IEEE Circuits and Systems Magazine - Q3 2019 - 36

may happen both in licensed and unlicensed bands, as
shown in Fig. 1, there is even a higher need for monitoring,
predicting, finding, and dynamically allocating the spectrum for all applications and services [6].
Classification and prediction can enhance recently
proposed DSA systems such as the licensed shared access (LSA) [7] and the citizens broadband radio service
(CBRS) that uses spectrum access system (SAS) [8] database concepts. Both systems are based on spectrum
databases, which provide a natural way to store historical
data over different time scales and consequently make
decisions over short and long time periods. By using
specific classification and learning techniques, the operation of secondary licensed users under both LSA and
SAS concepts as well as unlicensed general authorized
access (GAA) users of a SAS could be allocated to the
best available bands. In addition, used channels can be
evacuated proactively before appearance of the incumbent users. We have implemented both LSA and SAS systems in live LTE networks and used that information in
proposing practical improvements based on spectrum
and mobility prediction.
This article provides a short state-of-the-art survey of
the latest developments in spectrum prediction techniques
and applications, focusing specifically on database-assisted spectrum sharing concepts in cellular communications. There are some open challenges identified both
in our practical studies and in the recent survey papers.
Authors in [5] conclude that current network infrastructure should be enhanced with repositories for context

information and application profiles to assist realization
of novel predictive applications. Repositories are a natural part of LSA and CBRS networks. However, according
to [7], the current LSA concept is too static, enhancements and more dynamic operations are needed. We
extend the state-of-the art by defining how the current
LSA and CBRS concepts should be modified to realize
more dynamic approaches with the use of predictive
techniques. A predictive local-aware spectrum allocation with mobility management is used as a practical
example to show how more dynamic database-assisted
cellular networks could be realized.
This article is organized as follows. We first present
key applications of predictive methods and discuss techniques and metrics especially from the predictive channel selection point of view including a look to a recent
deep learning area. Computational costs of different algorithms are given. Different ways to improve database-assisted spectrum sharing in cellular systems are discussed
as open research challenges. Then, we study local-aware
resource allocation, giving some numerical results before
the conclusions.

IEEE CIRCUITS AND SYSTEMS MAGAZINE

Speed of Operations

Intelligence

Spectrum Prediction Applications for 5G Use Cases
Database-assisted operation and spectrum prediction
techniques can support all the main 5G use cases: 1)
Enhanced mobile broadband (eMBB) requires large
bandwidths. Database-assisted techniques can add supplemental frequency resources to the licensed band operation that is essential for eMBB. Also, other methods
such as license-assisted access
(LAA) will be used to enhance capacity. 2) Massive machine type
communication (mMTC) requires
Long-Term Memory
large coverage and good signal
High Intelligence, Low Speed
Long Term Trends and Predictions,
penetration through walls and
Large Time and Spatial
Wide Geographic Area
Scales e.g., Network
Monthly/weekly/daily Patterns
other obstacles, which mean acManagement System,
"Best Channel on Friday 3 p.m."
cess to frequency spectrum below
Spectrum DB Concepts
1 GHz. Predictive techniques can
assist in finding the most suitable
Short Term Spectrum
Short-Term Memory, Medium
time slots for low data rate transInformation Concerning
Intelligence
missions. 3) Ultra-reliable low laCurrent Data Transmission
e.g., a Base Station, Mobile
tency communications (URLLC)
"Time
Scale
of
a
Few
Edge Computing (MEC),
Seconds or Minutes"
require the use of licensed specSpectrum Databases
trum. Joint mobility and spectrum
prediction can be used, e.g., in
assisting the base station selecNo Memory, Low Intelligence,
High Speed
Instantaneous Spectrum
tion, minimizing the number of
Low Time and Spatial Scales
Information Locally
handovers, and improving the quale.g., Adaptive Sensing Based
Fast Adaptations
ity of vehicle-to-everything (V2X)
Device-to-Device (D2D)
communications.
The achievable intelligence levFigure 2. Hierarchy for intelligent spectrum use.
el in a communication network
THIRD QUARTER 2019

IEEE Circuits and Systems Magazine - Q3 2019

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q3 2019