Information Technology Reference
In-Depth Information
referencing the Lagrange and Euler dichotomy, Andrienko et al. (
2008
) present a
similar categorization of various ways of observing movement. They discuss
time-
based recordings
(regular interval sampling),
change-based recording
(record made
when position changes),
location-based recording
(when an object passes a beacon,
Eulerian perspective),
event-based recording
(position fix made with phone call),
and combinations of the above.
Records of mobile phone companies provide the most prominent source for
Eulerian movement data, although rigorous and justifiable privacy concerns limit
their availability. Typically, such data is also tied to a constrained movement space
(street network) and discrete. The work by the group around Rein Ahas on mobile
phone usage in the small European country Estonia may serve as a demonstrative
example for this line of work (Ahas et al.
2009
,
2010
; Silm and Ahas
2010
). The
group has access to passive mobile positioning data of a large fraction of the Estonian
population.
Passive
refers to the fact that instead of actively requesting records of
a moving object, here only times, anonymized caller ID, and cell ID with the geo-
graphical coordinates of the antenna are recorded when a user happens to make a
phone call or connect to the internet. Nevertheless, with an average of approximately
six fixed calls per user and day, such Eulerian movement paths can be constructed as
sequences of visited antennas. Exploiting this rather extensive coverage of a small
country's population, the group produces interesting work related to modeling home
and work locations (Ahas et al.
2009
,
2010
) or short-term population mobility (Silm
and Ahas
2010
).
Whereas many Lagrangian/GPS studies are based on experimental set-ups where
tracking devices are distributed and then monitored, many Eulerian studies piggy-
back on existing ICT infrastructure, e.g., people's individual phones and the GSM
(Global System for Mobile Communications) networks maintained by mobile phone
companies. As the Estonian example shows, such secondary exploitation has the
potential for accessing much larger numbers of individuals. For example, Versichele
et al. (
2012
) use proximity-based Bluetooth tracking at a large festival estimating
flow maps of up to 10 % of the festival's 1.5 million visitors. Delafontaine et al.
(
2012
) demonstrate with a similar Bluetooth setup the use of sequence alignment
methods for revealing variability in the visiting patterns at a trade fair, with patterns
formalizing the number and order of visited exhibition halls.
Computing movement descriptors
. Even though the various application fields
study movement with a rather diverse range of motivations, there seems to be a limited
set of parameters characterizing movement. For example, Andrienko et al. (
2008
)
list movement-related characteristics around
position
,
direction
,
speed
,
change of
direction
,
change of speed, acceleration
, and
travel distance
. Similarly, Dodge et al.
(
2008
) list
primitive parameters
(e.g., position
(
x
,
y
)
),
primary derivatives
(e.g.,
speed
f
). Hence,
whereas most authors agree on a basic set of parameters, it is much more ambigu-
ous how these are to be computed, and the implications of such choices when the
parameters serve as the fundamental ingredients of any subsequent analysis. A data
challenge based on GPS tracking of lesser black-backed gulls illustrated in 2011 a
rather impressive variation when several experts in movement analysis were asked
(
x
,
y
,
t
)
), and
secondary derivatives
(e.g., acceleration
f
(
speed
)
