Mobile Sensor Webs: Sweet Dreams vs' Nightmares

1
Mobile Sensor WebsSweet Dreams vs. Nightmares

• Phillip B. Gibbons, Intel Research Pittsburgh
• MobiSensors07
• Industrial Views and Experience

1/16/07
2
Mobile Sensor Web Vision

• Harnest the sensing, computing, and communication
  capabilities of a billion mobile wireless devices
  to support valuable apps

Pervasive multimedia sensing
3
A Pervasive Sensing Paradox

• While the value of a sensing system increases
  with the number capabilities of its sensors, so
  too do the nightmares
• A (mobile) sensor web is a distributed data
  management nightmare
• Workload is write-intensive, high volume, often
  time-critical
• Heterogeneity in data types, platforms, networks
• Mobility, Failures common, Shared

4
IrisNet An Architecture for a Worldwide Sensor
Web

• Planet-wide local data collection/storage
• Multimedia sensors from cell phones, PDAs, UMPCs
• Query all relevant sensor data at once
• Do for live data what Google does for content
• Robustness, data integrity privacy
• Easy deployment of sensing services
• IrisNet Approach
• Exploit processing power near the sensors
  Use application-specific sensor feed filtering
• Push queries to sensors compute answers
  in-network
• Provide load balancing, caching, fault tolerance,
  etc for all services

5
A Pervasive Sensing Paradox

• While the value of a sensing system increases
  with the number capabilities of its sensors, so
  too do the nightmares
• Multimedia sensors raise huge privacy concerns
  Everyone has big brother power
• Data management must address theseprivacy
  concerns, to the extent it can
• IrisNet uses privileged privacy filters
• Only tip of iceberg
• Leverage devices CPU power

6
Sweet Dreams vs. Nightmares

• The apps deriving the most benefit from
  collective, pervasive multimedia sensing will
  encounter the most resistance
• Collective-sensing apps
• Require dense multimedia sensing coverage
• Inferences observations of other people
  places
• U.S. legal climate heightened privacy
  sensibilities gt many valuable apps deployed only
  elsewhere

7
Sweet Dreams vs. Nightmares

• The apps deriving the most benefit from
  collective, pervasive multimedia sensing will
  encounter the most resistance
• Self-sensing apps are ok
• Collect data about a person solely for that
  person
• Or other trusted party guardian, health care
  provider, trainer, etc.

8
Examples of sensing, modeling and supporting
everyday behaviors

• Meeting Effectiveness
• - Sense speech dynamics of participants
• Model turn taking, interruptions, dominance,
  roles, level of engagement
• Support feedback on meeting dynamics,
  effectiveness, participation

• Autism
• - Sense speech dynamics/distortions (norms for
  F2F interaction patterns ? goal)
• Model turn taking, interruptions, intonation,
  level of engagement
• Support real time coaching to kids on how to
  modify behavior appropriately

• Eldercare, ADLs
• Sense everyday interactions with objects, rooms,
  locations
• Model high-level behavioral activities (ADLs)
• Support real time reminding, escalated alerting,
  feedback on wellbeing, patterns anomalies

• Fitness, physical activity
• Sense everyday physical activities (e.g.,
  walking, sitting) and their durations
• Model high-level physical activities (e.g., a
  run, a bicycle ride)
• Support real time awareness feedback - goals,
  automated journaling, patterns anomalies,
  methods to motivate sustained behavior change

Project at Intel Research Seattle
9
UbiFit

• Improving fitness through mobile devices

MSP - behavioral monitoring technology
UbiFit garden MSP persuasive technology
Project at U. Washington and Intel Research
Seattle
10
Patient Monitoring
Ear-piece SpO2, Skin temp 3d accel

• Integration of physiological, activity and
  environmental monitoring
• Ear-piece SpO2, skin temperature, 3d
  accelerometer
• ECG, 3d accelerometer
• Multi-Sensor Platform / Imote2 Activity
  Inferences
• IMote Environmental Temperature, humidity
• Phone on-body aggregator
• Connect to various sensors
• Data aggregation and processing
• Visualization of latest scalar data
• PC Displays historical data

ECG, 3d accel
MSP/Imote2 Activity Inference
Imote Environ- mental Temp Humidity
PC (visualization)
Project in Intel DHG