Accurate, Low-Energy Trajectory Mapping for Mobile Devices
[attachment=16918]

INTRODUCTION
With the proliferation of sensor-equipped smartphones,
the decades-long promise of location-based mobile services
and mobile sensing applications is finally becoming
real. Many location-based applications periodically
probe the device’s position sensor to obtain a stream of
position samples, and then process this stream to obtain
a trajectory. Examples include crowd-sourced traffic
and navigation applications [15, 33], personalized
trip management applications [28, 15], fleet management
applications [21], and mobile object/asset tracking
[11, 34, 7, 19, 25]. The fundamental problem in these
applications is trajectory mapping, where the goal is to
produce the most likely trajectory—a sequence of map
segments—traversed by the mobile device.



WHY CELLULAR?
One of the key motivations for CTrack is that it uses substantially
less energy than GPS. This is to be expected
from a theoretical standpoint because of the difference
in effective radiated power (ERP) for the two systems.
GPS satellites fly in an orbit 11,000 miles above the
earth, with a transmission power of 50 W, resulting in
2×10−11 mW/m2 at the receiver; in contrast, typical cellular
systems register an ERP of up to 10 mW/m2 [14].
This difference of 117 dB translates directly into energy
consumption at the receiver, as the difference must be
compensated by additional processing gain and amplification.
The ERP difference also explains why GPS signals
cannot be acquired without relatively unobstructed
line-of-sight to orbiting satellites, and why they are more
sensitive to weather conditions than GSM signals.

Energy Measurements
We performed a simple experiment to quantify the energy
consumption of each of the sensors of interest —
GPS,WiFi, GSM, the compass and the accelerometer on
an Android G1 phone. For each sensor, we wrote an Android
application to continuously sample the sensor at
some given frequency, as well as continuously query the
battery level indicator. We charged the phone to 100%,
configured the screen to turn off automatically when idle
(the default), and started the application.We used the Android
telephony API to retrieve nearby cell towers and
their associated signal strength values.


Other Energy Studies and Discussion
The numbers above are consistent many previous studies
conducted on a range of phones. For example, we
found [32, 31] that continuously sampling GPS on
iPhone 3G and 4 resulted in 3–10 hours total battery life
(iPhone 3G has lower battery life, and screen brightness
varied in the different papers, resulting in different run
times even without GPS). Leaving the phone on (with
screen on) resulted in 10–18 hours of lifetime (this would
be higher if we could turn the phone’s screen off, but at
the time, non-jailbroken iPhones did not support background
applications.)

Embedded Low-Power Applications
CTrack can also be applied outside the smartphone context
to embedded low-power tagging applications. For
these applications, minimizing cost and battery requirements
is essential. These applications benefit from using
GSM in place of GPS because of increased flexibility of
antenna placement for cellular systems, and resilience to
obstructed environments.