Mobility and Uncertainty - Mobility Data - page 84

relationships between entities in the three domains. For example, both the uncer-
tainty due to the finite representation of coordinates and the one due to unknown
positions fall into class (1), since they are caused respectively by the uncertainty
in the computer representation and in the human cognition (lack of knowledge/
memory) of entities.
The second branch (i.e., the uncertainty due to human, machine, and geog-
raphy/movement relationships) can be refined according to the kind of differ-
ence existing between the corresponding entities in the different domains. In
particular we can distinguish: inaccuracy/error , a deviation of a measurement
from the reality; incompleteness , caused by a partial description of the reality;
inconsistency , indicating the existence of different computational and cognitive
statements referring to the same entity (e.g., because of semantic mismatch
or contradiction, or simply due to different representations); and imprecision ,
which refers to a lack of exactness of computational or cognitive values. We can
further classify imprecision depending on its degree in: nonspecificity , meaning
that only a set containing the true value is known; ambiguity , when it is not
possible to define univocally a set containing the exact value; and vagueness ,
when it is not possible to define a set containing the exact value, because true
or false are just two of the possible truth values. We refer to fuzziness when the
truth of a value is replaced by a continuously changing degree of truth. In both
cases, no sharp/crisp boundary separates true and false values.
Uncertainty in Mobility Data
Using different position collection techniques entails different kinds of uncer-
tainty affecting recorded data. Some of the tracking methods described in Chap-
ter 2 have irrelevant errors on position and time measurement for most applica-
tion scenarios, whereas other ones are intrinsically less precise. In other cases,
the position is not measured, for example when it is manually inserted during
a data entry process. In this case the position could be inaccurate, because of
digitization errors, or vague, due to the nature of entities involved. For example,
a valley is a vague concept and it is hard to devise crisp borders that have separate
interior and exterior points. As a consequence, it is not possible to select the
trajectories that stopped inside a valley in an exact way. Similarly, due to the lack
of crisp borders of a zone frequently subject to avalanches, it is difficult to deter-
mine the number of skiers at risk even if we know exactly all of their trajectories.
Mobility data are characterized by several dimensions. In particular, in addi-
tion to space and time, data related tomovement semantics and user actions could
also be present. Each of these dimensions is potentially affected by one of the
above kinds of uncertainty. For example, the semantic annotation and segmen-
tation of trajectories could be affected by uncertainty in the spatial dimension.
Thus, in case the geometry of a place of interest (POI) is fuzzy, or the positions
of the objects are inaccurate, it could be difficult to assert that an object stopped
at a POI.