Table of Contents
N-SMARTS: Networked Suite of Mobile Atmospheric Real-Time Sensors
See the N-SMARTS homepage for more info.
Participants
N-SMARTS Platform
- Dr. Paul Aoki - Intel Research Berkeley
- Professor Eric Brewer - Berkeley CS
- Professor John Canny - Berkeley CS
- Professor Ronald C. Cohen - Berkeley, Chemistry
- R.J. Honicky - Berkeley CS
- Dr. John Huggins - CEO BSAC
- Dr. Alan Mainwaring - Intel Research Berkeley
- Dr. Eric Paulos - Intel Research Berkeley
- Professor Albert Pisano - Berkeley ME
- Paul Wooldridge - Berkeley, Chemistry
Enclosure Design
- Dr. Allison Woodruff - Intel Research Berkeley
- Professor Paul Wright - Berkeley ME
- Kyle Yeates - Berkeley ME
Advanced Sensing
- Dr. Justin Black - Berkeley EE
- Alex Elium - Berkeley EE
- Professor Albert Pisano - Berkeley ME
- Professor Richard White - Berkeley EE
A Mote on Steroids?
Why a mobile phone as a sensing platform
- Power solved
- Networking solved
- Existing economies of scale
- Mobile
- more coverage
- Periodic proximity offers opportunities
- Super-sampling for higher precision
- Automatic calibration
- Tracks a user
- Perfect for observing the world from a user's perspective
- “The action is where the people are at”
Challenges unique to the mobile phone as sensor
- Mobile
- User behavior / movement somewhat unpredictable
- Makes it difficult to observe a single location well
- especially locations infrequently travelled
- Sensor obstruction
- People put their phones in their pockets, purses, etc.
- Local environmental bias
- e.g. driving in your polluting bucket
- makes it difficult to make observations relevant to other people
The N-SMARTS Platform
The sensor board
- Show and tell
- Schematics
- TI MSP430 Microcontroller
- CSR BlueCore 2 bluetooth radio
- regulated power supply
- accepts any voltage 3.4 to 6.5 volts
- low noise
- Dual CO/NOx MEMS sensor
- Fuel cell CO sensor
- FreeScale 3-axis accelerometer
- temperature sensor
- Removed from this version
- USB connector
- Transflash storage
- Next revision
- designed for large enclosure
- several detachable sensor modules
- includes digital temperature/humidity sensor
Mechanical design
- Bluetooth chosen over RS232 or USB to avoid repeated mechanical integration, dongle
- Will fit in the battery well
- with the battery too!
- Also doing a large enclosure design
- stationary or vehicle deployments
Phone platform
- Two current models
- Nokia N95
- Expensive, feature rich
- Symbian OS
- Top of the line
- Includes GPS
- slow, outdoor fixes only
- program with Python!
- LG VX9800
- BREW
- Very, very good AGPS
- fast, indoor fixes
- ~15dB better than unassisted GPS!!!
- Only possible with CDMA phones (this type of AGPS)
Current research topics
- Atmospheric modeling (Demo)
- Super-sampling
- Automatic calibration
- Obstruction and context inference
- Plume detection and avoidance
- Particulate mass sensing
- Social impact of “Community science”
- Awareness of environment through games
Context Inference
Why
- Detect obstructions to the sensor
- Label data with context
- indoors/outdoors
- driving
- riding a bike
- phone on table
- phone in pocket
- Answer questions like
- What is the median exposure to NOx for bicycle commuters on Shattuck ave on tuesdays
How?
- Mic
- use feature extraction (like PCA) on spectrum of ambient noise sample, then classify (SVM?)
- Mic + Speaker
- Use LMS (or other) channel estimation, then classify (SVM?)
- Accelerometer, GPS
- motion and location vectors can also be used, if relevant (ala feature extraction)
- lots of prior work here
- Other sensor inputs (temp, humidity, etc?)
In the pipeline ...
- Outstanding grants or awarded grants for
- Pollution sensing in West Oakland
- Spike in asthma cases when ships are late
- Plume detection and safety guidance
- Environmental games
A pitch
- We have $$$ for a masters student:
- Figure out how to operate the MiCS sensor
- Very interesting signal processing problem
- lots of people would apparently benefit from this (lots of citations)
