AN APPROACH FOR TRACKING WILDLIFE USING WIRELESS SENSOR NETWORKS - PowerPoint PPT Presentation

AN APPROACH FOR TRACKING WILDLIFE USING WIRELESS SENSOR NETWORKS Francine Lalooses* † Hengky Susanto* Chorng Hwa Chang* * Tufts University † MITRE Corporation Approved for Public Release; Distribution Unlimited. Case Number 07-0512

Outline • Overview • Overview – Introduction – Introduction – Tracking moving target – Tracking moving target – Related work – Related work • Sensor network for habitat monitoring • Sensor network for habitat monitoring – ZebraNet – ZebraNet – Great Duck Island – Great Duck Island • Distributed Predictive Tracking Algorithm • Distributed Predictive Tracking Algorithm • Problem with Target Tracking • Problem with Target Tracking • Tracking Failure Recovery • Tracking Failure Recovery • Quandary of Recovering from Failure • Quandary of Recovering from Failure • Future Work • Future Work

Introduction: Wireless Sensor Networks • Consist of sensors and base stations • Purpose of tracking in sensor networks is a necessity for many applications – Including computer vision, tactical battlefield surveillance, perimeter security, emergency response, animal tracking

Analogy Illustration

Purpose of Tracking • For wildlife habitat monitoring • Sensors only monitor land targets (animals) • Animals are tagged • Sensors purposely placed at certain locations • Saves time finding target in the region • Better understanding of region/animal relationship

Tracking Moving Target • Sensors: – Hibernate to conserve energy – Awaken when target detected entering the monitored area – Sensors record information of the target – Relay information between sensors and follows target movement

Tracking Moving Target in Wildlife Awake Hibernating

ZebraNet • Track animals long term and over long distances • GPS enabled • All nodes mobile • Zebras are tagged with RFID • Peer-to-peer routing and data storage

Great Duck Island • Biologists put sensor devices on Maine's Great Duck Island – In underground nests – 4-inch stilts placed just outside their burrows • Record data about the birds • Sensors relay information to a gateway node (the base station) – Gateway node transmits information to • a laptop • then to satellite dish • ultimately, to an Intel Research lab at Berkeley California

Great Duck Island

Distributed Predictive Tracking • No central point • Cluster based architecture • Assumptions: – Randomly distributed sensors – Default to normal beam – Hibernation mode • Predictive mechanism • Wakeup all sensors for recovery A Protocol for Tracking Mobile Targets using Sensor Networks, RPI

Outline • Overview – Introduction – Tracking moving target – Related work • Sensor network for habitat monitoring – ZebraNet – Great Duck Island • Distributed Predictive Tracking Algorithm • Problem with Target Tracking • Tracking Failure Recovery • Quandary of Recovering from Failure • Future Work

Problem Tracking Moving Target in Wildlife What if . . .

Lost Target Conditions • Network failure : When node that is currently monitoring target fails to wake the next node • Prediction failure: Failure predicting where the target is heading • Multiple sensors interaction: Failure to recognize multiple targets at once • Hardware malfunction: Hardware may malfunction or the battery is weakening

Outline • Overview – Introduction – Tracking moving target – Related work • Sensor network for habitat monitoring – ZebraNet – Great Duck Island • Distributed Predictive Tracking Algorithm • Problem with Target Tracking • Tracking Failure Recovery • Quandary of Recovering from Failure • Future Work

Approach in Tracking Failure Recovery • Goal – Finds the monitored target after failure occurs • Purpose: – Conserve energy – Quick and efficient method to recover from failure

How Does The Algorithm Work? • Analogy of lost tourist in Marrakesh – Keeping track the traveler – Lost the traveler in Hotel Kenzi Farah – Where to find the traveler • Check the popular places around the hotel – Casino, Jemaa El Fna, Mosque, Koutoubia • Establish the search region – Perform search for the lost traveler

Why This Approach • Studies show: – Animals tend to move in certain patterns • Along river • Tree – Animals are sensitive to temperature and chemical elements

Algorithm Summary • Establish search region – To limit the search area – Take advantage of hierarchical cluster • Finding popular place surrounding where the animal last seen • Perform search algorithm

Establish Search Region • Computes the diameter of search region – Uses approximate velocity of animal – Radius d = time t * V • Uses hierarchical cluster to find popular place in that region – Place that is often visited by animals such as {River, Trees, etc}

Hierarchical Cluster • Nodes form clusters after deployment • Each cluster select a leader or a Cluster Head (CH) • Only CH communicate to each other • Reduces traffic • Eliminates network collisions

Hierarchical Cluster • Cluster based algorithm • Hierarchical approach Master CH CH d First level CH CH CH d • Variables: – Distance d = Velocity * time

Search Algorithm • x = CH of the target’s last seen position • y = Identified CH popular place • Computes distance y to x y y X y y

Broadcast and Activate Nodes • X = the object’s last seen position • Broadcast from X • Activate sensors • Establish minimized region X

Minimizing Search Region • Calculation based on maximum hop and popularity Lost region 1h 2h 2d 4h • Variables: – h = CH hop count

Understanding Necessary Conditions • Coverage area • Sensor placement and terrain • Sensor limitation • Sensor reliability • Multiple sensor interaction and identification

Oops!! • What if the animal is not in the desired search region? – Awake all nodes in the initial search region

Outline • Overview – Introduction – Tracking moving target – Related work • Sensor network for habitat monitoring – ZebraNet – Great Duck Island • Distributed Predictive Tracking Algorithm • Problem with Target Tracking • Tracking Failure Recovery • Quandary of Recovering from Failure • Future Work

Quandary of Recovering from Failure • Long description to describe a single target – Due to limited information from sensor itself – Takes large amount of bandwidth – Requires processing power – Requires storage space in sensor – Not energy efficient • Moving target’s speed – Difficult to estimate the speed accurately

Outline • Overview – Introduction – Tracking moving target – Related work • Sensor network for habitat monitoring – ZebraNet – Great Duck Island • Distributed Predictive Tracking Algorithm • Problem with Target Tracking • Tracking Failure Recovery • Quandary of Recovering from Failure • Future Work

Future Work • More in depth understanding on – Relation between animal’s behavior and accuracy of establishing search region – Performance analysis evaluation – Network traffic analysis during recovery process

In Progress • Minimizing target description – More efficient description to describe the target • Improving the test bed in our sensor lab

Thank You Merci Shukran