Aerospace and Electronic Systems Magazine May 2018 - 44

Feature Article:

DOI. No. 10.1109/MAES.2018.170108

Enabling RF Data Analytics Services and Applications
via Cloudification
Todor Cooklev, Wireless Technology Center, Purdue University, Fort Wayne, IN, USA
Vladimir Poulkov, Ilia Iliev, Technical University of Sofia, Bulgaria
Daniel Bennett, Adaptive RF, Fort Wayne, IN, USA
Krasimir Tonchev, Technical University of Sofia, Bulgaria

INTRODUCTION
Data analytics is related to extracting information from observations, measurements, or experiments about a phenomenon or subject of interest for different purposes, such as interpretation of the
data, decision making, diagnosis, and prediction [1]. A classification of the entire field of big data analytics is difficult to make
and can be based on a variety of criteria. One classification related
to the depth of the analysis is presented by Blackett [2], identifying three levels: descriptive, predictive, and prescriptive analytics.
Descriptive analytics uses past data to analyze what has occurred,
while predictive analytics focuses on forecasting future trends and
determining probabilities of occurrence mainly through the application of different statistical techniques or data mining algorithms
for pattern extraction. Prescriptive analytics facilitates decision
making and optimization of complex systems.
Cloudification and big data techniques have brought different
applications of data analytics, such as structured data analysis, text
analytics (email, documents), web content (text, image, video) and
web usage (browser logs, transactions) analytics, multimedia analytics, network analytics, and mobile analytics. [1]. These technologies
are based on the processing of huge amounts of data generated by
many sensors, instruments, or web applications. Although the application domains are different, some commonality among the methods
has emerged: data visualization, statistical analysis, data mining, and
deep learning. Data visualization is intended to communicate the information clearly and effectively through graphical means. Statistical analysis is based on probability theory and statistics and allows

Authors' current addresses: T. Cooklev, Wireless Technology
Center, Purdue University Fort Wayne, 2101 E Coliseum Blvd.,
Fort Wayne IN 46805, USA, E-mail: (tcooklev@purdue.edu).
V. Poulkov, I. Iliev, K. Tonchev, Faculty of Telecommunications,
Technical University of Sofia, 8, Kliment Ohridsky Blvd., 1756
Sofia, Bulgaria. D. Bennet, Adaptive RF, 4617 Innovation Blvd.,
Fort Wayne, IN 46818, USA.
Manuscript received May 31, 2017, revised November 6, 2017,
and ready for publication December 27, 2017.
Review handled by L. Ligthart.
0885/8985/18/$26.00 © 2018 IEEE
44

drawing inferences from large collections of data. Data mining is
related to discovering patterns in large data sets, applying algorithms
for the classification, clustering, regression, statistical learning, association analysis, neural networks, and machine learning (ML).
Deep learning essentially is the implementation of ML methods on
the basis of learning representations for analysis of data [3], [4], [5].
The cloudification of radio frequency (RF) applications is another emerging trend [6]-[11]. The technical challenges of this
task include the acquisition, storage, and processing of RF or
baseband signals. Current technology can begin to answer these
challenges. However, although terms such as "spectrum cloud,"
"wireless cloud," and "cloud radio access network" (C-RAN) have
appeared, their meaning and associated architectures have been
different. There are two main approaches in the use of these terms.
The first approach is in the context of cognitive radio and dynamic
spectrum access. Wireless clouds are being proposed as heterogeneous radio access networks that support multiple air interfaces in
which the mobile terminals can autonomously decide on the most
appropriate air interface and configuration [6]. Cloud-based architectures have also been suggested as a bridge between spectrummonitoring devices and spectrum-using devices for the upcoming
Internet of spectrum devices [7]. The implementation of such an
approach requires a number of new interface standards, some of
which are being developed by IEEE P1900. The second main approach is the cloudification of the radio architecture: the resources
for signal processing and higher-layer protocols are pooled in a
cloud separately from the RF front-end [8]-[11].
In this article, we briefly describe our collaborative cloud-based
platform enabling the introduction of RF data analytics services. The
platform allows the collection of RF data, sharing of the data, algorithms, and services. As a first application, we use the platform for
long-term spectrum monitoring. We describe possible services and
applications by using the collected data from the monitoring related
to the different types of data analytics descriptive, predictive, and prescriptive analytics. As an example of the outcome and advancement
of the research on the collected long-term spectrum monitoring data,
we illustrate how by the use of different RF data analytics approaches, sources of interference and causes of signal-to-interference-plusnoise ratio (SINR) degradation could be identified and determined.
Some societal aspects of RF data analytics are also discussed.

IEEE A&E SYSTEMS MAGAZINE

MAY - JUNE 2018

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine May 2018