Calling all cars:
cell phone
networks and the
future of traffic
Presentation by Scott Corey
Article written by Haomiao
Huang
The Future of Cars
Self-driving
cars?
Boosting the brainpower of the
environment cars drive in
Traffic monitoring has been revolutionized
An intelligent highway
Reducing
the effect of traffic jams and
accidents
Traffic control schemes to react to real
time data
Aid in planning for the future
Sensors
Monitor
traffic
Parking availability
Air pollution
Have traditionally been static sensors
Inductive Loop Detectors
Traffic Cameras
RFID tags
Problems
Expensive
to deploy, operate, repair
Placed only at key locations
Mobile
sensors are a necessity
Mobile Phones
Equipped
with GPS and Internet access
Smartphones enable more widespread
source of data
Worldwide,
there are more cell phones in
use than toothbrushes
Mobile Millennium
One
of the first large-scale phone-based
traffic monitoring projects in the US
Run by Nokia, NAVTEQ, and UC Berkeley
Gathering data, but privately
User
privacy is key for user acceptance
Two main needs:
Preventing the path of a vehicle to be
reconstructed
Separating the identification of the phone
from the data
Anonymity
Data
from phones is tagged with user
information
The data packet is encrypted at
transmission
Proxy server cannot decrypt packet, but
can strip identifying information
Sent to traffic servers after information
stripped
Reconstructing paths
Uses
virtual trip lines instead of constant
reporting
VTL spacing varies based on speed to
maximize number of cars
Randomizing measurements
Making sense of it all
UC Berkeley tasked to fuse all the data
together
GPS from phones
GPS data from dedicated vehicles
Static sensors
Given all of the measurements being
gathered and a stretch of road of interest,
what is the best estimate of the number of
cars on that road, and how fast they're
going?
Combining data with maps
GPS
tracks are useless alone – need to
combine with maps to know what road
network you are monitoring
Measurements have to use machinelearning methods to correct for people
walking with phones, parked cars
The flow of traffic
Tracking
thousands of cars individually is
difficult and expensive
Traffic researchers treat movement of cars
as liquid flowing through tubes
Fluid Dynamics
Requires
initial conditions and rate of cars
entering/leaving roadway
Fluid dynamics model works well with
fixed sensors
Cameras can determine initial conditions
Sensors attached to on and off ramps
Disruptions
Drivers
are not perfect
Accidents
Unnecessary slow-downs
Adding
GPS dramatically increases the
versatility of the fluid model
GPS incorporated as internal conditions
for the flow to satisfy
Mobile Century
Proof
of concept test
100 cars with mobile phones mixed into
traffic
Ran for 10 hours with 150 student drivers
Despite accounting for 2-5% of cars on
the highway, speed and density of cars
measured at a high resolution
Accident was detected and reported in
less than a minute
Till all are one
Concepts
and technology are now
widespread
Mobile sensors used to identify potholes in
roads
Connections to vehicle sensors
Mobile sensing is the future