MetroSense

People-Centric Urban Sensing

Andrew T. Campbell, Shane B. Einseman, Nicholas D. Lane, Emiliano Miluzzo, Ronald Perterson

Dartmouth College/ Columbia University

Focus of Sensor Network Research

• Industrial, structural, environmental monitoring systems, military systems, etc.

Characteristics of Existing Systems

• Small-scale, short-lived, mostly-static

• Application-specific

• Multi-hop wireless

• Very energy-constrained

• Mobility not an issues of driving factor

• People out of the loop

Sensor networks at cross roads?

They don’t impact our everyday lives, why?

Follow the people. Follow the money.

New Frontier for Sensing: Urban Timescape

Ron Fricke, timescape is a day in the life of a city (edited version)

QuickTime™ and aYUV420 codec decompressor

are needed to see this picture.

People-centric, mobility counts, scale matters.

Urban Sensing Apps.

• Noise mapping• http://www.noisemapping.org/

• Emotion mapping• http://biomapping.net

• Congestion charging• http://www.cclondon.com/

Congestion Map London

London Noise Map

Emotion Maps of London

Downing Street Noise Map

People-Centric Sensing Apps.

• Heath care applications• Emergency care (Codeblue), assited

living (AlarmNet)

• Recreational applications• Running (Nikeplus), dancing

(interactive dance ensembles)

• Urban gaming • http://www.comeoutandplay.org/

index.php

Emerging Urban Sensing Classes

• “Personal Sensing” for individuals• e.g., nikeplus, sensor-enabled cellphone apps., health care• Great potent for commercial success (catch the ipod generation) -

youthful early adopters

• “Peer Sensing” for groups• e.g., urban gaming, peers apps. • Could be an explosive growth because of existing gaming users

• “Utility Sensing” (system-wide) provides utility to a large population of potential users

• e.g., noisemapping, others• Providers (e.g., towns, organizations, enterprises) will have to

invest in build out - costly

Need to exploit existing computing, sensing, wireless breakthroughs and infrastructure to support these emerging urban sensing classes

What is MetroSense?

Architecture for large-scale sensing based on mobile sensors.

Captures interaction between people, and, between people and their surroundings.

Enables general purpose programming of the infrastructure.

Based on three design principles that promote low cost, scalability, and performance.

Importantly, mobile people-centric sensors run their own apps. (i.e., personal sensing, peer sensing), and, in parallel support “symbiotic sensing” (i.e., utility sensing, peer sensing) of other users in a transparent manner.

Gains scalability and sensing coverage via people-centric mobility, and its adaptive “sphere of interaction” design.

Goal is to study and evaluate an “opportunistic sensor network” paradigm

Characteristics of Existing Systems

• Small-scale, short-lived, mostly-static

• Application-specific

• Multi-hop wireless

• Very energy-constrained

• Mobility not an issues of driving factor

• People out of the loop

Characteristics of MetroSense

• Large-scale, long-lived, mostly-mobile

• Application-specific

• Multi-hop wireless

• Very energy-constrained

• Mobility not an issues of driving factor

• People out of the loop

Characteristics of MetroSense

• Large-scale, long-lived, mostly-mobile

• Application-agnostic

• Multi-hop wireless

• Very energy-constrained

• Mobility not an issues of driving factor

• People out of the loop

Characteristics of MetroSense

• Large-scale, long-lived, mostly-mobile

• Application-agnostic

• Very limited multi-hop wireless

• Very energy-constrained

• Mobility not an issues of driving factor

• People out of the loop

Characteristics of MetroSense

• Large-scale, long-lived, mostly-mobile

• Application-agnostic

• Very limited multi-hop wireless

• Not energy-constrained

• Mobility not an issues of driving factor

• People out of the loop

Characteristics of MetroSense

• Large-scale, long-lived, mostly-mobile

• Application-agnostic

• Very limited multi-hop wireless

• Not energy-constrained

• Mobility is a driving factor

• People out of the loop

Characteristics of MetroSense

• Large-scale, long-lived, mostly-mobile

• Application-agnostic

• Very limited multi-hop wireless

• Not energy-constrained

• Mobility is a driving factor

• People in the loop

Characteristics of MetroSense

• Large-scale, long-lived, mostly-mobile

• Application-agnostic

• Very limited multi-hop wireless

• Not energy-constrained

• Mobility is a driving factor

• People in the loop

• Security, trust, and privacy important

My sense of “urban” has changed … from here

.. to here

Sensing across a Large Area is Challenging

Sensing across a Large Area is Challenging Imagine the

Green Is Time Square ;-)

Sensing across a Large Area is Challenging• Some simple questions

one might ask• How many people are

sitting, running, walking on the Green?

• Where is Andrew on the Green?

• Noise, temperature, allergies distribution across the Green [now, 10AM-10PM, etc.]

• Others

Sensing across a Large Area is Challenging - what scales?

Ubisense

Tiered

Mesh

MetroSense

Building Campus Town City Scale

Co

st (

$)

Ubisense

Tiered

Mesh

MetroSense

Building Campus Town City Scale

Fid

elit

y (S

amp

les/

Are

a)

What is MetroSense?• Relies on the random or not so random

mobility of people• Task sensors to “collect” sensor data

and deliver it opportunistically• Offers in delay-tolerant sensing• Infrastructure

• Sensor Access Points (SAPs)• Mobile Sensors (MSs)• Static Sensors (SSs)

• Operations• Opportunistic Tasking, Sensing,

Collection

• Opportunistic Delegation Model (ODM)

Sensing across a Large Area is Challenging - a proposed Campus-wide Sensor Network

SAP locations - Aruba APs

Sensing Coverage using MetroSense

MetroSense InfrastructureSensor Access Point (SAP) People-centric sensing apps

Dartmouth Pulse BikeNetSensor devices

Interacts with static sensor clouds

MetroSense Operations

opportunistictasking

opportunisticcollection

opportunisticsensing

limitedpeering

comms & ground-truthsensing

Opportunistic Delegation Model (ODM)

sensing “space” of interest

• Goal is to extend sensing coverage• Application requires sensed modality ß from

“space” during [t1, t2]• Delegate “limited” responsibility for “limited” time• Direct and indirect delegation of roles

• Sensing, tasking, collection and “data muling”

• Enables new services (ODM Primitives)• Virtual sensing range, virtual collection range,

virtual static sensor, virtual mobile network

• Design challenges• Sensing range is dependent on modality• Limited comms. “rendezvous” time • Candidate sensor selection is challenging• Likelihood of a mobile reaching a targeted sensing

space is probabilistic in nature• Delay tolerant characteristics of sensing and

collection processes

TX range

sensing range

direct delegation

indirect delegation

Data mule

Virtual Sensing Range - an ODM Primitive/Service

Area of Interest - “sensing space”

Virtual Sensing Range

TX range

sensing range

Virtual Sensing Range - Experimental Result

Virtual Sensing Range

New Transports for Opportunistic Tasking and Collection

• Reliable, secure needs• Limited “rendezvous” time,

mobility, probabilistic sensing/collection, delay tolerant collection

• New transport needs• Lazy uploading• Lazy tasking• Direction-based muling

• Adaptive multihop

Mote broadcasting syncmsgs from GPS unit

GPS Unit

Measuring terrain slope

Measuring pedal speed

Measuring how fast i kick the assof inconsiderate motorists

Bikenet Road Warrior

Skiscape - @ Dartmouth Skiway

Existing Urban Sensing Initiatives

• Nokia’s SensorPlanet• http://www.sensorplanet.org/

• CENS Urban Sensing Summit (May 2006)• http://bigriver.remap.ucla.edu/remap/index.php/

Urban_Sensing_Summit

• CitySense (BBN/Harvard)• MetroSense (Dartmouth/ Columbia)• Others?

Conclusion

• Next wave in sensor networks is “people in the loop and not out of loop” sensor networks

• Scale and mobility matters and are challenging in terms of architectural design

• Didn’t talk about security, privacy, and trust that are central to this effort

What is MetroSense?

Ultimately, its about a new wireless sensor edge for Internet

Thanks for listening!