Sensor Networks for Sensor Networks for Engineering ApplicationsEngineering Applications

Sensor Networks for Sensor Networks for Engineering ApplicationsEngineering Applications

Dr. Saman K. HalgamugeDr. Saman K. Halgamuge

Associate Professor and Reader (Mechatronics) Associate Professor and Reader (Mechatronics) University of Melbourne, AustraliaUniversity of Melbourne, Australia

Current Research ProjectsCurrent Research Projects Driver Support Systems using Image Processing (funded by Robert Bosch Driver Support Systems using Image Processing (funded by Robert Bosch

Germany)Germany) Bioinfomatics: Protein data analysis (funded by Australian Research Council)Bioinfomatics: Protein data analysis (funded by Australian Research Council) Soft Computing applied to Brake by wire Systems (funded by Australian Soft Computing applied to Brake by wire Systems (funded by Australian

Research Council and Pacifica Group Technogies Ltd)Research Council and Pacifica Group Technogies Ltd) Sensor Networks (funded by Australian Research Council and Ajax Fastners)Sensor Networks (funded by Australian Research Council and Ajax Fastners) Optimised Control Systems for Integration of Wind Farms (Funded by Rowden Optimised Control Systems for Integration of Wind Farms (Funded by Rowden

White Foundation)White Foundation) Sequence Data Analysis (funded by Australian Research Council)Sequence Data Analysis (funded by Australian Research Council) Recearch Network on Intelligent Vehicles and Road Infrastructure- Seed Recearch Network on Intelligent Vehicles and Road Infrastructure- Seed

funding stage (funded by Australian Research Council)funding stage (funded by Australian Research Council) Projects completed in 2003Projects completed in 2003

– Self Organising Neural Networks (funded by Australian Research Council)Self Organising Neural Networks (funded by Australian Research Council)

– Neuro-Fuzzy Systems (funded by Advanced Engineering Centre for Manufacturing)Neuro-Fuzzy Systems (funded by Advanced Engineering Centre for Manufacturing)

OutlineOutline

Introduction to Sensor NetworksIntroduction to Sensor Networks– Recent advancesRecent advances– Smart SensorsSmart Sensors– Attributes of Sensor NetworksAttributes of Sensor Networks– Research issuesResearch issues

Smart Bolt sensor systemSmart Bolt sensor system– Tension monitoringTension monitoring– Wired, wireless and hybrid systemsWired, wireless and hybrid systems– Potential applicationsPotential applications

Advances in Sensor TechnologyAdvances in Sensor Technology– Miniaturization, Materials and Miniaturization, Materials and mechanics of processesmechanics of processes

– New sensor capabilitiesNew sensor capabilities– Wireless, low power, smartWireless, low power, smart

Advances in Information ProcessingAdvances in Information Processing– Real-time assessmentReal-time assessment– Information fusion algorithmsInformation fusion algorithms– Various communication protocolsVarious communication protocols

Recent Advances in the areaRecent Advances in the area

““Intelligent” or “smart” Intelligent” or “smart” sensorssensors

Sensors with an onboard Sensors with an onboard microcontroller/Electronics and some microcontroller/Electronics and some means for communicationmeans for communication

Capable of local or distributed Capable of local or distributed information processinginformation processing

Thousands of smart sensors can Thousands of smart sensors can collaborate to achieve a larger goalcollaborate to achieve a larger goal

Nodes link each other to form a Nodes link each other to form a communication framework called communication framework called Sensor Sensor NetworkNetwork

What is a Sensor NetworkWhat is a Sensor Network A connected network of sensor A connected network of sensor

nodesnodes Sensor networks can be classified Sensor networks can be classified

into 3 groups based on their into 3 groups based on their medium of communication:medium of communication:– Fully wired sensor networks, e.g. Fully wired sensor networks, e.g.

intelligent cars, robotsintelligent cars, robots– Hybrid (Wired and Wireless) Hybrid (Wired and Wireless)

sensor networks, e.g. large sensor networks, e.g. large mechanical structures such as mechanical structures such as bridgesbridges

– Wireless sensor networks Wireless sensor networks

CHCH

BASE

A clustered hybrid sensor network

7

Ad hoc Sensor NetworksAd hoc Sensor Networks Wireless sensor networks divided into:-Wireless sensor networks divided into:-

– infrastructure supported networks infrastructure supported networks – ad hoc networksad hoc networks

Ad hoc sensor networks:Ad hoc sensor networks:– A network formed on the fly when a group of Smart A network formed on the fly when a group of Smart

Sensors are deployed in a target environmentSensors are deployed in a target environment– Wireless is the most common form of Wireless is the most common form of

communication linkcommunication link– mobility is allowed and usually battery poweredmobility is allowed and usually battery powered– Some Wireless Protocols: Bluetooth, IEEE 802.15, Some Wireless Protocols: Bluetooth, IEEE 802.15,

Customized solutions Customized solutions

Research NeedsResearch Needs New sensors development and deploymentNew sensors development and deployment Smart materials for sensors and actuatorsSmart materials for sensors and actuators (controllable solids and fluids, electric-(controllable solids and fluids, electric-

restrictive & magneto-restrictive, restrictive & magneto-restrictive, piezoelectric, optical fibers….)piezoelectric, optical fibers….)

Innovative information fusion and protocolsInnovative information fusion and protocols– networking strategies and protocolsnetworking strategies and protocols– multisensor data fusionmultisensor data fusion– signal processingsignal processing

Battery technologyBattery technology

Smart Sensor ArchitectureSmart Sensor Architecture

P S O U W PE PR L Y

SENSORS

ADC

MICRO PROCESSO

R

MEMORYRADIO

REAL TIME OS

ALGORITHMS

1 kbps – 1 Mbps, 3 – 100

m, Lossy Transmission

Limited Lifetime Processing

Element

Require Supervisio

n

Slow Processing

Limited Memory

Sensing Element

Communication

Some ImplementationsSome Implementations

Micro Strain

ISC-iISC-iBlueBlue RF MOTE

EMBERJPL Sensor

WEB

((a) Direct a) Direct TransmissionTransmission

(b) Multi-hop (b) Multi-hop TransmissionTransmission

(c) Clustered (c) Clustered Transmission*Transmission*

* * Cluster Head (CH) Cluster Head (CH) coordinates the coordinates the communication communication within a cluster, within a cluster, process received process received information, and information, and maintain the link maintain the link with base station.with base station.

(a) clusterhead

(c)

Figure: Three Examples of Wireless Sensor Networking Protocols

head

ApplicationsApplications Detection of moisture levels in soilDetection of moisture levels in soil Gathering of environmental and weather Gathering of environmental and weather

informationinformation Information gathering in in-hospitable Information gathering in in-hospitable

physical environments such as remote physical environments such as remote geographic regions (e.g. Antarctic) or toxic geographic regions (e.g. Antarctic) or toxic urban locationsurban locations

Hazard Monitoring (Environmental, Hazard Monitoring (Environmental, Chemical/Biological)Chemical/Biological)

Homeland Security, safety, and border Homeland Security, safety, and border protection, battlefield monitoring protection, battlefield monitoring

Issues in Sensor NetworksIssues in Sensor Networks CommunicationCommunication

– Limited Range (~10 m)Limited Range (~10 m)– Limited Bandwidth (1 Mbps)Limited Bandwidth (1 Mbps)

Computation ResourcesComputation Resources– Memory (No large Routing Tables)Memory (No large Routing Tables)– CPU (No complex route discovery protocols)CPU (No complex route discovery protocols)

Fault ToleranceFault Tolerance– Link FailureLink Failure– Node FailureNode Failure

Power SupplyPower Supply– Battery operated (No periodic broadcast for Battery operated (No periodic broadcast for

route discovery)route discovery)

How to Form ClustersHow to Form Clusters

How sensors How sensors organise organise themselves into themselves into clusters. clusters.

Node distances: Node distances: Intracluster (dIntracluster (dijij), ), Cluster-base (DCluster-base (Djj))

Assume no Assume no intercluster intercluster communicationscommunications

Identification of Identification of CHj, nCHj, njj,&k for ,&k for all j is expectedall j is expected

n2 sensors

j=2

j=1

n1 sensors

Dj

BASE

n3 sensors j=3

dij

CH1

CH3

CH2

Cluster formation k=3

Energy Efficient Cluster Placement (Centralised)Energy Efficient Cluster Placement (Centralised) The total energy loss for the sensor system (EThe total energy loss for the sensor system (Etotaltotal):):

Research based on Energy ModelResearch based on Energy Model

))VCqNrb100(K)VCNb100)(kn()n

Dd((b1.0E 2

avgcyc2

avgcyctotalj

2jk

1j

n

1i

2ijtotal

Etotal is the total energy dissipated by all k cluster heads and all n-k wireless sensors

Comparisons for 70 sensor Comparisons for 70 sensor nodesnodes

FCM (left) and Subtractive (right) algorithms on the same 70 sensor nodes

Energy Loss due to distanceEnergy Loss due to distance

0

2

4

6

8

10

12

14

16

18

10 20 30 40 50 60 70 80 90

K-meanFCM

No. of nodes

Energy loss

No. of cluster createdNo. of cluster created

0

1

2

3

4

5

6

7

8

10 20 30 40 50 60 70 80 90

K-meanFCM

No. of nodes

No of clusters

Energy Efficient Cluster Energy Efficient Cluster Placement (Decentralised)Placement (Decentralised)

Highly mobile sensor nodesHighly mobile sensor nodes rotating cluster heads (occasional births and rotating cluster heads (occasional births and

deaths)deaths) only local information is used (LEACH only local information is used (LEACH

protocol):protocol):– each qualified sensor will offer itself to become a each qualified sensor will offer itself to become a

CHCH– locally selected CHs broadcast their statuslocally selected CHs broadcast their status– CH status information is lost in the next stepCH status information is lost in the next step

SMARTBOLT™SMARTBOLT™ System System

SMARTBOLT™ is a fully wired sensor system SMARTBOLT™ is a fully wired sensor system that finds the tension in a mechanical bolt.that finds the tension in a mechanical bolt.

SMARTBOLT™ can also be used as a part of SMARTBOLT™ can also be used as a part of a hybrid sensor system.a hybrid sensor system.

a prototype of a wireless SMARTBOLT™ a prototype of a wireless SMARTBOLT™ sensor system with an interface to PDA is sensor system with an interface to PDA is developed.developed.

Why we need this new systemWhy we need this new system

Convenience of operating a smaller, lighter unit (e.g. Convenience of operating a smaller, lighter unit (e.g. Palmtop) in data acquisition --> more economical.Palmtop) in data acquisition --> more economical.

Handheld computing is advancing at an exponentially Handheld computing is advancing at an exponentially increasing rate --> familiarity, marketingincreasing rate --> familiarity, marketing

Single point of control for multiple units --> large Single point of control for multiple units --> large number of units can be easily and accurately number of units can be easily and accurately managed through a single Palmtop.managed through a single Palmtop.

Applications of SmartboltsApplications of Smartbolts

Large mechanical structures Large mechanical structures (bridges, aircraft, vehicles used in (bridges, aircraft, vehicles used in mining sites…)mining sites…)

Manufacturing equipmentManufacturing equipment buildingsbuildings automotive partsautomotive parts ……....