Database Reference
In-Depth Information
In this chapter, after discussing the principal causes of uncertainty in mobility
data, we address trajectory uncertainty and discuss two models for its represen-
tation: the cylinder and the space-time prisms model. We also address trajectory
uncertainty for movement constrained to road networks. In this context, we show
how the space-time prisms model can be used to address the map-matching prob-
lem introduced in Chapter
2
. Finally, we also discuss how uncertainty can be
accounted for in trajectory clustering.
5.2 Causes of Uncertainty in Mobility Data
Appropriate accounting for uncertainty requires being aware of its sources, both
in data collection and data processing. This identification is crucial to decide
if uncertainty should be accounted for in a given situation and how to manage
it. Therefore, before moving on to the representation of uncertainty, we briefly
discuss its main causes, distinguishing the uncertainty in the movement data
per se from that introduced by postprocessing or deliberate accuracy/specificity
reduction. Further, we analyze the observational error introduced by the main
trajectory-tracking techniques suitable for mobility data.
Uncertainty in Localization
The uncertainty introduced when measuring moving object positions depends
both on the technique adopted and on the context in which it is applied, as
we detail later in this section. Regardless of the specific method used to track
object positions, we can identify two kinds of sources of uncertainty: (1) those
related to the nonspecificity of the acquired position, and (2) those related
to the inaccuracy in the position measurement process. A presence sensor, for
example, reveals the identity of objects that are within its range. Thus, by design,
the spatial extent containing an object is known but the actual position of such
object is unknown; therefore, the resulting position is affected by nonspecificity.
On the other hand, the results of GPS position and time measures are precise, but
affected by context-dependant stochastic errors, making them inaccurate. Note
that for some position-tracking technology, both aspects may coexist. Consider a
wireless communication equipment (GSM, WiFi, RFID, Bluetooth, etc.) used to
detect when objects enter its range. In this case the position of the spotted object
is a vague region, due to the possibly mutating environment. For example, some
kind of obstacle may be on the line of sight of the receiving antenna, hindering
the communications and thus potentially causing the object to be out of the range
of the equipment.
Uncertainty Due to Intentional Accuracy Degradation
A measured position, by itself imprecise and inaccurate to some extent, can be
further degraded either at collection time or later, before subsequent processing
or disclosure. This apparently surprising choice is usually determined by privacy
