Hong Lu , Nicholas D. Lane, Shane B. Eisenman, Andrew T. Campbell Bubble-Sensing: A New Paradigm for Binding a Sensing Task to the Physical World using Mobile Phones Hong Lu , Nicholas D. Lane, Shane B. Eisenman, Andrew T. Campbell Dartmouth College
Idea Problems Idea Bubble-Sensing hong@cs.dartmouth.edu Goal Attach sensing request to the area of interest Problems Hold the bubble in the area of interest Recover from lost bubble Exploit heterogeneous devices Idea Opportunistically leverage the devices that remain in the area or pass through to fulfill the sensing request and maintain the binding Bubble-Sensing hong@cs.dartmouth.edu
Bubble Sensing Virtual roles Idea Sensing Bubble Bubble creator Opportunistically leverage cell phone density within the area to maintain coverage Sensing Bubble (action, region, duration) Virtual roles Bubble creator Bubble anchor Bubble carrier Sensing node Take a photo here !! Bubble-Sensing hong@cs.dartmouth.edu
Bubble Creation Phase Bubble-Sensing hong@cs.dartmouth.edu Bubble Server Bubble Anchor Mobile Sensor Bubble Creator Bubble-Sensing hong@cs.dartmouth.edu
Bubble Maintenance Phase Bubble Server Bubble Anchor Bubble Anchor Simulation of three representative techniques: Hop counting Centroid MCL Zebranet based mobility model. Techniques sensitive to beacon or node density. Failure to produce location estimates (or estimates are highly inaccurate in the case of MCL) Although required, existing localization schemes are weak. One of the few examples of these sensor networks – zebra net used for this simulation. Point was to examine the performance of three approaches to localizatoin and how they would fare In this environment. They perform poorly. Quickly describe the simulation. Hey what are the axis. Tell tell them exactly what is shown. And which are the schemes and the setup. And note the MCL issue. Question someone may ask: - be ready for it - what is MCL? What is centroid? What is Amphrhous??? MAKE SURE YOU RE-READ THESE PAPERS!!! Be nice to maybe talk about other applications of monte carlo??? In other areas? Mobile Sensor Mobile Sensor Bubble-Sensing hong@cs.dartmouth.edu
Bubble Restoration Phase Bubble Server Mobile Sensor Bubble Anchor Bubble-Sensing hong@cs.dartmouth.edu
Bubble Restoration Phase Bubble Server Bubble carrier Mobile Sensor Bubble Anchor Bubble-Sensing hong@cs.dartmouth.edu
Test bed Implementation Device Nokia N95, Nokia N80 + BlueCel dongle Pys60 WiFi based communication Ad-Hoc mode infrastructure mode 4mw Beacon based WiFi localization The sensing task: Take sound clips Bubble-Sensing hong@cs.dartmouth.edu
Sensing Coverage over Time Make sure you discuss how you built the classifier and what it does. What is the feature vector? Then discuss the performance of it. Make sure you read up about decision trees and the specifics of j48 before you forget the details. Bubble-Sensing hong@cs.dartmouth.edu 9
Fidelity Bubble-Sensing hong@cs.dartmouth.edu Make sure you discuss how you built the classifier and what it does. What is the feature vector? Then discuss the performance of it. Make sure you read up about decision trees and the specifics of j48 before you forget the details. Bubble-Sensing hong@cs.dartmouth.edu
Related Work Mobile phone as a sensing device UCLA, UIUC, MIT, Ohio State Nokia, Motorola, Intel, Microsoft Exploit location information Routing, Geocast geographic storage, DCS, GHT Theoretical work Virtual static node Where to next? Analysis of ABL within a full scale testbed Stocastic analysis of system Realistic outdoor experiments. Using metrosense testbed. Bubble-Sensing hong@cs.dartmouth.edu
Thanks for Listening Bubble-Sensing hong@cs.dartmouth.edu
Sensing Coverage over Time Make sure you discuss how you built the classifier and what it does. What is the feature vector? Then discuss the performance of it. Make sure you read up about decision trees and the specifics of j48 before you forget the details. Bubble-Sensing hong@cs.dartmouth.edu