The BikeNet Mobile Sensing System for Cyclist Experience Mapping

1
The BikeNet Mobile Sensing System for Cyclist
Experience Mapping

• Shane B. Eisenman, Emiliano Miluzzo, Nicholas
  D. Lane
• Ron A. Peterson, Gahng-Seop Ahn and Andrew T.
  Campbell
• Dartmouth College, Columbia University

2
Sequence

• The MetroSense Project BikeNet
• The sensing system
• Sensor data!
• Lessons
• Related work
• Wrap up

3
MetroSense

• People-centric Sensing
• Bringing sensor networks into mainstream use by
  the general population
• Sensing systems applied to everyday activities
• BikeNet
• Representative of this class of sensing systems
• Focused on recreational sensing

4
BikeNet

• Recreational Sensing Cyclist Experience Mapping
• 57 million cyclists in the U.S.
• A diversity of requirements

Athletic Training
Fun and Leisure
Means of Transport
5
BikeNet

• Demonstrating the faces of people-centric sensing
  systems

Air Quality
Braking
Cyclist Experience Mapping
Cyclist Community
Sensing power for the people
Coasting
Noise
Distance
Car Density
6
The Sensing System
Physical Bike Area Network (BAN)
7
The Sensing System
Logical Bike Area Network (BAN)
8
The Sensing System
Simplifying the prototype
9
The Sensing System
Hardware Prototypes
10
The Sensing System

• Sampling meaningful sensor data required sensor
  type specific consideration of
• Mounting
• Housing
• Calibration
• Meeting these requirements were as challenging as
  any part of the system.
• Example
• Tilt Sensor

11
The Sensing System

• Example Tilt Sensor (slope of path)

• Used 2-D Accelerometer
• Complicated by
• Noise from bike frame vibration
• Difference in precise orientation angle.
• Bike specific error characteristics demanding
  bike specific calibration
• 3 point calibration process with known stationary
  angles

12
The Sensing System
BANs
Hanover, NH USA
13
The Sensing System
BANs
BANs
14
The Sensing System
Sensor Access Points (SAPs)
15
The Sensing System
Backend Services
16
The Sensing System
Tasking
17
The Sensing System
Sensing
18
The Sensing System
Delivery
19
The Sensing System
Presentation Sharing
20
Sensor Data!

• Data collection began in the summer of 2006
• Participants included members of the sensor lab
  and the general public
• More than 100 kilometers of data collected
• Anonymized traces available soon on Crawdad
  archive

21
Performance Index
22
Performance Index
Distance
Duration
Path Slope
Coasting
Speed
23
Performance InputsSlope and Coasting
24
Health Index
25
Health Index
Noise
Traffic Density
C02 Level
26
Health Input Car Density
27
Health Input C02 Level
28
BikeView Present and Share
29
Public Utility Sensing CO2 Map Hanover NH
30
Lessons

• Mobility and people bring new challenges to
  experimental system development.
• How to debug and perform evaluation?
• Experiments require much more time and effort to
  perform
• Experiments are less predictable with people in
  the loop
• Difficulties exist in finding an experimental
  methodology (i.e., repeatability).

31
Lessons
Debugging on the go!
32
Lessons

• Moving from protocols to caring about the payload
  changes everything!
• Noisy data.
• Vibrations from the bike frame.
• Consider physical solutions (i.e. improving the
  mounting) before attempting post processing
  solutions
• Validating inferences and collected sensor data
  requires time and effort.
• Counting cars by hand with button clicks from a
  bike (tricky and dangerous)
• Manual measurement of road angles
• Ground Truth Helmet

33
Lessons
Sometimes it takes 190 odd kilometers to get it
right
34
Lessons

• Moving from protocols to caring about the payload
  changes everything!
• Noisy data. Vibration in the bike frame.
• Determining appropriate sampling rates.
• Consider physical solutions (i.e. improving the
  mounting) before attempting post processing
  solutions
• Validating inferences and sensor data requires
  time and effort.
• Counting cars by hand with button clicks from a
  bike (tricky and dangerous)
• Manual measurement of road angles
• Ground Truth Helmet

35
Lessons
Ground-Truth Validation Helmet
36
Related Projects

• Existing Cyclist Systems
• Stovepipe commercial solutions
• Body Area Networks and Personal Area Networks
• SATIRE, MIThrill
• DTNs, Mobile Sensing Systems
• Haggle, Cartel, ZebraNet
• People-Centric Sensing
• MIT Media Labs, UCLA, UIUC, Nokia Research, Intel
  Research, Microsoft Research, Motorola

37
Wrap Up

• BikeNet
• Platform for experimentation with mobile sensing
  systems supporting
• Personal Sensing
• Sharing sensor data within Social Networks
• Public Utility Sensing

38

• Cheers for listening
• http//bikenet.cs.dartmouth.edu

Sponsors