Microsoft Research Faculty Summit 2007

Aman KansalResearcherNetworked Embedded Computing, MSR

PARK with people …and phones

UploadPictures, Video,Audio

1. Is the court wet?

3. Which bird soundsreported?

APPLICATION

GROUP MEMBER

Stitched view

SMS: Click picture of court.

Group Points: 400

SenseWeb

(Data centric coverage model)

2. What play structures are there?

2.14 billion phones and growingMobility

reach where static sensor cannotincreased spatial coverage

Phone exists for voice/data apps: Piggybacking sensing is cost effective

Human assistanceCan sometimes help detect or aim at interesting phenomenon

Client on phoneAllows users to take picturesAutomatically uploads data to serverLocation stamps using inbuilt/Bluetooth GPS

SenseWebSenseWeb

ServerIndexes images by location and time (SQL Server database)Web service API for phones and apps.Supports several sensor types

Example App: PortalDisplays sensor data by location and sensor typePublicly accessible at http://atom.research.microsoft.com/sensormap

Web service API’s allow building other apps.

Information valueWhich data to collect and share: battery and bandwidth constraints

Coverage managementWhich phone sensed where app needs coverage

Sensor tasking for application demandsIncentive mechanismsData verifiability, user privacy

Entropy of a single image: H(X) = -(p.log(p)) [p: image histogram]

Value among multiple imagesConsider common spatial coverage

H(X|Y) = -E[log2p(X|Y)]H(X|Y1,…,Ym) = H(X|Z) (Z: common spatial coverage)

Commonality: found using key feature based algorithm

10 20 30 40 50 60 70 80 900

0.5

1

1.5

2

2.5

3

Relevance Value Cut-off

Data

Siz

e (

MB

)

Relevance Value Cutoff (%)

Data

Siz

e (

MB

)

Buildings

Kitchen

Value based selection

Details: ACM Sensys WSW 2006

Which sensors does app accessWho sensed in required region during required time window?

Mobile Sensor Swarm

Which sensors does app access

Who sensed in required region during required time window?

Solution: locationSamples are geo-stampedApps do not track device

TrajectoryConnectivitySharing preferences

Device ID anonymized

Data CentricAbstraction

Mobile Sensor Swarm

Application 1

Application n

Several location technologies

GPS: does not work everywhereCell tower: coarseWifi: coarseHuman entered tags: approximate, high manual effort

Leverage camera data to enhance location

Refine location granularityRoom within building, aisle within store

Associate data when location not availableVerify location

i

jMij

AlgorithmImages within vicinity organized as a graphEdge weight by matchRelation R(i,j) by highest weightRefined location zone: Transitive closure of R

Details: ACM NOSSDAV 2007

Minimize sensing task overhead on phonesSense to be most accurate on most used regions

Good model: determine where sensing neededLearn most used: where apps need dataTask phones: battery, bandwidth, privacy, intrusion costs

Phenomenon

Demand Sensing cost

Details: Andreas Krause, Intern project report

Set V of possible observations For each subset A of V, define utilityU(A) = Σi E[Di (Var(Si) – Var(Si | A)) ]

Expectation over demand Di and observations A

0 20 40 60 80 1000

0.005

0.01

0.015

0.02

0.025

0.03

Number of observations (out of 534)

Dem

and-

wei

ghte

d va

rianc

e

Random selection

Optimized for variance reduction

Optimized fordemand-weighted variance

Theorem: U(A) is submodularTheorem [Nemhauser et al]: For submodular U:

U(greedy solution) > (1-1/e) U(optimal)

Mobile phones enable many sensing appsArchitecture to use a highly volatile swarm of mobile devices as a sensor network

Information value based data selectionLocation based data centric abstraction

Coverage management and data addressingAvoids burdening applications with managing device motion, connectivity, sharing

Efficient sensor tasking

Contact: kansal@microsoft.com

© 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.

The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after

the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.