, 6, or 3 Degree 4, 6, or 3

provides redundancyprovides redundancy Which is best?Which is best?

a

(a) Square tiles

(b) Triangle tile

(c) Hexagon tile

Computing with Computing with N, r, N, r, δδ

Standard formulas for tile area (Standard formulas for tile area (αα) and for ) and for distance to the center of the tiledistance to the center of the tile

Distance to center < Distance to center < δδ Distance between nodes < rDistance between nodes < r Each node is part of c = (6, 4, or 3) tilesEach node is part of c = (6, 4, or 3) tiles N = (A/N = (A/αα)/c, where A/)/c, where A/αα is the number of is the number of

tilestiles

Main Results for Regular GridsMain Results for Regular Grids

N is proportional to the surface area of AN is proportional to the surface area of A if if r < r < δδ, hexagonal deployment minimizes , hexagonal deployment minimizes

NN, and , and NN is inversely proportional to is inversely proportional to rr22

If If δδ < r < r, triangular deployment minimizes , triangular deployment minimizes NN, and , and NN is inversely proportional to is inversely proportional to δδ22

Triangular, square, or hexagonal are within Triangular, square, or hexagonal are within a factor of two of each othera factor of two of each other

Sparse GridsSparse Grids

If If r < r < δδ, we can , we can reduce the number of reduce the number of nodes by going to nodes by going to sparse grids (sparse sparse grids (sparse meshes)meshes)

Communication Communication distance remains distance remains smallsmall

the number of nodes the number of nodes may drop may drop substantiallysubstantially

3 nodes per side, s=33 nodes per side, s=3

S=3

Main Results for Sparse GridsMain Results for Sparse Grids

Communication radius Communication radius rr, tile side , tile side a = r * sa = r * s NN is inversely proportional to is inversely proportional to aa and to and to rr The degree of most nodes is two, so The degree of most nodes is two, so

reliability is reduced – the same as for reliability is reduced – the same as for linear deploymentslinear deployments

1-Dimensional Deployment1-Dimensional Deployment

• Many common applications: along Many common applications: along streams, roads, ridgesstreams, roads, ridges

• Requires relatively few nodesRequires relatively few nodes

• With the least number of nodes for a With the least number of nodes for a given given rr, network fails if a single node , network fails if a single node failsfails

• How well can we do if we double the How well can we do if we double the number of nodes?number of nodes?

Protection against node Protection against node failuresfailures• PairedPaired • InlineInline

r

r

Paired and Inline Paired and Inline PerformancePerformance

• For inline, two successive node For inline, two successive node failures disconnect the networkfailures disconnect the network

• For paired, failure of the two nodes of For paired, failure of the two nodes of a pair disconnects the networka pair disconnects the network

• The former is about twice as likelyThe former is about twice as likely

Sampling a GradientSampling a Gradient

If we know the gradient, a linear If we know the gradient, a linear deployment is sufficientdeployment is sufficient

A gradient can be computed from A gradient can be computed from three samples in a trianglethree samples in a triangle

Variable gradients need more and Variable gradients need more and longer baselines, as do threshold longer baselines, as do threshold determinationsdeterminations

Grids and sparse grids measure Grids and sparse grids measure gradients wellgradients well

Quantifying a gradientQuantifying a gradient

The differences between pairs of samples help determine the gradient

Minimizing the number of nodesMinimizing the number of nodes

The ultimate The ultimate sparse grid: a circlesparse grid: a circle

Tolerates single Tolerates single node failuresnode failures

Even sampling in Even sampling in all directionsall directions

Lines outward from Lines outward from the center: a starthe center: a star

Center is well Center is well coveredcovered

Star-3, Star-4, Star-Star-3, Star-4, Star-5, Star-m5, Star-m

SummarySummary

• Many regular deployments

• Generally, N and r are given, sampling distance is allowed to vary

• Tradeoff between N and redundancy: sparse grids allow large sampling distance

• Lines, circles, stars are optimal when N is small, can provide information about gradients

AcknowledgementsAcknowledgements

• Kim Bridges

• Brian Chee and many students on the Pods project, including Michael Lurvey and Shu Chen

• DARPA (Pods funding)

• Hawaii Volcanoes National Park