2 of 26

Why this paper?

Demonstrates how some of the information you are collecting might be analyzed/used
Interesting routing problem
Controversial? application

SCADA - (Supervisory Control and Data Acquisition)

3 of 26

Introduction

Sensors are deployed quite pervasively

thermostats
cars
scientific applications

download data via USB

"How can we make wide-area communication more accessible to sensornets?"

collect data that is currently only available at the sensor/in a restricted region

thermostat/energy meter

allows data collection without manual intervention

field deployments

4 of 26

Introduction

If sensors could communicate over the wide-area, data could be sent to a centralized collection point
Wide-area technologies

Satellite

too expensive
requires line-of-sight

Cell data

also expensive
some coverage issues

WiFi

spotty coverage
"free"?

5 of 26

Recall: Typical Architecture

from Fidelity and Yield paper

6 of 26

Data Muling

Convenient acronym: Mobile Ubiquitous LAN Extension

Idea: a mule opportunistically (or intentionally) collects data from a sensor without WAN coverage

perhaps can also deliver data to a sensor
mule then transfers when WAN is accessible

7 of 26

Contributions

Mobile phone implementation
Four deployments
Feasibility study based on human mobility

8 of 26

Why phones?

Ubiquity

muling can come for "free"

General purpose computing platforms

multiple radios and WAN connectivity

Can provide an interface to sensors

Discussion question: useful for all possible applications?

9 of 26

Applications

Garbage Bins

can tells a hiker that it is full and hiker reports the information when WAN is available

Habitat Monitoring

lower cost by replacing 3G/satellite connection

Car Blackboxes

use bluetooth to transfer car stats to driver's phone

Personal Energy Monitoring

collect a view of energy usage at home/work/public spaces

Hiking Water Quality
Others?

10 of 26

Challenges

How effective are current mobile phones as data mules?
Is data muling feasible for traditional sensornet applications?
How well does human mobility support data muling?

11 of 26

Architecture

Sensors
Data mules
Gateways

12 of 26

Architecture

The assumption is that sensors all have bluetooth

Reasonable assumption?

Mesh + mule
Mule-to-mule communication

Discussion question: is this a reasonable general purpose architecture?
Discussion question: what factors might impact design of the architecture?

properties of a particular application

13 of 26

Data Sets

Mobility/MIT and Mobility/ISI

bluetooth contact information

Subsidence/Oilfield

GPS units deployed in oilfield
human intentionally drives to deployment

Subsidence/Urban

controlled experiment

People/ISI

used bluetooth to track humans in office/home

14 of 26

More Discussion

They use stationary bluetooth devices seen as "stand-ins" for sensors
Discussion question: is this a good choice?

15 of 26

Feasibility Results

815 stationary bluetooth devices seen in MIT trace
Useful result?

16 of 26

Feasibility Results

Contact patterns

17 of 26

Oilfield Deployment

The mule must intentionally go to the deployment
Four sensors communicate with mule via bluetooth
"muling halved the time outages were unknown, from 60 sensor-days to 27"

agree or disagree?

18 of 26

Urban Sensing

Oilfield sensors deployed behind a house
Only 7% of contacts were opportunistic (using bluetooth)

WiFi experiment?

19 of 26

Office Sensing

"opportunistic muling works very well when radio range and mobility patterns are well matched"

20 of 26

Observations

So far...focus primarily on contact opportunities
Fine for intentional muling, but what about opportunistic?
How long do I have to be near the sensor?
Will this eat my battery?

21 of 26

Loiter Time

Time to transfer data to mule via bluetooth
17 second setup time - connection and initial request

22 of 26