WIRELESS SENSOR NETWORKSPart of slides from Mobicom 2002 tutorials

OutlineOutline

IntroductionMotivating applicationsEnabling technologiesUnique constraintsqApplication and architecture taxonomy

Embedded Networked Sensing PotentialEmbedded Networked Sensing Potential

Micro-sensors, on-board processing, and wireless interfaces all feasible at very small scale

can monitor phenomena “up close”

Seismic Structure ContaminantWill enable spatially and temporally denseenvironmental monitoring

Seismic Structure response

Contaminant Transport

environmental monitoringEmbedded Networked Sensing will reveal previously unobservable

Marine Microorganisms

Ecosystems, Biocomplexity

previously unobservable phenomena

App: Ecosystem MonitoringApp: Ecosystem Monitoring

S iScienceUnderstand response of wild populations (plants and animals) to habitats over time.Develop in situ observation of species and ecosystem dynamics. Develop in situ observation of species and ecosystem dynamics.

TechniquesData acquisition of physical and chemical properties, at various spatial and

l l i h i d h bitemporal scales, appropriate to the ecosystem, species and habitat.

Automatic identification of organisms(current techniques involve close-range (cu e ec ques vo ve c ose a ge human observation).

Measurements over long period of time,taken in-situ.

Harsh environments with extremes in t t i t b t ti temperature, moisture, obstructions, ...

Field Experiments Field Experiments

Monitoring ecosystem processesImaging, ecophysiology, and environmental sensorsStudy vegetation response to climatic trends and diseases.y g p

Species MonitoringVisual identification, tracking, and population measurement of birds and other vertebratesmeasurement of birds and other vertebratesAcoustical sensing for identification, spatial position, population estimation.

Education outreach V h dEducation outreachBird studies by High School Science classes (New Roads and Buckley Schools).

Vegetation change detection

Avian monitoring Virtual field observations

ENS Requirements for Habitat/EcophysiologyApplications

Diverse sensor sizes (1-10 cm), spatial sampling intervals (1 cm - 100 m), and temporal sampling intervals (1 μs - days), depending on habitats and

iorganisms.

Naive approach → Too many sensors →Too many data.

I t k di t ib t d i l iIn-network, distributed signal processing.

Wireless communication due to climate, terrain, thick vegetation.

Ad ti S lf O i ti t hi li bl l li d ti i Adaptive Self-Organization to achieve reliable, long-lived, operation in dynamic, resource-limited, harsh environment.

Mobility for deploying scarce resources (e.g., high resolution sensors).Mobility for deploying scarce resources (e.g., high resolution sensors).

Transportation and Urban MonitoringTransportation and Urban Monitoring

Disaster Responsep

Intelligent Transportation Project (Muntz et al.)

Smart Kindergarten Project: Sensor based Wireless Networks of ToysSensor-based Wireless Networks of Toysfor Smart Developmental Problem-solving Environments

Middleware Framework

WLAN AccessPoint

SensorManagement

SensorFusion

SpeechRecognizer

Database& Data Miner

Wired Network

NetworkManagement

High-speed Wireless LAN (WLAN)WLAN-Piconet

BridgeWLAN-Piconet

Bridge

SensorsModules

Piconet Piconet

Sensor Badge

Networked ToysNetworked Toys

Enabling TechnologiesEnabling Technologies

Embed numerous distributed devices to monitor and interact with physical world

Network devices to coordinate and perform higher-level tasks

Embedded NetworkedExploit

physical world higher-level tasks

Control system w/Small form factorUntethered nodes

ExploitcollaborativeSensing, action

Sensing

Tightly coupled to physical worldTightly coupled to physical world

Exploit spatially and temporally dense, in situ, sensing and actuation

SensorsSensors

Passive elements: seismic, acoustic, infrared, strain, salinity, humidity, temperature, etc.

Passive Arrays: imagers (visible, IR), biochemical

Active sensors: radar sonarActive sensors: radar, sonar

High energy, in contrast to passive elements

T h l t d f IC t h l f i d b t Technology trend: use of IC technology for increased robustness, lower cost, smaller size

COTS d t i f th d i k i t b d COTS adequate in many of these domains; work remains to be done in biochemical

Some Networked Sensor Node DevelopmentsSome Networked Sensor Node Developments

LWIM III

UCLA, 1996

G h RFM

AWAIRS I

UCLA/RSC 1998

G h DS/SSGeophone, RFM

radio, PIC, star

network

Geophone, DS/SS

Radio, strongARM,

Multi-hop networksnetwork Multi hop networks

WINS NG 2 0WINS NG 2.0Sensoria, 2001Node development

UCB Mote, 20004 Mhz, 4K Ram512K EEProm platform; multi-

sensor, dual radio,Linux on SH4,

512K EEProm,128K code, CSMA

Processor

,Preprocessor, GPShalf-duplex RFM radio

New Design ThemesNew Design ThemesLong-lived systems that can be untethered and unattended

Low-duty cycle operation with bounded latencyExploit redundancy and heterogeneous tiered systems

Leverage data processing inside the networkThousands or millions of operations per second can be done using energy of sending a bit over 10 or 100 meters (Pottie00)Exploit computation near data to reduce communication

Self configuring systems that can be deployed ad hocUn-modeled physical world dynamics makes systems appear ad hocMeasure and adapt to unpredictable environmentExploit spatial diversity and density of sensor/actuator nodes

Achieve desired global behavior with adaptive localized algorithmsCant afford to extract dynamic state information needed for centralized control

Sample Layered Architecture p y

User Queries, External Database

Resource constraints call for more tightly In-network: Application processingintegrated layers

O Q i

In network: Application processing, Data aggregation, Query processing

D t di i ti t hiOpen Question:

Can we define anI t t lik

Data dissemination, storage, caching

Internet-like architecture for such application-specific systems??

Adaptive topology, Geo-Routing

systems??MAC, Time, Location

Phy: comm, sensing, actuation, SP

Systems Load/Event Metrics

• Spatial and Temporal Scale– Extent

Taxonomy• Frequency

– spatial and temporal

ModelsMetrics

• Efficiency– System

– Spatial Density (of sensors relative to stimulus)D t t f ti lii

spatial and temporal density of events

• Locality – spatial, temporal

System lifetime/System resources

• Resolution/Fidelity– Data rate of stimulii

• Variability– Ad hoc vs. engineered

system structure

spatial, temporal correlation

• Mobility– Rate and pattern

/– Detection,

Identification• Latencysystem structure

– System task variability– Mobility (variability in

space)

p– Response time

• Robustness– Vulnerability to

• Autonomy– Multiple sensor

modalitiesComputational model

node failure and environmental dynamics

S l bili– Computational model complexity

• Resource constraints– Energy, BW

• Scalability– Over space and

timegy,

– Storage, Computation

OutlineOutline

Survey of sensor node platformsEnergy issues for wireless sensor networksgy

Sources of energy consumptionEnergy management techniquesEnergy management techniques

Sensor Node H/W-S/W PlatformsSensor Node H/W S/W PlatformsIn-node

Wireless communication

In node processing

Event detection with neighboring nodes

sensors CPU radioAcoustic, seismic, image, magnetic etc interface

Electro-magnetic interfacemagnetic, etc. interface

batteryLimited battery supply

Energy efficiency is the crucial h/w and s/w design criterion

Variety of Real-life Sensor Node PlatformsVariety of Real life Sensor Node Platforms

RSC WINS & Hidra

Sensoria WINS

UCLA’s iBadge

UCLA’s Medusa MK-II

Berkeley’s Motes

Berkeley Piconodes

MIT’ AMPMIT’s μAMPs

And many more…

Different points in (cost, power, functionality, form factor) space

Berkeley MotesBerkeley Motes

Devices that incorporate communications, processing, sensors, and batteries into a small package

Atmel microcontroller with sensors and a communication unit

RF i l d l RF transceiver, laser module, or a corner cube reflector

temperature, light, humidity, p , g , y,pressure, 3 axis magnetometers, 3 axis accelerometers

Ti OSTinyOS

light, temperature,

II-19

g , p ,10 kbps @ 20m

The Mote FamilyThe Mote Family

Ref: from Levis & Culler, ASPLOS 2002

UCLA iBadgeUCLA iBadge

Wearable Sensor Badgeacoustic in/out + DSPtemperature, pressure, humidity, magnetometer, accelerometeracce e o e eultrasound localizationorientation via magnetometer and accelerometerbluetooth radiobluetooth radio

Sylph Middleware

BWRC’s PicoNode TripWire Sensor NodeBWRC s PicoNode TripWire Sensor Node

Chip encapsulation

Solar Cell(0.5 mm) Battery (3.6

mm)

PCB (1 mm)

(1.5 mm)

Components and batteryVersion 1: Light Powered Components and batterymounted on back

Size determined by powerdissipation (1 mW avg)

Version 2: Vibration Powered dissipation (1 mW avg)

Ref: from Jan Rabaey, PAC/C Slides

Sensor Node Energy RoadmapSensor Node Energy Roadmap

10,00010,000

• Deployed (5W) Rehosting to Low

1,0001,000

r (m

W)

ep oyed (5 )

• PAC/C Baseline (.5W)

Power COTS(10x)

100100

1010ge P

ower

• (50 mW) -System-On-ChipAdv Power 1010

11Ave

rag -Adv Power

ManagementAlgorithms (50x)

.1.1

(1mW)

20002000 20022002 20042004

Source: ISI & DARPA PAC/C Program

Comparison of Energy SourcesComparison of Energy Sources

Comparison of Potential Energy Scavenging Techniques

Power (Energy) Density Source of Estimates

B tt i (Zi Ai ) 1050 1560 Wh/ 3 (1 4 V) P bli h d d t f f tBatteries (Zinc-Air) 1050 -1560 mWh/cm3 (1.4 V) Published data from manufacturersBatteries(Lithium ion) 300 mWh/cm3 (3 - 4 V) Published data from manufacturers

Solar (Outdoors)15 mW/cm2 - direct sun

0.15mW/cm2 - cloudy day. Published data and testing.

Solar (Indoor).006 mW/cm2 - my desk

0.57 mW/cm2 - 12 in. under a 60W bulb TestingVibrations 0.001 - 0.1 mW/cm3 Simulations and Testing

3E-6 mW/cm2 at 75 Db sound levelAcoustic Noise

3E-6 mW/cm at 75 Db sound level 9.6E-4 mW/cm2 at 100 Db sound level Direct Calculations from Acoustic Theory

Passive Human Powered 1.8 mW (Shoe inserts >> 1 cm2) Published Study.

0 0018 W 10 d C di t P bli h d St dThermal Conversion 0.0018 mW - 10 deg. C gradient Published Study.

Nuclear Reaction80 mW/cm3

1E6 mWh/cm3 Published Data.

With aggressive energy management, ENS might live off the environment.

Source: UC Berkeley

Communication/Computation T h l P j iTechnology Projection

1999 (Bluetooth 2004

Technology)(150nJ/bit) (5nJ/bit)1 5 W* 50 W

Communication1.5mW* 50uW

~ 190 MOPS(5 J/OP)

Computation

Communication

Assume: 10kbit/sec. Radio, 10 m range.

(5pJ/OP)p

Large cost of communications relative to computation continues

Source: ISI & DARPA PAC/C Program

Where Does the Energy Go?Where Does the Energy Go?

Processing

excluding low-level processing for radio, sensors, actuators

Radio

Sensors

A t t

II-26

Actuators

Power supply

ProcessingProcessing

Common sensor node processorsCommon sensor node processors:Atmel AVR, Intel 8051, StrongARM, XScale, ARM Thumb, SH Risc

Power consumption all over the map, e.g.16 5 mW for ATMega128L @ 4MHz16.5 mW for ATMega128L @ 4MHz75 mW for ARM Thumb @ 40 MHz

But, don’t confuse low-power and energy-efficiency!ExampleExample

242 MIPS/W for ATMega128L @ 4MHz (4nJ/Instruction)480 MIPS/W for ARM Thumb @ 40 MHz (2.1 nJ/Instruction)

Other examples:0.2 nJ/Instruction for Cygnal C8051F300 @ 32KHz, 3.3V0.35 nJ/Instruction for IBM 405LP @ 152 MHz, 1.0V0.5 nJ/Instruction for Cygnal C8051F300 @ 25MHz, 3.3V0.8 nJ/Instruction for TMS320VC5510 @ 200 MHz, 1.5V0.8 nJ/Instruction for TMS320VC5510 @ 200 MHz, 1.5V1.1 nJ/Instruction for Xscale PXA250 @ 400 MHz, 1.3V1.3 nJ/Instruction for IBM 405LP @ 380 MHz, 1.8V1.9 nJ/Instruction for Xscale PXA250 @ 130 MHz, .85V (leakage!)

A f ffAnd, the above don’t even factor in operand size differences!However, need power management to actually exploit energy efficiency

Idle and sleep modes, variable voltage and frequency

RadioRadio

Energy per bit in radios is a strong function of desired communication performance and choice of modulation

R d BER f i h l diti ( i lti th d D l f di )Range and BER for given channel condition (noise, multipath and Doppler fading)

Watch out: different people count energy differentlyE.g.

Mote’s RFM radio is only a transceiver, and a lot of low-level processing takes place in the main CPU

While, typical 802.11b radios do everything up to MAC and link level encryption in the “radio”

Transmit, receive, idle, and sleep modes

Variable modulation, coding

C tl d 150 J/bit f h t Currently around 150 nJ/bit for short ranges

More later…

Computation & CommunicationComputation & Communication

Encode Decode TransmitDecode

Energy breakdown for voice Energy breakdown for MPEG

Encode Decode TransmitEncode

Decode

TransmitReceive Receive

Processor: StrongARM SA-1100 at 150 MIPSRadio: Lucent WaveLAN at 2 Mbps

Radios benefit less from technology improvements than processorsTh l ti i t f th i ti b t th t The relative impact of the communication subsystem on the system energy consumption will grow

SensingSensing

Several energy consumption sourcestransducerfront-end processing and signal conditioning

analog, digitalADC conversionADC conversion

Diversity of sensors: no general conclusions can be drawnLow-power modalitiesp

Temperature, light, accelerometerMedium-power modalities

Acoustic, magneticHigh-power modalities

Image, video, beamformingImage, video, beamforming

ActuationActuation

Emerging sensor platformsMounted on mobile robotsAntennas or sensors that can be actuatedAntennas or sensors that can be actuated

Energy trade-offs not yet studiedSome thoughts:

Actuation often done with fuel, which has much higher energy density than batteries

E.g. anecdotal evidence that in some UAVs the flight time is longer than E.g. anecdotal evidence that in some UAVs the flight time is longer than the up time of the wireless camera mounted on it

Actuation done during boot-up or once in a while may have significant payoffssignificant payoffs

E.g. mechanically repositioning the antenna once may be better than paying higher communication energy cost for all subsequent packetsE g moving a few nodes may result in a more uniform distribution of E.g. moving a few nodes may result in a more uniform distribution of node, and thus longer system lifetime

Power Analysis of RSC’s WINS NodesPower Analysis of RSC s WINS Nodes

• Summary• Processor

– Active = 360 mWActive 360 mW• doing repeated

transmit/receive

– Sleep = 41 mWSleep 41 mW– Off = 0.9 mW

• Sensor = 23 mW• Processor : Tx = 1 : 2• Processor : Rx = 1 : 1• Total Tx : Rx = 4 : 3 at maximum

range– comparable at lower Tx

Power Analysis of Mote-Like NodePower Analysis of Mote Like Node

Some ObservationsSome Observations

Using low-power components and trading-off unnecessary performance for power savings can have orders of magnitude impactNode power consumption is strongly dependent on the operating mode

E.g. WINS consumes only 1/6-th the power when MCU is asleep as opposed to active

At short ranges, the Rx power consumption > T power consumptionmultihop relaying not necessarily desirablep y g y

Idle radio consumes almost as much power as radio in Rx modeRadio needs to be completely shut off to save power as in sensor networks idle time dominates

MAC protocols that do not “listen” a lotMAC protocols that do not listen a lot

Processor power fairly significant (30-50%) share of overall powerIn WINS node, radio consumes 33 mW in “sleep” vs. “removed”

Argues for module level power shutdown

Sensor transducer power negligibleUse sensors to provide wakeup signal for processor and radioNot true for active sensors though…

Energy Management ProblemEnergy Management Problem

Actuation energy is the highestActuation energy is the highestStrategy: ultra-low-power “sentinel” nodes

Wake-up or command movement of mobile nodes

Communication energy is the next important issueStrategy: energy-aware data communication

Ad h i f h i i d i Adapt the instantaneous performance to meet the timing and error rate constraints, while minimizing energy/bit

Processor and sensor energy usually less important

Transmit 720 nJ/bit Processor 4 nJ/op

R i 110 J/bit 200 /bit

MICA moteBerkeley

Receive 110 nJ/bit ~ 200 ops/bit

WINS node Transmit 6600 nJ/bit Processor 1.6 nJ/opW NS odeRSC

p

Receive 3300 nJ/bit ~ 6000 ops/bit

Processor Energy ManagementProcessor Energy Management

K bKnobsShutdownDynamic scaling of frequency and supply voltagegMore recent: dynamic scaling of frequency, supply voltage, and threshold voltage

All of the above knobs incorporated into d OS h d l

Application

PA-APIsensor node OS schedulers

e.g. PA-eCos by UCLA & UCI has Rate-monotonic Scheduler with shutdown and DVS

Gains of 2x-4x typically in CPU power with

PA-Middleware

POSIX PA-OSLGains of 2x 4x typically, in CPU power with typical workloadsPredictive approaches

Predict computtion load and set

OperatingSystem

OperatingSystem

Modified OS Services

PA HALpvoltage/frequency accordinglyExploit the resiliency of sensor nets to packet and event lossesNow losses due to computation noise

Hardware Abstraction Layer

PA-HAL

HardwareNow, losses due to computation noise

Radio Energy ManagementRadio Energy Management

Tx Rx

?

?

time

During operation, the required performance is often less than the peak performance the radio is designed for

How do we take advantage of this observation, in both the sender and the receiver?

Energy in Radio: the Deeper Story….Energy in Radio: the Deeper Story….

Tx: Sender Rx: Receiver

ChannelIncomingi f i

Outgoinginformation information

TxE RxEEelecE elecERFETransmit

electronicsReceive

electronicsPower

amplifier

Wireless communication subsystem consists of three components with b t ti ll diff t h t i tiwith substantially different characteristicsTheir relative importance depends on the transmission range of the radiothe radio

ExamplesExamples

N k C021 M d S N d

nJ/bit nJ/bit nJ/bit

GSM Nokia C021 Wireless LAN

Medusa Sensor Node (UCLA)

4000

6000

8000

200

300

400

600

0

2000

4000

0

100

0

200

TxelecE Rx

elecERFE TxelecE Rx

elecERFE TxelecE Rx

elecERFE~ 1 km ~ 50 m ~ 10 m

• The RF energy increases with transmission range• The electronics energy for transmit and receive are typically comparablegy yp y p

Energy Consumption of the SenderEnergy Consumption of the Sender

Parameter of interest:

energy consumption per bitTx: Sender

I i PIncominginformation

bitbit T

PE =

TxelecP

RFP TotalP

ergy rgy

ergy

RFDominates

Electronics Dominates

Ene

Ener Ene

Transmission timeTransmission time Transmission time

Effect of Transmission RangeEffect of Transmission Range

Short rangeL Short-rangeLong-range

Ener

gy

Medium-range

E

Transmission timeTransmission time

Power Breakdowns and TrendsPower Breakdowns and Trends

Radiated powerRadiated power63 mW (18 dBm)

Intersil PRISM II

Power amplifier 600

e s SM (Nokia C021 wireless LAN)

Analog electronics240 mW

Digital electronics170 mW

Power amplifier 600 mW

(~11% efficiency)Trends:

Move functionality from the analog to the digital electronicsDigital electronics benefit most from technology improvements

Borderline between ‘long’ and ‘short’-range moves towards shorter transmit distances

Radio Energy Management #1: Shutdown

PrincipleOperate at a fixed speed and power level

Poweravailable time

Shut down the radio after the transmissionNo superfluous energy consumption timei i iNo supe uous e e gy co su p o

GotchaWhen and how to wake up?M l t

timetransmission time

More later …

Energy

transmission time

shutdown

no shutdown

allowed time

Radio Energy Management #2: Scaling along the Performance-Energy Curve

II-44 Poweravailable time

Principle

– Vary radio ‘control knobs’ such as modulation and

timetransmission time

such as modulation and error coding

– Trade off energy versus t i i ti

Modulation scalingfewer bits per symbol

transmission time

RFEfewer bits per symbol

Code scalingmore heavily coded

TxelecE Rx

elecEmore heavily coded

Ener

gy

Ener

gy

transmission time transmission time

When to Scale?When to Scale?

RF dominates Electronics dominates

ergy

Ene

E

transmission time

Emin

t*

• Scaling results in a convex curve with an energy minimum EminI l k l d i i i t* di • It only makes sense to slow down to transmission time t* corresponding to this energy minimum

Scaling vs. ShutdownScaling vs. Shutdown

yRegion of Region of

• Use scaling while it reduces the energy

• If ti i En

ergy scaling shutdown • If more time is

allowed, scale down to the minimum

Emin

energy point and subsequently use shutdown

t* time

Power

allowed time

Power

allowed time

Powerallowed timetransmission time

timetransmission time = t*

timetransmission time = t*

time

transmission time

Long-range SystemLong range System

II-47

The shape of the curve depends on the relative i f RF d

li bl i

importance of RF and electronics

This is a function of the

Ener

gy

realizable region transmission range

Region of scaling

Long-range systems have an operational region where they benefit from scaling

transmission time t*

y g

transmission time

Short-range SystemsShort range Systems

II-48 • Short-range systems have an operational have an operational region where scaling isnot beneficial

nerg

y realizable region

Region of

• Best strategy is to transmit as fast as possible and shut down

E Region of shutdown

i i it* transmission timet*

Sensor Node Radio Power Management Summary

Sh t li k Short-range links

• Shutdown based

rgy

• Turn off sender and receiver• Topology management schemes exploit this

e.g. Schurgers et. al. @ ACM MobiHoc ‘02

Ener

g S g @ CM M 0transmission time

Long-range links

Scaling based

y

Slow down transmissions

Energy-aware packet schedulers exploit this R h h l @ ACM ISLPED ‘02

Ener

gy

e.g. Raghunathan et. al. @ ACM ISLPED ‘02transmission time

Another Issue: Start-up TimeAnother Issue: Start up Time

Ref: Shih et. al., Mobicom 2001

Wasted EnergyWasted Energy

Fixed cost of communication: startup timeHigh energy per bit for small packets

Ref: Shih et. al., Mobicom 2001

Sensor Node with Energy-efficient Packet Relaying [Tsiatsis01]

• Problem: sensor noes often simply relays packets– e.g. > 2/3-rd pkts. in some sample tracking simulations

• Traditional : main CPU woken up, packets sent across bus– power and latency penalty

• One fix: radio with a packet processor handles the common case of relaying– packets redirected as low in the protocol stack as possible

Ch ll h d h l ll d• Challenge: how to do it so that every new routing protocol will not require a new radio firmware or chip redesign?

– packet processor classifies and modifies packets according to application-defined rules l d h bi i f k i h d d i f i

…zZZ– can also do ops such as combining of packets with redundant information

MultihopPacket

MultihopCommunication

Subsystem

Radio

GPS

Micro

Rest of the Node

CPU Sensor

Packet CommunicationSubsystem

RadioM d

GPS

Micro

Rest of the Node

CPU Sensor

Packet

Modem ControllerCPU Sensor Modem Controller

CPU Sensor

Traditional Approach Energy-efficient Approach

Putting it All Together: Power-aware Sensor Node

Sensors RadioCPU

Dynamic Voltage & Freq.

Scaling

Scalable Sensor

Processing

Freq., Power, Modulation, & Code ScalingScaling Processing Code Scaling

Coordinated Power Management

PA-APIs for Communication, Computation, & Sensing

Energy-aware RTOS, Protocols, & Middleware

PASTA Sensor Node Hardware Stack

OutlineOutline

Exploiting Node Mobility in Wireless Sensor NetworksL li i i h N d M biliLocalization with Node MobilityMobile Assist Event Collection

54

Mobile Sensor NodeMobile Sensor Node

55

Mobile Sensor NetworksMobile Sensor Networks

Traditional Static Sensor NetworksSensing, monitoring, and data collection applicationsFixed deployment

Mobile Sensor NetworksAnalyze the effect of node mobility on system capability, design and performance

Research DirectionsUncontrollable Mobility (U-Mobility)

E bl li ti ith t ll bl d t i t d Enable sensor applications with uncontrollable and uncertainty node movement

Controllable Mobility (C-Mobility)y ( y)Leverage controllable mobile node to improve the performance

56

Challenges & IssuesChallenges & Issues

In static sensor network localization is considered as a basic In static sensor network, localization is considered as a basic capability of sensor node.

It is essential to interpret the collected data

Upper layer protocols (e.g., Geographic Routing) are running with node’s location

Th f d t l i t f bil t kThe fundamental impact of mobile sensor networksThe location of node keeps changing

Reconsidering localization with U-Mobility is a critical challenge

Quick re-localizable

Distributed

Accurate and GPS-free (low cost)57

Challenges & Issues (cont )Challenges & Issues (cont.)

L li tiLocalizationwith node mobility

Coverage Application

1) Probabilistic coverage2) Maintain the coverage with C-Mobility against with C-Mobility against U-Mobility

• Probabilistic Coverage Map▫ A place is not just “yes” or “no”, but “probably be” covered.

• Maintain the coverage

58

g▫ Break by U-Mobility; Recover by C-Mobility.

Challenges & Issues (cont )

L li ti

Challenges & Issues (cont.)

Localizationwith node mobility

Coverage ApplicationEvent CollectionApplication1) Probabilistic coverage2) Maintain the coverage with C-Mobility against

1) Mobile assist in energy balancing

2) Design feasible path

pp

with C-Mobility against U-Mobility2) Design feasible path

of C-Mobility

• Mobile Assist in Energy Balancing▫ Balance the query cost among sensing nodes to extend the network lifetime.

• Feasible Path Design of C-Mobility▫ Feasible path: velocity requirement, total travelling distance

Challenges & Issues (cont )

L li ti

Challenges & Issues (cont.)

Localizationwith node mobility

1) Probabilistic coverage

Coverage ApplicationEvent CollectionApplicationOther Applications

) g2) Maintain the coverage with C-Mobility against U-Mobility

1) Mobile assist in energy balancing

2) Design feasible path

pp

1) Data Aggregation2) Mobile Routing3) V hi l t k2) Design feasible path

of C-Mobility3) Vehicular networks4) …

• We focus on exploiting node mobility of• We focus on exploiting node mobility of▫ Localization▫ Coverageg▫ Event Collection

OutlineOutline

I t d tiIntroductionLocalization with Node Mobility

Ji Luo and Qian Zhang, “Relative Distance Based Localization for Mobile Sensor Networks”, GlobeCom 2007, Best Paper Award.

Mobile Assist Event CollectionMobile Assist Event CollectionFuture WorkC l iConclusion

Challenges & IssuesChallenges & Issues

In static sensor network, localization is considered as a basic capability of sensor node.

It is essential to interpret the collected dataIt is essential to interpret the collected dataUpper layer protocols (e.g., Geographic Routing) are running with node’s location

Th f d l i f bil kThe fundamental impact of mobile sensor networksThe location of node keeps changing

Reconsidering localization with U-Mobility is a critical challenge

Q k l l blQuick re-localizableDistributedAccurate and GPS-free (low cost)( )

62

L li tiLocalizationExisting Localization Algorithm Existing Localization Algorithm

Static Sensor NetworksDV-HOP (Niculescu & Nath, 2003)( , )APIT (Tian He, et. Al, 2003)... ...Not suitable for mobile sensor networks

Mobile Sensor NetworksMCL (Hu & Evan, 2004)MCL (Hu & Evan, 2004)Simple Monte Carlo approachNot accurate enough

We face sensor nodes moving in a region uncontrollablyOnly some of them known their locations (e.g., seeds), others are GPS-freeH l bili i l li iHow to leverage mobility in localization

L li iR l i Di C i

LocalizationRelative Distance Constraints

Constraint from static location information1 h t i t1-hop constraint

Seed B in the list of node A’s 1-hop neighbors

Sensor A

Inequality:

2 hop constraint

222 )()( Ryyxx BABA ≤−+− R

Sensor B

2-hop constraintSeed B in the list of node A’s 2-hop neighbors but not in the list of A’s Sensor A1-hop neighborsInequality: 2222 4)()( RyyxxR BABA ≤−+−≤

R Sensor B

Localization

R l i Di C i

Localization

Relative Distance ConstraintsConstraint from static location informationC i l i bili i f i 2VConstraint leveraging mobility information

Motion on boundary constraintSeed B in the list of A’s 1-hop neighbors Sensor A

2Vmax

Seed B in the list of As 1 hop neighborsbut not in the list of A’s 1-hop neighborsat previous time slotI li

Sensor A

R

Inequality:

Inherited Constraint

2222max )()()2( RyyxxVR BABA ≤−+−≤−

Sensor B

Inherited ConstraintIf we have inequality at previous time slot:

dcybxa AA ≤−+−≤ 22 )()(

We can obtain new inequality for current time slot:

( )2max222

max )()(}),0{(max VdcybxVaof BA +≤−+−≤−

LocalizationLocalization

Relative Distance ConstraintsConstraint from static location information

C i l i bili i f iConstraint leveraging mobility information

Our solution: MIL (Mobile Inequality Localization)Basic IdeaBasic Idea

Leverage relative distance constraints (knowledge from neighbor and history motion) to restrict current node’s position in a small region

Technical problemSolve set of ring inequalities

( ) ( )⎧ 22( ) ( )( ) ( )

⎪

⎪⎨

⎧

≤−+−≤≤−+−≤

22

22

22

12

12

11

dcybxadcybxa

⎪⎩ KK

Localization

MIL i d i h d

Localization

MIL, estimated weighted centerNumerical method

maxYWeighted Center

(0,0)

T i li i

Outer BoundaryminX maxX

minY

222 5)0()0( ≤−+− AA yx2222 5)0()5(2 ≤+≤ yx

Two inequality constraints

55,55 ≤≤−≤≤− AA yx=>

55100 ≤≤−≤≤ yx=> 5550

≤≤−≤≤

A

A

yx

5)0()5(2 ≤−+−≤ AA yx 55,100 ≤≤≤≤ AA yx=> Ay

LocalizationLocalization

EvaluationEstimated error of MIL in theory

( )( ) ( ) 1,1111)(222

≤⎟⎠⎞

⎜⎝⎛ −−=−−≥≤ εεεεδ π

SdmP

Simulation

Accuracy Impact of velocityy p y

OutlineOutline

I t d tiIntroductionLocalization with Node MobilityMobile Assist Event Collection

Ji Luo, Xing Xu, and Qian Zhang, “Delay Tolerant Event Collection for Undergro nd Coal Mine sing Mobile Sinks” IWQoS 2009 (best paper Underground Coal Mine using Mobile Sinks”, IWQoS 2009. (best paper award candidate)

Xing Xu, Ji Luo, and Qian Zhang, “Delay Tolerant Event Collection in Sensor Networks with Mobile Sink”, InfoCom 2010.

Future WorkConclusion

Mobile Assist Event CollectionMobile Assist Event Collection

Basic ScenarioBasic ScenarioStatic sensors deployed in underground tunnel to monitor events

Tramcars move in a fixed path with fixed velocity and fixed time intervalTramcars move in a fixed path with fixed velocity and fixed time interval

Path of tramcarLocation

Path of tramcar

event

Sensors

event

TimeTime interval

Mobile Assist E ent CollectionMobile Assist Event Collection

Ob iObservationTramcars could be exploited as mobile sink to relay delay tolerant events

Tough communication environment (static sensor could not communicate with each other)Tough communication environment (static sensor could not communicate with each other)

Selectively query portion of static sensors but not allSpatial-Temporal Correlation: one event may last for a period of time and might be sensed by multiple static sensorsmultiple static sensors

More about CorrelationMore about Correlation

Collect sensor readings from all sensor nodes

Redundancy due to Correlation of Sensor readings:

Spatial correlation: nearby sensors have similar sensor readings

Temporal correlation: sensor readings would not change significantly during a short period of timeshort period of time

Leveraging Correlation: Reading EventLeveraging Correlation: Reading Event

Spatial-Temporal Correlation of sensor readingsSpatial-Temporal Correlation of event:Spatial Temporal Correlation of event:

Temp. in LabT > 36 °C

Mobile Assist E ent CollectionMobile Assist Event Collection

Wi h hi b iWith this observationSelectively query portion of static sensors but not all

Spatial Temporal Correlation one event may last for a period of Spatial-Temporal Correlation: one event may last for a period of time and might be sensed by multiple static sensors

Event Collection Problem: set of events to be collected

Design a set of event collection actionsΩ

ΨTarget

Collect all events:

Maximize network lifetime

Mobile Assist Event CollectionMobile Assist Event Collection

Samples of Learning Sample SetTime 1

Collected events −Ω 0t

Sensor Readings

Predict Recent Sensor

ProcessProcess

Sample Set

P

Time 2 Generate the inputby prediction

+Ω 0t

Online

Use existing samples to predict mean and variance of recent sensor readings1) Local optimization is NP-hard

Readings Spatial-Temporal

Correlation Model

Process

Sample SetTime 3

e Process

1) Local optimization is NP hard2) Heuristic Solution

add event collection action withsmallest

of Event RegionGenerate Possible Event Regions

Sample SetTime k

Local Optimization

singto maximize the minimal residual energy

Make the Decision of Event Collection Actions

Knowledge Obtained by

Process

Sample SetUpdate Samples

Knowledge Obtained by Current Event Collection Actions

Mobile Assist E ent CollectionMobile Assist Event CollectionEvaluationEvaluation

Three metrics:• Network Lifetime

Double lifetime• Energy balance• Event collection rate

Low energy deviation

Stable event collection rate after initial stagerate after initial stage

Mobile Assist Event CollectionMobile Assist Event Collection

P i iPrevious scenarioU-Mobility: tramcar is uncontrollable but the path is known in advance

Another scenarioC M bili h h d if ld l h bil i k C-Mobility: what happened if we could control the mobile sink to move and collect the event

Selective querying still work for spatial-temporal correlationq y g p pFeasible path design for C-Mobility is required

Velocity requirementTotal travelling distance

Event Collection in 2D RegionEvent Collection in 2D Region

Static sensors are deployed to sense the environmentSpecial sensor readings indicate some event

C ll t d b iCollect event data to a base station

Lifetime is main concern

Potential Collection SchemePotential Collection Scheme

Forwarding packets Great energy consumption

Asymmetry location Imbalanced consumption

Full connectivity Costly & unrealistic

Introduce a Mobile SinkIntroduce a Mobile Sink

No forwarding tasksForwarding packets Great energy consumption

g

No imbalanced consumption due to forwarding

No need to maintain full connectivity

Asymmetry location Imbalanced consumption

Full connectivity Costly & unrealistic

Potential Reporting SchemePotential Reporting Scheme

Collect sensor readings from all sensor nodesRedundancy due to Correlation of Sensor readings:

Spatial correlation: nearby sensors have similar sensor readingsTemporal correlation: sensor readings would not change significantly Temporal correlation: sensor readings would not change significantly during a short period of time

Again Leveraging CorrelationAgain: Leveraging Correlation

Spatial-Temporal Correlation of sensor readingsSpatial-Temporal Correlation of event:Spatial Temporal Correlation of event:

Temp. in LabT > 36 °C

Leveraging Correlation (2)Leveraging Correlation (2)

Event Region:Boundary of connected events, indicating the “same” event

Event region contains an Upright CylinderS ti l T l C l ti f tSpatial-Temporal Correlation of event

Collect Event Region, instead of all events

Energy Efficient Event Collection SchemeEnergy Efficient Event Collection Scheme

1. Full sensor readingsAll pointsAll points

2. Data indicating events Highlighted red pointsHighlighted red points

3. Representative data of event regionevent region

One S-T point for each cylinder

Less collections Energy saving

Problem FormulationProblem Formulation

Event Collection Problem: set of event regions to collect

: find a set of query actions (collections)Objectives:j

(1) Provide certain collection rate of ;(2) Maximize the network lifetime .

Maximize the minimum residual energy of static sensorsMaximize the minimum residual energy of static sensors

S l ti ith P f t K l dSolution with Perfect Knowledge

AssumptionSet of event regions is given in advance

Sensor Selection ProblemSelect a set of sensors to query

Selected sensor will consume energyAfter the selection, the lifetime can be maximized

Network Flow Graph for Lifetime MMaximization

Framework with Perfect KnowledgeFramework with Perfect Knowledge

Perfect Knowledge• Given Event Regions

Unrealistic Assumption• Given Event Regions p

Sensor Selection Problem• Selected Sensor SetSelected Sensor Set

Design Route• Perform CollectionsPerform Collections

P ti l S hPractical Scheme

Predicted Event Regions• Based on collected readings

Predict

• Based on collected readings

Sensor Selection Problem• Selected Sensor Set

UpdateSelected Sensor Set

Process

Design Route• Perform CollectionsPerform Collections

E l tiEvaluation

Simulation Setting100 static sensorsSensor readings

Intel Berkeley Research Lab Trace Data

Competitors

Our Scheme Stochastic CoverageOur Scheme Stochastic Coverage

S-T Correlation √ √ √

Prediction √ √ ╳╳

LifetimeMaximization

√ Randomly N.A.

Lifetime Performance (1)Lifetime Performance (1)

Similar acceptable event collection rate: 85%2.5X lifetime comparing to Stochastic Schemep g5X lifetime comparing to Coverage Scheme

Lifetime Performance (2)Lifetime Performance (2)

Standard-deviation of residual energy of all static sensorsConverged standard-deviation, Others keep increasing

Balanced energy consumption

E ent Collection Concl sionEvent Collection - Conclusion

Energy-Efficient Event Collection SchemeLeverage

Mobility of mobile sinkSpatial-temporal correlation of event

Concept of event region for more efficient collection scheme

Practical event collection schemeB ild di i d lBuild prediction modelProved guaranteed event collection rate in theoryVerified good performance on real trace dataVerified good performance on real trace data

ConclusionConclusion

Categorize node mobilityU-Mobility and C-Mobility

E bl l li i i M bil S N kEnable localization in Mobile Sensor NetworksAddress several issues

CoverageCoverageProbabilistic coverage mapDouble mobility to maintain coverage

Event CollectionSensor selection with Spatial-Temporal CorrelationFeasible Movement Path DesignFeasible Movement Path Design

Identify several future research topics based on the current progressp g