Transcript Mobiscopes for Human Spaces

By: Tarek Abdelzaher, Yaw Aanokwa, Peter Boda, Jeff Burke,
Deborah Estrin, Leonidas Guiba, Aman Kansal, Samuel
Madden, Jim Reich
Presentation By:
Ankit Gupta
About the talk:
 General Idea
 Why Mobiscopes?
 Classes of Mobiscopes
 Common Requirements
 Mobility and Sampling coordination
 Heterogeneity
 Privacy
 Networking Challenges
 Human Factors & Social Implications
 Conclusion
General Idea
 Federation of distributed mobile sensors
 Why?
 Covering large areas can be challengeing


Unavailability of wired power
Expense of purchasing & maintaining enough devices
The paper focuses on the challenges and opportunities
Mobiscopes pose in human spaces.
Classes of Mobiscopes
 Vehicular Mobiscopes
 For traffic and automotive monitoring
 Equipped vehicle senses various surrounding conditions
Benefit:
Exploit oversampling provided by dense vehicle traffic
Examples:
Inrix, EZCab, NavTeq, TeleAtlas etc.
 HandHeld Mobiscopes
 Could be useful for



Monitoring health impact of exposure to highway toxins,
Monitoring an individual’s use of transportation systems,
Gather real time information about civic hazards & hotspots.
Common Requirements
 Data persistence must be assured
 Data access tends to be spatially correlated with the
user’s location & can change rapidly
 Human in the loop as an actuator, sensor, interpreter,
or responder
 Sensors & data to be shared by many public and
private entities
 Trust, coordinated deployment and respect of users’s
privacy
 This all leads to:
 General architecture and design guidelines for future
Mobiscopes
 Component reuse and reduction in development costs
 Interoperability amongst future systems
Mobility and Sampling
Coordination
 Performance depends on patterns of transporters
 Highly structured (Road traffic)
 Less structured (foot traffic)
 Sensor densities
 Sensing device’s availability can depend on user behavior or
device characteristics
 Application Adaptation
 Must adapt to network’s available communication
characteristics
 Could buffer data when connectivity unavailable
 Actuated Mobility


Task some or all nodes to visit a specific location to collect
information on demand
Task actuators to visit some areas either one at a time or as part of
a circuit
 Opportunistic connectivity
 Building low-level network protocols to quickly identify
and associate with nearby node (or networks)
 Routing algorithms to deliver data through such
opportunistic connections
 Prioritization


Buffered data to be prioritized
Prioritization to avoid wasting valuable bandwidth when different
nodes cover overlapping geographic areas
Challenges and opportunities of
heterogeneity
 Mobiscopes take on various topologies & structures
 Federate devices with different capabilities
 Draw together components with varying levels of trust
& credibility
 Benefits:
 Immune to weaknesses of sensing modalities
 Robust against defective, missing or malicious data
sources
 Heterogeneity of Ownership
 Individually owned devices
 Owners might not be trustworthy
 Might not maintain their devices in good condition
 Data Resolution & Types



Derive & maintain metrics at multiple resolutions
Simple interpolations (smoothly varying, temperature)
Complex models (faster varying or sparse data)
 Robustness
 Model driven approaches like Kalman filters & Particle
filters adapt well to irregular sampling
Tackling data Privacy
 People’s ability to control information flow about
themselves
 Definition
 Inability to publicly associate data with sources could
lead to los of context
 Revealing too much context can potentially thwart
anonymity, violating privacy requirements
 Local Processing
 Putting the selectivity and filtering capabilities on the
end-user
 Verification


Important to develop systems where users can verify data’s
correctness without violating the source’s privacy
Proper incentives to promote successful participation, prevent
abusive access with the purpose of “Gaming the system”
 Privacy preserving data mining
 User isn’t willing to share his or her data, but might be
interested in the result of aggregation over the target
community
 Could use additive random noise to perturb data
withour affecting the statistics to be collected
Networking Challenges
 Shifts the networks main utility from data
communication to information filtering
 Need for network storage as a key service because
aggregation and filtering both imply a need to buffer
Human Factors and Social
implications
 Considering broader policy precedents in information
privacy
 Extending popular education on IT’s new observation
capabilities
 Facilitating individual’s participation
 Helping users understand & audit their own data
uploads
 User Interfaces
 Missing from traditional embedded systems
 Opportunity for ambient and explicit feedback to the
user
 Help users configure their sensing participation
 Provide feedback on operational status
Conclusion
 Much research still needs to be done
 Much work still needs to be done on
 Platforms & API’s that offer efficient, robust, private &
secure networking & sensory data collection in the face
of heterogeneous connectivity and mobility
Questions
 ???