Movement Spaces and Movement Traces - Computational Movement Analysis - page 15

cells or Bluetooth beacons (Versichele et al. 2012 ), a discrete space can also be a
deliberate design choice.
Proposing an approach for compressing GPS tracking logs, Richter et al. ( P18 .
2012 ) exploit the fact that urban transit is bound to a transportation network. First,
instead of storing raw and highly redundant GPS logs, movement is abstracted to a
time-stamped passage through a urban transit link (e.g., “along tram line #3” from
stop #402 to #405 from 10:32 to 10:45). Second, following similar concepts intro-
duced in the wayfinding literature, consecutive semantically equal passages are chun-
ked together (e.g., a sequence of visited street edges is chunked to “along Bismar-
ckstrasse” plus additional specifications). Here, the use of the discrete (and also
constrained for that matter), semantically annotated transportation network allowed
for the development of a compression technique for movement data. Similarly, move-
ment events of fish from one river zone into another one are discrete movement events
( P20 . Bleisch et al. 2014 ).
The choice of a certain conceptual movement space can also allow for the adaption
of related methods from neighboring research fields. For example, the abstraction of
movement to a series of visited discrete places allows for the adoption of sequence
and time series analysis. Du Mouza and Rigaux ( 2005 ) propose sequence queries
for trajectories represented as sequences of visited GSM cells. Similar techniques
are used in Dodge et al. ( P14 . 2012 ), however in that study not the movement space
is discrete, but the trajectory is discretized in a segmentation process for similarity
analysis (see Chap. 3 ) .
2.3 Computing Movement Descriptors
Depending on the data capture procedure and the respective conceptual movement
model, raw movement data comes as a stream of location data in the form of
lists of GPS fixes or as time stamped visits to checkpoints. Apart from mapping
movement traces for exploratory analysis, computing descriptive statistics capturing
the essence of the studied movement process is a frequent entry point to CMA.
Many tracking systems produce parameters describing the observed movement,
such as instant speed, accelerometer readings, bearing and signal strength. Such
system-produced data can carry useful information, however, the algorithmic basis
of its computation is all too often unclear or even undocumented by the producer
( P13 . Laube and Purves 2011 ). The research summarized in this topic suggests that
maximal control and hence transparency is achieved when the movement descriptors
underpinning CMA are (re-)computed from raw locational data.
2.3.1 Trajectory Operators
Laube et al. ( P3 . 2007 ) proposed the notion of trajectory operators, then called
lifeline context operators , adopting Tomlin's map algebra for two-dimensional field
data (Tomlin 1990 ) for the case of one-dimensional streams of movement fixes. A set