Movement Spaces and Movement Traces - Computational Movement Analysis - page 23

to quantify the measurement error for estimates of turning angle and step length as
a function of distance between consecutive locations. By contrast, in the database
community, imperfect tracking data has for some time been the driving force for a
significant research strand around the handling and querying of uncertain positional
data in moving object databases (MOD, for example, Trajcevski et al. 2004 ).
2.5 Concluding Remarks
This chapter has summarized contributions from a series of articles underpinning
the methodological fundamentals of computational movement analysis—conceptual
modeling and abstraction, and representation and description of movement spaces as
well as the moving entities embedded therein. It was shown that from a growing diver-
sity of technologies allowing for the tracking of individuals emerges a wide range of
different forms of movement data, adhering to an equally diverse range of conceptual
models and data structures. Movement of individuals is captured from GPS, WiFi,
Bluetooth, cell phone logs, ticketing and intelligent public transit cards, as well as
gantry stations. Most research analyzing movement is very problem driven, even
data driven, and often the infrastructure setting dictates how the world is abstracted,
and hence the conceptual data models are chosen. In other words, the way data is
collected often rules the abstraction of the world, which then has implications on the
analysis process. However, since there is little comparison between methods, there is
little insight about the implications of such crucial design choices. CMA studies how
the diversity of how we perceive and model movement has implications on how we
describe and quantify movement, and hence how we progress in the CMA process
enriching movement data to process knowledge, as will be studied in the following
chapter on movement mining.
Computational movement analysis contributes to the theory of GIScience by
adapting and adopting core concepts of spatio-temporal modeling and analysis to
movement data as a relatively new form of geographic information. Work was por-
trayed adapting 2D field operators to 1D streams of location fixes, both based on a
deeply geographic notion of spatial respectively temporal dependence or neighbor-
hood. Other studies borrowed from the methodological toolbox of geomorphometry
and performed multi-scale analysis addressing sampling issues and Monte Carlo
simulation investigating uncertainty.
Movement data often inherits its conceptual model and sampling regime from a
given tracking system or research design from the application scientist collecting the
data in the first place. However, I argue that computational movement analysis can
and must move beyond accepting such preliminaries as unchangeable constraints
and rather consider them as design choices and systematically study the implications
of such design choices. Furthermore, as became evident, for example, in the work on
semantic trajectory compression, the characteristics of conceptual movement spaces
can at the same time be a limitation but also an opportunity. This will become even
more evident in the following two chapters.