Trajectories and Their Representations - Mobility Data - page 10

recorded positions: the information on the movement of the object is missing. If
this is accidental (e.g., because of a device malfunction) we say there is a hole
in the track. The typical case where this still happens is when a GPS is taken
through a tunnel. The connection is cut as long as the GPS doesn't get out of
the tunnel. Short-duration holes may sometimes be “filled,” using, for example,
linear interpolation algorithms that compute the missing positions. In this case
the hole disappears.
If missing points are not due to some data acquisition accident (whatever
the cause), it follows that their absence is due to a decision by the application
designer to interrupt data acquisition during some specific periods. For example,
a company running daily tourist tours in Paris may decide to track tourists'
positions during its hours of operation (say from 8 A.M. to 6 P.M.) but not during
lunchtime (say from 12:30 P.M. to 2 P.M.) when tourists on a tour are free to
do whatever they want. Consequently, tourists' daily trajectory tracks will be
filled with positions from 8 A.M. to 12:30 P.M. and from 2 P.M. to 6 P.M., and no
positions during the lunchtime break. This lunchtime break is not an accidental
hole in the trajectory; we call it a semantic gap (its semantic in this case is that
of the lunch period).
A trajectory with semantic gaps is defined for a set of disjoint time intervals
instead of a unique time interval. For the sake of simplicity, in the rest of the
chapter we will deal only with trajectories defined on a single interval (i.e.,
without semantic gaps).
1.3 From Raw Trajectories to Semantic Trajectories
The two representations of trajectories defined above come directly from the
movement track. It is why they are often called raw trajectories . They are
well fitted if, for example, the aim of the application reduces to locating some
moving objects (e.g., where was Mr. Smith on the evening of June 12, 2012?) or
computing statistics on the spatio-temporal characteristics of the trajectory (e.g.,
which percentage of daily tourist trajectories show a global speed over 7 km/h?).
On the other hand, many applications need more informative results, such as
those that can be computed by combining raw data with the contextual data
(e.g., geo-objects and events that show a spatial or temporal relationship with
the trajectory data), and with the thematic data available for the moving object
itself (e.g., age, gender). These applications can reach this goal by following
one of two approaches:
1. The application dynamically accesses the contextual data during its compu-
tations.
2. The application first preprocesses the trajectory representations, enriching
them with contextual data and appropriate restructurings, and after that it
computes its results by using the enriched trajectories.