IEEE Circuits and Systems Magazine - Q3 2019 - 43

constant QoS for critical services. A database reliant
system is also particularly necessary in high mobility
situations [34], such as with vehicles. It has been shown
that prediction methods using only short scale channel
state information (CSI) become unreliable very quickly
with the speed of the mobile node as shown in Fig. 4b
and Fig. 4c.
When only local CSI between a BS and a mobile
user is used to predict future performance, the historical channel variations are used to model the current behavior. Figure 4b shows how the actual channel variations can be modeled as a Gaussian Process
with good accuracy for a LTE downlink user moving
at 10 km/h. The number of previously observed CSI
values dictates the accuracy of the regression process
used to predict future values; generally, the longer the
observation history the more reliable the prediction
will be. On the other hand, for high mobility users the
temporal correlation between CSI information is lost
as the channel varies greatly each moment with the
user location. Figure 4c shows the root mean square
error (RMSE) of the CSI prediction for different observation windows for an LTE user moving at 60 km/h.
The figure shows that even though there is an observation window which minimizes the RMSE, the error
remains considerably large for all values of the observation windows. The situation can be improved by using also long-term data, not only short term CSI data.
A distributed approach which considers average channel behavior over different base stations in the path of
the user can be used to provide powerful corrective
measures to local predictions and thus enhance channel prediction. The database concept could be applied
also to other mobile users such as drones [38].
Future Challenges Including Integrated
Satellite-Terrestrial Systems
We identified a number of needed improvements in
Table 2. Implementation of these improvements might
be challenging and research and development work is
needed to achieve the targets. Some ideas for solutions
were described in the previous sections. Base stations
will be developed to support faster frequency changes
due to the evolution in standards and a general need for
flexible and fast resource allocations. Equipment manufacturers have taken part in the research projects aiming to identify the future development needs, e.g. in [8]
and findings are used to develop both SW and HW solutions. An interesting future challenge is to develop and
use prediction methodologies together with licensed
spectrum sharing models taking into account network
traffic and mobility variations. This would enable proactive channel selections and route management.
THIRD QUARTER 2019

There are challenges related to the infrastructure and
costs related to effective spectrum prediction. This is described in [37] as "Setting up a database to enable spectrum sharing between wireless networks is not a trivial
task. It requires technical, economic and political effort.
Since operators may not be willing to share their information between each other and reveal how they are actually
using the precious radio resources, there is a need for involvement of a third party as a spectrum database operator. Selecting that third party and building trust towards
this database operator or operators does not happen in
the blink of an eye. Regulatory authorities test and verify commercial operators and their database solutions."
There are already some LSA and SAS operators fulfilling
the requirements such as Google and Fairspectrum operating in multiple countries. However, this is still a challenge and good business models and cost sharing models
are needed, see e.g. [39].
Future 5G and beyond systems will aim to integrate
satellite and terrestrial systems. The envisioned system
will use a common core network, and partly common
spectrum resources in the operation. This creates interesting challenges from a spectrum sharing point of view
[37], [40], [41]. Similar to the model described in Figure 2,
authors in [40] have described a hierarchical spectrum
management concept for resource management in integrated satellite-terrestrial systems. The concept provides
some starting points for researchers aiming to solve
spectrum sharing problems with different time scales.
One challenge is to build a database-assisted testbed to study spectrum sharing scenarios between 5G
satellite and terrestrial networks taking into account
realistic network traffic dynamics. The testbed should
demonstrate the use of licensed spectrum sharing and
other interference mitigation techniques to provide additional capacity and to guarantee interference protection and QoS for both satellite and terrestrial services.
Mobility and spectrum prediction are enablers for
sharing the spectrum between small satellites in Low
Earth Orbit (LEO) sharing the band with Geostationary satellites using the same band. This is clearly a research problem that needs analytical, simulation, and
practical development work before the system can be
realized. The problem is very timely due to appearance
of several megaconstellation projects such as OneWeb
and SpaceX aiming to launch hundreds of small satellites to the LEO orbit.
Conclusions
This paper explores the use of historical data and predictive approaches for intelligent spectrum use and
shows the potential applications and advantages of the
proposed methods especially from a cellular systems
IEEE CIRCUITS AND SYSTEMS MAGAZINE

IEEE Circuits and Systems Magazine - Q3 2019

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