Wireless big data: transforming heterogeneous networks to smart networks 21
is given. In section 3, we define WBD and give possi-
ble solutions to handle it based on its characteristics.
In section 4, the framework of the smart networks is
given in detail. The open problems of designing the
smart networks will be discussed in section 5. Fi-
nally, we conclude the paper in section 6.
next generation
cognitive radio
SDN/NFV
smart networks
wireless big data
Figure 1 Enabling technologies of the smart networks
2 Brief overview of enabling tech-
niques for smart networks
In this section, the enabling technologies for smart
networks are briefly reviewed, including CR, NFV,
and data-driven AI methods.
2.1 CR
CR was first proposed in Ref. [5], and became pop-
ular after FCC (Federal Communications Commis-
sion) recommended it as a promising solution for
supporting dynamic spectrum access
[6]
. Since then,
a lot of researches have been carried out
[7-9]
. Here,
we focus on two CR paradigms related to HetNet:
OSA (Opportunistic Spectrum Access) and CSA
(Concurrent Spectrum Access). In the OSA model,
a spectrum band can be reused by secondary systems
only if it is idle. In the CSA model, the spectrum
band is shared by primary system and secondary sys-
tem while secondary system has to obey certain con-
straints for protecting the primary system.
The cognitive cycle of CR consists of spectrum
awareness, analysis, decision, and spectrum exploita-
tion. The loop will be repeated until the system
fully reaches the optimal state. Spectrum awareness
is powered by spectrum sensing
[10]
, which aims to
tell whether the specific spectrum is idle. Analysis
and decision focus on the solution of the most effec-
tive resource allocation
[11]
. Although CR has been
investigated for over a decade, practical imperfec-
tion is unavoidable
[12]
. The imperfection mainly lies
in the channel/noise uncertainty, signal uncertainty,
noise/channel correlation and transceiver design. To
cope with the problem brought by uncertainty, some
robust schemes, such as eigenvalue based spectrum
sensing, have been proposed, and a good survey is
given by Sharma et al.
[12]
.
Reconfiguration of operational parameters, such
as waveform, transmitting power, modulation, car-
rier frequency, bandwidth, coding, etc. is the en-
abler for CR. Self-adaptive reconfiguration of oper-
ational parameters are implemented so as to reuse
the available spectral opportunities
[2]
. In Ref. [13],
SDAI (Software Defined Air Interface) is proposed
as a framework of 5G air interface. In SDAI, mod-
ules and related parameters are adapted in real time
to flexibly fit all sorts of applications for 5G, and to
realize agility and efficiency.
2.2 NFV
Wireless networks are supported by RAN (Radio Ac-
cess Network) and CN (Core Network), where the
RAN is in charge of user access and related ser-
vice for each base station while the CN controls
the interconnection among base stations and Inter-
net. In traditional wireless networks, physical equip-
ments such as routers, base stations and switches
are coupled heavily with the software running on
them. Such coupling makes it difficult to upgrade
services through updating evolving software or net-
work protocols. To solve this problem, SDN (Soft-
ware Defined Network) which utilizes virtualization
technology to abstract away the softwares originally
running on physical devices is proposed to offer a
new paradigm to manage networking services
[3]
. The
main purpose of SDN is to decouple the physical
network forwarding devices and the CPFs (Control
Plane Functions)
[14]
.
NFV has been promoted and advocated by