2. Mobile Participatory Computing
Foursquare allows 25 million users to save and
share information about the places they
visit, participating in creating a location-aware
service.
How can we generalize this model and the
supporting infrastructure?
How can we further exploit sensors, and
intelligence via mobile phones?
3. This Presentation
1. Lessons from the Past
2. The Mobile Participatory Computing Revolution
3. Pilot experiments
4. Towards a software ecosystem
4. This Presentation
1. Lessons from the Past
2. The Mobile Participatory Computing Revolution
3. Pilot experiments
4. Towards a software ecosystem
5. Before Peer-to-Peer (P2P)
Until the late 1990s the client-server model
dominated the internet:
Powerful servers provide services.
Dumb clients consume services.
This made sense in the early days, when home
computers were incapable of providing
services.
6. P2P Principles
People took notice of the growing pool of
untapped resources at the edge.
If edge resources also provide services:
Server costs would be greatly reduced.
Users could participate in sourcing media.
We move from dumb client and powerful servers
to networks of equal peers.
7. Revolutionary Applications
Napster:
User spare storage space at the edge and users files to
create a music sharing service.
Peaked at 1 million users in 2000.
Seti@Home:
Uses spare compute power at the edge to search radio
signals for evidence of alien life.
Holds record for worlds longest computation.
All of this achieved using free edge resources and
without expensive servers.
8. Lessons of the P2P Revolution
1. Well-connected resources can be used as a
platform to build services.
1. Given incentives, users participate in
providing services.
2. Media and other businesses should be ready
for disruptive technologies.
9. This Presentation
1. Lessons from the Past
2. The Mobile Participatory Computing
Revolution
3. Pilot experiments
4. Towards a software ecosystem
10. The Coming Revolution
Mobile resources exceed those that supported
the P2P boom:
PC 1999: 300MHz, 32MB RAM, 8GB HDD.
Phone 2012: 600MHz, 256MB RAM, 4GB SD.
Connections are also improving:
PC 1999: 512kbps cable modem.
Phone 2012: 700kbps UMTS connection.
Once again, we have a large pool of
untapped resources at the edge.
11. Key Differences
Mobile devices are relatively restricted in terms
of computation, memory and storage.
Yet, they have other notable features:
Ubiquitous connection to users.
Sensing of the environment.
Mobile roaming over large areas.
Can we use mobile devices, their sensors
and their users to implement services?
12. Animals as Mobile Sensors
How can you sense the
temperature of arctic waters
at various depths?
You could build a complex
robot… or you could just ask
a seal to do it for you.
What should we ask people
to do for us?
13. Use Case: Social Reporting
Belgium is flooding. “TV News-Show” wishes to
report on the floods in real-time:
A network of trusted individuals is dynamically
established in each city and asked to report on flood
conditions; there is bad flooding in Ghent.
A 2nd network is dynamically established to GPS units of
cars driving in Ghent. Analysis shows that traffic Is
stationary in the city centre.
Personalized news reports are issued to citizens of
Ghent with real-time traffic and weather data.
TV-News-Show provides personalized reports on a
news event using citizen reporters.
14. Use Case: Road Monitoring
Road Quality Monitoring:
When users activate GPS directions on their phone, the
accelerometer reports location-tagged vibration data.
Aggregated vibration data is used to map road quality.
Users who provide data are rewarded with free access to
the road quality map, while other users pay for the service.
The „killer app‟ will not be written by computer
scientists, we must make it easy for domain
experts to create such apps.
15. This Presentation
1. Lessons from the Past
2. The Mobile Participatory Computing Revolution
3. Pilot experiments
4. Towards a software ecosystem
16. Supporting Middleware
Consistent component-based development
approach on all platforms:
Embedded experts develop reusable components.
Application developers rapidly assemble components to
form distributed applications.
Support for remote management:
On demand deployment of new software.
Runtime reconfiguration of applications.
A federated security model allows for trusted
deployment, and use of 3rd party hardware.
17. The LooCI Middleware
Originally developed for Wireless Sensor
Networks.
Runs on very limited resources:
20MHz CPU, 16KB RAM, 48KB Flash.
Platform and language independent:
C on Contiki OS
Java ME on Sun SPOT
Java SE on Android
18. The LooCI Middleware
Originally developed for Wireless Sensor
Networks.
Runs on very limited resources:
20MHz CPU, 16KB RAM, 48KB Flash.
Platform and language independent:
C on Contiki OS
Java ME on Sun SPOT
Java SE on Android
19. The LooCI Middleware
Originally developed for Wireless Sensor
Networks.
Runs on very limited resources:
20MHz CPU, 16KB RAM, 48KB Flash.
Platform and language independent:
C on Contiki OS
Java ME on Sun SPOT
Java SE on Android
20. Build on Social Networks
Access to millions of users:
Facebook: > 800 million users.
Twitter: > 150 million users.
A mechanism to recruit users for apps.
Already deployed on mobile devices.
Avoids NAT and firewall issues.
21. LooCI on Social Networks
Event Bus is designed for IEEE 802.15.4:
Limited packet size and
limited number of transmissions.
Twitter has similar constraints, so mapping the event
bus to twitter was easy.
Now LooCI components talk to each other and
users via Twitter (soon also Facebook).
The LooCI binding API needed no modification.
22. A Live Application
Follow my heart-rate on Twitter during this
session at: www.twitter.com/WSNTeam4
23. A Small Experiment:
4 users, 3 days, 2 countries
UDP and Twi er Device Availability
100
80
Availabiliy (%)
60
40
20
0
User1 User2 User3 User4
Details in: “Enabling Massive
Scale Sensing with the @LooCI
Mobile Sensing Framework”, to
appear in proc. of EUC‟12.
24. Related Work
Relevant applications: FourSquare, Google
Maps, Bodyblogger, etc.
Useful middleware:
LooCI, RUNES, Pogo, AnonySens, Cartel, Pris
m, etc.
Work from P2P field on trust, security and
economic models.
All of these are pieces of the puzzle, but we
need a full software ecosystem.
25. Results of Pilot Experiment
We have the basic infrastructure for creating
mobile participatory applications.
LooCI runs on any Android phone, providing
consistent central deployment, administration
and reconfiguration.
We have a small but growing test-bed in
Belgium and soon also Australia.
26. This Presentation
1. Lessons from the Past
2. The Mobile Participatory Computing Revolution
3. Pilot experiments
4. Towards a software ecosystem
27. A Job for iMinds!
We need models of data ownership: the user
should retain control of their data once it leaves their
device.
Economic models to encourage user participation
in mobile applications.
Accounting and control of user costs such as
battery, bandwidth & disruption.
Rich models of trust and privacy to encourage user
participation.