207The Future of Intelligent Transport Systems. http://dx.doi.org/10.1016/B978-0-12-818281-9.00019-X
Copyright © 2020 Elsevier Inc. All rights reserved.
Chapter 19
Big data analytics for intelligent
transportation systems
19.1 Introduction
One way to perceive data is as a collection of instances that match a certain
schema. The schema describes the attributes and properties (e.g., ranges, types,
etc.) that instances have. Under this perception, “big data” describes data that
are too huge (i.e., they contain large volumes of instances) or too complex (i.e.,
they have many and complex attributes) and require advanced techniques to
process. The sources of big data may vary from sensors and cameras to social
networks and from telecommunication networks to financial transactions. In the
case of transportation systems, big data can be generated by sensor devices and
smart meters that track the moving objects and report their position and status,
from wide-area remote sensing systems that monitor the transportation network
and supervise all vehicles, or even from social networks and web-based sys-
tems. As a result, the decision support system must be able to collect, process,
analyze, and redistribute momentarily, complex data in the size of Terabytes
to Petabytes, which is impossible for a traditional data processing system to
handle.
From a data perspective, the basis for the development of efficient transpor-
tation systems is the aggregation of timely and reliable traffic data (e.g., traffic
load, speed, density, traffic composition, etc.), which can be collected either
by on-site (Leduc, 2008) or by remote sensing techniques (Barmpounakis,
Vlahogianni, & Golias, 2016) and usually have to be transferred to a central
repository for further processing. Depending on the means of transportation,
different techniques are used for data collection, which implies the perception
of raw data, an optional preprocessing and the transmission to a higher level
network as depicted in Fig. 19.1.
19.2 Data collection
19.2.1 In road networks
“On-site” sensing techniques measure traffic data by embedding “detectors” in
the road infrastructure and can be either intrusive or non-intrusive (Zeng, 2015).
The former combines a data recorder and a sensor (e.g., pneumatic road tubes,