Page 1
2 Cognitive Radio overview 16
2. Cognitive Radio Overview
2.1. Background and history of Cognitive Radio
Spectrum is a scarce commodity, and considering the spectrum scarcity faced by the
wireless-based service providers led to high congestion levels. The main reason that
leads to inefficient utilization of the radio spectrum is the spectrum licensing system
itself. If the allocated radio spectrum is not used by licensed users, it cannot be
utilized by unlicensed users [M. M. Buddhikot]. Due to this static and rigid allocation,
wireless systems have to work only on a dedicated band of spectrum, and cannot
change the transmission band as changing the environment. For example, if one
channel of spectrum band is heavily used, the wireless system cannot change to work
on another more lightly used band.
The authorized access of the spectrum is usually defined by owner of spectrum
(i.e. licensee), transmit power, frequency, space, type of use, and the license duration.
In general, a license is allocated to one licensee, and the use of spectrum by this
owner must have the specification e.g. maximum power of transmit, base station
location. In the present spectrum-licensing system, the license cannot change the
application or giving the right to another licensee. This restriction causes in low
utilization of the frequency spectrum. A spectrum hole is a band of frequencies
assigned to a licensee, but, at a particular time and particular geographic location, that
user is not utilizing the band [S. Haykin paper].
Page 2
2 Cognitive Radio overview 17
The distribution of radio spectrum is under control of the central government;
the Federal Communications Commission (FCC) published a report in November
2002, prepared by the Spectrum-Policy Task Force, aimed at improving this precious
resource. The allocation of the unlicensed frequency bands has resulted in the
congestion of these bands. The most of the usable frequency spectrum already has
been assigned for licensed user, resulting in a scarcity of spectrum for new and
up-and-coming wireless applications. To resolve this crisis, regulators and policy
makers are working on new spectrum management strategies. Particularly, the U.S.
Federal Communications Commission (FCC) is tackling the problem in three ways
[Federal Communications Commission, Docket no. 02-135]: spectrum reallocation,
spectrum leases, and spectrum sharing. In spectrum reallocation, bandwidth from
government and other long-standing users is reassigned to new wireless-base services
such as mobile communication, broadband Internet and video distribution. In
spectrum leases, the FCC relaxes the technical and business limitations on existing
spectrum licenses by permitting existing licensees to use their spectrum flexibly for
various services or even lease their spectrum to third parties. Spectrum sharing has
allocation of an unmatched amount of spectrum that could be used for unlicensed or
shared service.
This thesis focuses on spectrum allocation in order to improve the efficiency
of spectrum usage. The FCC is considering a new spectrum-sharing pattern, where
licensed bands are opened to unlicensed operations on a non-interference basis.
Because some licensed bands (such as TV bands) are under-utilized, spectrum sharing
in empty sections of these licensed bands can fill the spectrum shortage problem. This
spectrum-sharing model frequently referred to as dynamic spectrum access (DSA).
Licensed users are referred to primary users (PU), whereas unlicensed users that
Page 3
2 Cognitive Radio overview 18
access spectrum opportunistically are referred as secondary users (SU). The spectrum
utilization can be improved notably by making it possible for a SU to access a
spectrum hole that unoccupied by the PU. The spectrum holes have been utilized to
promote the efficient use of the spectrum by taking advantage of the existence of
spectrum holes. This concept is known as Cognitive Radio coined by Joe Mitola,
after introducing the note of software radio in 1991, together with Gerald Maguire,
used the term cognitive radio (CR) for the first time in 1999. They exposed a CR as an
improvement of a software radio and state: “Radio protocol is the set of RF bands, air
interfaces, protocols, and spatial and sequential patterns that moderate the use of radio
spectrum. Cognitive radio extends the software radio with radio-domain model-based
reasoning about such protocols.” Mitola introduces radio cognition cycle, explicitly
elaborated in [J. Mitola’s dissertation] as shown in Figure. 2.1.
Figure 2.1 Cognitive cycle .
2.2. Software-defined radios (SDRs)
A software-defined radio (SDR) is a reconfigurable radio in which the transmission
parameters such as modulation mode, frequency band, and protocol may be adapted
dynamically. This adaptability function is obtained by software-controlled signal
Page 4
2 Cognitive Radio overview 19
processing algorithms. This additional ability has the radio being assembled with a
radio frequency front end, a down converter to an intermediate frequency (IF) or base
band processing, an A/D converter, and then a processor shown in Figure 2.2 as
general structure of SDR transceiver. The capacity of processing limits the complexity
of the signal that can be holds.
Figure 2.2 SDR transceiver
SDR is a main part to implementing CR. The major functions of SDR are as follows [F. K.
Jondral, pp. 275–283,]:
Multiband function: SDR will carry wireless data transmissions over a
different frequency spectrum used by different wireless access systems (e.g.
cellular band, ISM band, TV band).
Multiple standard: SDR will support various standards such as GSM, CDMA,
WiFi, WiMAX.
Multiple services: SDR will be able to work on various types of services such
as cellular telephony or broadband wireless Internet access.
Multi-channel access: SDR will be able to work on multiple frequency bands
simultaneously.
As shown in Figure 2.2, the transmission parameters in a SDR transceiver may be
reconfigured according to the communication requirements and specifications: The
radio parameters such as standard to be operated and frequency band can be set before
Page 5
2 Cognitive Radio overview 20
the device is delivered to the user. However, the parameters cannot be changed when
the device is configured once. Even though dynamic reconfiguration of the device is
not supported in SDR. The parameters can be seldom reconfigured, when the network
infrastructure modified or added on a connection basis. For example, when a user
wants to start a wireless Internet connection, the transceiver parameters can choose
from the availability of different wireless access networks such as GSM, WiFi or
WiMAX, based on performance and cost. The radio parameters can also be
dynamically changed on a basis of time-slot. For example, the transmission power can
be changed when the level of interference varied. The unlicensed user (SU) can
change the operating band of frequency when the activity of the licensed user (PU) is
detected. This SDR can supports from 800MHz to 5 GHz frequency spectrums.
2.3. Cognitive Radio characteristics and capabilities
A CR includes both a sensing and an adaptation component to the software-defined
radios, provides methods for intelligent spectrum sensing, spectrum management, and
access of spectrum for cognitive radio users (SU). A suitable description is found in
[Hykin paper]: “Cognitive radio is an intelligent wireless communication system that
is aware of its surrounding environment i.e. its outside world, and uses the
methodology of understanding by building to learn from the environment and adapt
its internal states to statistical variations in the incoming radio frequency (RF) stimuli
by making corresponding changes in certain operating parameters e.g. transmit power,
carrier frequency, and modulation strategy in real time. In other words, a CR is an
extended Software defined radio (SDR) that additionally senses its environment,
tracks changes, and possibly reacts upon its findings. A CR network facilitates to set
Page 6
2 Cognitive Radio overview 21
up communications among CR users/nodes. The wireless communication parameters
can be attuned according to change in the environment, topology, operating or
requirements of user. The two key objectives of CR are: (1) to attain highly reliable
and highly capable wireless communications, and (2) to improve the frequency
spectrum utilization.
2.3.1. Cognitive Radio architecture
The CR protocol architecture is shown in Figure 2.3. In the SDR transceiver are
implemented at physical layer as a RF front-end. The adaptive protocols in the
application, transport, network, and MAC, layers have to be aware of the variations in
the CR environment. In particular, the traffic activity of primary users, the
transmission requirements of SU and changes in channel feature, etc has to consider
by the adaptive protocols. To connect all components, a CR control is used to set up
interfaces with the adaptive protocols, SDR transceiver, and wireless use and services.
This CR component uses intelligent algorithms to process the calculated signal from
the physical layer, and accept the requested information on transmission requirements
from the SU to control the protocol parameters in the different layers [Dynamic
Spectrum Access and Management in Cognitive Radio Networks, EKRAM HOSSAIN].
Physical layer includes carrier frequency, duty cycle, transmit power, digital
modulation mode, processing gain, spectrum bandwidth, channel coding rate and
type, and waveform of the transmitted signal.
MAC layer includes packet size, packet type, channel/time slot allocation, data
rate, retransmission probability, scheduling scheme, and MAC protocol.
Page 7
2 Cognitive Radio overview 22
A network and transport layer includes network scheduling algorithm in the
routing layer, routing metric and parameters of congestion control (TCP window size)
and rate control parameter (token bucket size rate control) in the transport layer.
Lastly, Application layer includes encryption algorithm and source coding.
Figure 2.3 CR protocol stacks.
2.3.2. Functions of Cognitive Radio
The key functions of CR, to carry efficient and intelligent dynamic spectrum access,
are as follows [I.F. Akyildiz et al.]:
Spectrum sensing: The aim of spectrum sensing is to observe the status of the
spectrum and the movement of the licensed or primary user (PU) by
periodically sensing its frequency band. In this case, a CR transceiver senses a
spectrum hole or idle spectrum (i.e. time, location and band) and also observe
the way of accessing it (i.e. access duration and transmit power) without
interfering with the communication of a PU. Spectrum sensing is two types;
centralized and distributed. In case of centralized spectrum sensing, a base
Page 8
2 Cognitive Radio overview 23
station or access point act as sensing controller that senses the target frequency
band, and the information as a result obtained is shared with other nodes in the
cognitive radio network. The complexity of user terminals can reduce by
centralized spectrum sensing, since all the sensing functions are carrying out at
the sensing controller. However, centralized spectrum sensing having problem
from location multiplicity. For example, the sensing controller may not be able
to sense a secondary user (SU) at the periphery of the cell. In case of
distributed spectrum sharing, SU carry out spectrum sensing separately, and
the spectrum-sensing outcome can be either used by individual CRs (known as
non-cooperative sensing) or shared with other SUs (as a cooperative sensing).
Even though cooperative sensing acquires a communication and operating
cost, the accuracy of spectrum sensing is superior to that of non-cooperative
sensing.
Spectrum analysis: The SUs use the information obtained from spectrum
sensing to schedule and plan spectrum access. In this case, the transmission
requirements of SUs have to be optimizing the transmission parameters. Main
mechanisms of spectrum management are spectrum analysis and spectrum
access optimization. In spectrum analysis, information from spectrum sensing
is analyzed to get knowledge concerning the spectrum holes such as duration
of availability, interference estimation, and collision probability with a PU due
to sensing error. Then, a decision to access the bandwidth, frequency, transmit
power, modulation mode, time duration and location of spectrum is made by
optimizing the system performance given the preferred objective i.e. maximize
the throughput of the SUs (this thesis focused on these) and maintain the
interference caused to PUs lower the target threshold.
Page 9
2 Cognitive Radio overview 24
Spectrum access: Before a decision is made on spectrum access based on
spectrum analysis, the unlicensed users access the spectrum holes. Spectrum
access is carries out based on a cognitive medium access control (MAC)
protocol, which proposes to avoid conflict with PUs and also with other SUs.
The CR transmitter is also required to make compromise with the CR receiver
to synchronize the transmission in order that the transmitted data received
successfully. A cognitive MAC protocol either is based on a fixed allocation
MAC (such as FDMA, TDMA, CDMA) or a random access MAC (such as
ALOHA, CSMA/CA) [Dynamic Spectrum Access and Management in Cognitive
Radio, EKRAM HOSSAIN].
Spectrum mobility: Spectrum mobility is a task related to the change of
operating frequency band of SUs. When a PU starts accessing a radio channel,
which is currently being used by an SU, the SU switched to another idle
spectrum band. This change in channel of operating frequency band is referred
to as spectrum handoff. In spectrum handoff, the different layers in the
protocol stacks, the protocol parameters have to be adjusted to match the new
channel of operating frequency band. Spectrum handoff must attempt to make
certain that the communication by the SUs can carry on in the new spectrum
band. This observable fact is most appealing in this thesis.
2.3.4. Dynamic spectrum access (DSA)
SUs exploits the implementation of cognitive radio will be based on DSA. Dynamic
spectrum access can be defined [Hykin’s paper] as a method to fine-tune the spectrum
resource handling in a real time approach in response to the changing objective and
Page 10
2 Cognitive Radio overview 25
environment (e.g. type of applications and available channel), changes of radio state
(e.g. battery condition, transmission mode, and location), and changes in external
constraints and environment (e.g. operational policy and radio propagation). There are
three main models of DSA, namely, general-use, shared-use, and private-use models.
In the general-use model, the spectrum is open for access to all users. This model is
already been using in the ISM band. In the shared-use model, licensed users (PU) are
allocated the frequency bands, which are opportunistically accessed by the unlicensed
users (SU) when the PU does not occupy them. In the private-use model, a PU agrees
to access of a particular frequency band to a SU for a certain time. This model is more
flexible than the spectrum-licensing model related with traditional command-and-
control, because the type of use and the spectrum licensee is able to changed
dynamically. In opportunistic spectrum access SU can exploit idle in-band sections
without causing interference to the active PUs. Spectrum overlay and spectrum
underlay are the two approaches for opportunistic spectrum access. The spectrum
overlay approach (or opportunistic spectrum access) does not essentially force any
strict constraint on the transmission power by SU. It allows SU to identify and utilize
the spectrum holes defined in space, time and frequency. This approach is well
matched with the existing spectrum allocation thus the PU systems can continue to
work without being affected by the SUs. The spectrum underlay approach limits the
transmission power of SU so that they work below the interference temperature limit
of PU. One possible approach is to transmit the signals in an ultra wide frequency
band (UWB) transmission in order that a high data rate is achieve with very low
transmission power. It is the worst-case hypothesis that the PU transmits all the time.
Thus, probably it does not utilize spectrum holes. However, second approach is out of
scope of this thesis. In the general-use model, dynamic sharing can be between
Page 11
2 Cognitive Radio overview 26
homogenous networks (IEEE 802.11a uses in the 5 GHz band) or between
heterogeneous networks (coexistence between IEEE 802.11b and 802.15.1 Bluetooth
networks). When all the networks in a heterogeneous environment have cognitive or
adaptive capabilities, it is referred to as symmetric sharing. On the other hand, when
there is one or more network without cognitive/adaptive capabilities (e.g. coexistence
of legacy technology with cognitive radio technology, coexistence of powerful 802.11
networks with low-power 802.15.4 networks) is referred to as asymmetric spectrum
sharing. DSA is divided into two major parts, i.e. spectrum investigation (sensing and
analysis) and spectrum utilization (decide and handoff). Different design techniques
of CR can be used in these parts.
2.3.5. Cognitive Radio Components
The major components of cognitive radio, which functions and adapt the transmission
parameters according to the varying environment [J. Mitola and , T. Christian James
Rieser, Dissertation]. The different components in a CR that apply these functionalities
are shown in Figure 2.4.
Figure 2.4 Components in a node of cognitive radio.
Page 12
2 Cognitive Radio overview 27
Transmitter and receiver: This SDR-based wireless transceiver is the main
component by means of the functions of signal transmission and reception. In
addition, a wireless receiver is also sense the frequency spectrum used (observe
the activity on the frequency spectrum). The parameters of transceiver in the CR
can be dynamically changed as directed by the upper layer protocols.
Spectrum analyzer: The spectrum analyzer uses measured signals to analyze the
spectrum usage. This is to detect the signal signature from a PU and to locate
spectrum holes for SU to access. The spectrum analyzer should take care that the
transmission of a PU is not interfered with if an SU decides to access the
spectrum. In this case, different signal-processing methods can be used to obtain
spectrum-handling information.
Knowledge extraction/learning: Learning and knowledge extraction use the
information on spectrum usage to understand the behavior of RF environment of
PU. A knowledge base of the spectrum access environment is built and preserve,
which is consequently used to optimize and adjust the transmission parameters to
get the desired objective under different restrictions. Machine learning algorithms
taken from the area of artificial intelligence (AI) can be utilized for learning and
knowledge extraction.
Decision-making: After the knowledge of the spectrum usage is available, the
decision on accessing the spectrum has to be made. There are various techniques
used to achieve a best solution. For example, optimization technique such as GA
and PSO can be applied when the system can be modeled as a single entity with a
multiple objective. Stochastic optimization may be applied when the system states
are random. Some of examples has been given in chapter 4.
Page 13
2 Cognitive Radio overview 28
2.3.6. Spectrum sensing
The goal of spectrum sensing is to identify the existence of transmissions from PU.
There are three types of spectrum sensing, i.e. non-cooperative sensing, cooperative
sensing and interference-based sensing shown in Figure 2.5. The survey on various
types of spectrum sensing is explored in [Yahya Rahmat-Samii, Dynamic Spectrum
Access and Management in Cognitive Radio Networks by EKRAM HOSSAIN]. Non-
cooperative spectrum sensing is used by a SU to detect the transmitted signal from a
PU by using local observations and local measurements.
Energy detection is the best method for spectrum sensing when the
information from a PU is unavailable [A. Sahai et al.]. In the method of energy
detection, the output signal from a band pass filter is squared and integrated over the
inspection interval. .A decision algorithm compares the integrator output with a
threshold level [J. G. Proakis, Digital Communications] to decide whether a PU exists
or not. In general, the energy detection performance degrades when the SNR
decreases.
Figure 2.5 various types of spectrum sensing in the CR physical layer
The limitations of energy detection are: susceptible to the uncertainty of noise
power and it can only detect the presence of the PU signal but unable differentiate the
Page 14
2 Cognitive Radio overview 29
type of signal such signals from SUs sharing the same channel with the PU. Thus, the
error of detection would be high in presence of signal sources other than the PU.
Matched filter detection is commonly used to identify a signal by comparing a
reference signal (or pattern) with the input signal. A matched filter will maximize the
received SNR for the calculated signal. Thus, if the signal information (e.g. packet
format or modulation) from a PU is known a matched filter is an optimal detector in
stationary Gaussian noise [A. Sahai et al]. Since a reference signal is used for signal
detection, a matched filter needs only a small amount of time to operate. However, if
this reference signal does not exist or is erroneous, the spectrum sensing performance
degrades significantly. Matched filter detection is suitable when the transmission of a
PU has pilot, preamble, synchronization word or spreading codes, which can be used
to construct the pattern for spectrum sensing.
In cyclostationary feature detection, when the transmitted signal from a PU
usually has a periodic pattern. This periodic pattern is views as cyclostationarity, and
can be used to detect the presence of a PU [Dynamic Spectrum Access and Management
in Cognitive Radio, EKRAM HOSSAIN]. A signal is cyclostationary if the
autocorrelation is a periodic function. The transmitted signal from a PU can be
distinguished from noise with periodic pattern, which is a wide-sense stationary signal
without correlation. In general, cyclostationary detection can provide a more accurate
sensing result and it is robust to variations in noise power. However, the detection is
complex and requires long observation periods to obtain the sensing result. A pattern
recognition scheme based on a artificial intelligence can be used to implement
cyclostationary feature detection for spectrum sensing.
Page 15
2 Cognitive Radio overview 30
In cooperative sensing, an SU transmitter may not always be able to sense the
signal from a PU transmitter due to its channel fading and geographic separation. For
example, as shown in Figure 2.6, the transmitter and receiver of the SU cannot sense
the signal from the transmitter of the PU because they are out-of-range. This is view
as the hidden node problem. While the transmitter of the SU transmits, it will interfere
with the receiver of the PU. To overcome the hidden node problem in non-cooperative
transmitter sensing, cooperative spectrum sensing may be used. In cooperative
sensing, information of spectrum sensing from multiple SUs are exchanged with each
other to sense the presence of PUs. The cooperative spectrum sensing architecture can
be either centralized or distributed [G. Ganesan and Y. G. Li, pp. 137–143.]. Using
cooperative exchange of spectrum sensing information, the hidden node problem can
be solved and the sensing probability can be appreciably enhanced in a heavily
shadowed environment.
Figure 2.6 Hidden node problem
On the other hand, this acquires a greater communication and computation operating
cost compared with non-cooperative sensing. The scheme of interference-based
sensing has been given by FCC. In this case, the sensing algorithm will measure the
noise and interference level from all sources of signals at the receiver of the PU. This
Page 16
2 Cognitive Radio overview 31
information is used by an SU to control the spectrum access by computing expected
interference level without violating the interference temperature limit. Alternatively,
an SU transmitter may observe the feedback signal from a SU receiver to gain
knowledge on the interference level. However, spectrum-sensing topic is out of scope
of this research work.
2.3.7. Spectrum analysis and spectrum decision
Spectrum analysis is required for the description of different spectrum bands in terms
of PU activity, operating frequency, interference, bandwidth, and channel capacity. In
the spectrum underlay approach, the interference temperature limit at the PU receiver
and operating frequency, the allowable transmission power at the CR can be
determined. Consequently, the capacity of channel can be estimated. Spectrum
analysis models can be based on either present spectrum sensing (real-time) results or
past spectrum results. The architecture of spectrum analysis can be either non-
cooperative or cooperative. A cooperative architecture, which can be either distributed
or centralized, can get better the accuracy of the spectrum usage model. A cooperative
architecture requires exchanging information between CRs (pay additional operating
cost) and could suffer from the problem of scalability. This is out of scope of the
thesis and assumption made as the result of spectrum analysis has been obtained.
The spectrum decision deals, how to utilize the spectrum holes, in other
words, what power level and modulation to use, how to allocate the spectrum holes
with CRs. This is mainly a medium access control (MAC) problem for a CR. In
addition, spectrum access decisions may require to be communicated between the CR
Page 17
2 Cognitive Radio overview 32
nodes and the intended receivers. Spectrum decisions may be made based on either a
local or a global optimization principle. In case of local optimization, the spectrum
access decision is made in a way of non-cooperative. In the case of global
optimization, a cooperative spectrum access decision is made either in a distributed or
a centralized way. In a non-cooperative spectrum access approach each CR node is
responsible for its own decision. If the miss-detection probability is large, the access
policy subjected to conservative. If the false alarm probability is large, the access
policy subjected to aggressive. Therefore, the access approach can be jointly
optimized with the sensing approach. A non-cooperative spectrum access strategy has
minimal communication requirements (lower operating cost), but it can result in poor
spectrum exploitation.
In a cooperative centralized approach, a centralized server maintains a
database of availability of spectrum and access information that depends on the
information received from a group of SUs through a dedicated control channel. Thus,
spectrum management is simpler which enables efficient spectrum sharing. In both
the centralized and the distributed approaches, the PU may or may not cooperate.
Once a decision is made to access the spectrum opportunities, various issues related to
radio link control and resource management have to be determined. These consist of
pulse shaping, choice of the number of spectrum bands to access, transmission power
control and the set of suitable bands, adaptive modulation and coding etc.
Page 18
2 Cognitive Radio overview 33
2.3.8. Applications of Cognitive Radio
CR concepts can be applied to different types of wireless communications, a few of
which are explained below:
Next generation wireless networks: CR is expected to be a core technology for
next generation heterogeneous wireless networks. This will provide intelligence to
both the user-side and service-side equipments to manage the network and air
interface efficiently. At the user-side, mobile equipment with multiple air
interfaces (e.g. cellular, WiFi and WiMAX) can observe the condition of the
wireless access networks (e.g. transmission quality, delay, throughput, and
congestion) and make a decision on selecting the access network to connect
according to cost. At the service-side, radio resource from multiple networks can
be optimized for the given set of mobile users and their QoS requirements. Based
on the mobility and traffic outline of the users, efficient load balancing systems
can be implemented at the infrastructure of service provider to allocate the traffic
load among multiple available networks to reduce network congestion [I.F.
Akyildiz at al.].
Coexistence of different wireless technologies: New wireless technologies (e.g.
IEEE 802.22-based WRANs [33]) are being developed to reuse the radio spectrum
allocated to other wireless services such as TV service. CR is a solution to provide
coexistence between these different technologies and wireless services. IEEE
802.22 based WRAN users can opportunistically use the TV band when a TV
station is not broadcasting. Spectrum sensing and spectrum management will be
Page 19
2 Cognitive Radio overview 34
critical mechanism for IEEE 802.22 standard WRAN technology to avoid
interference to TV users and to maximize throughput for the WRAN users.
eHealth services: Different types of wireless technologies are implemented in
healthcare services to improve efficiency of the healthcare and patient care
management. However, using wireless communication devices in healthcare
application are limited by electromagnetic compatibility (EMC) and
electromagnetic interference (EMI) requirements. Since the bio-signal sensors and
medical equipments are sensitive to EMI, the transmit power of the wireless
devices has to be cautiously controlled. Also, different biomedical devices (e.g.
surgical equipment, monitoring and diagnostic devices) use RF transmission. The
spectrum handling of these devices has to be carefully selected to avoid
interference with each other. The CR concepts can be applied for many wireless
medical sensors that are designed to operate in the ISM (industrial, scientific, and
medical) band, which can implement CR concepts to choose suitable transmission
bands to avoid interference [J. Mitola dissertation].
Intelligent transportation system: Intelligent transportation systems (ITS) widely
use different wireless access technologies to improve the safety and efficiency of
transportation by vehicles. Two different types of communications take place in
ITS system – vehicle-to-vehicle (V2V) communication and vehicle-to-roadside
(V2R) communication. In V2V communications, a special form of ad hoc
network, i.e. a vehicular ad hoc network is formed among vehicles to exchange
safety-related information. In V2R communications, information is exchanged
between the roadside unit and the onboard unit in a vehicle. High speed of the
vehicles and quick variations in network topologies cause significant challenges to
efficient V2V and V2R communications. CR concepts can be used in both
Page 20
2 Cognitive Radio overview 35
onboard units and roadside units so that they can adapt their transmissions to deal
with the rapid variations in the ambient RF environment [J. F. Hauris, pp. 427–
431.]. With multiple-radio abilities at the onboard units, they must be able to
adaptively opt the radio to communicate with the roadside units.
Emergency networks: Emergency and public safety networks can take advantages
of the CR concepts to provide flexible and reliable wireless communication. In the
case of disaster situation, the infrastructure of standard communication may
possibly not exist, and then, a CR network as an emergency network may require
to be established to carry disaster recovery. Such a network may use the CR
concept to make possible wireless transmission and reception over a wide range of
the radio spectrum.
Military networks: With CR, the parameters wireless communication can be
dynamically adapted based on the location and time as well as the mission of the
militaries. Suppose, if some frequencies are noisy or jammed, the CR transceiver
can search for and access another frequency bands for communication.
Furthermore, location-aware CR can manage the transmitted waveform in a
particular area to keep away from interference to the high priority military
communication systems.
2.4. Machine learning
Machine learning is one of the branches in artificial intelligence. This deals with the
design and development of learning algorithms by means of test data or past
occurrence to optimize the performance of a system given definite objectives and
limitations. Machine learning exploits the hypothesis of mathematics and statistics to
Page 21
2 Cognitive Radio overview 36
build inference models from test data so that algorithms may be considered based on
these models. Two main steps in machine learning are training and making decision.
In the training step, test data or past knowledge are used to make knowledge about the
system or the environment. In this step, able algorithms are essential to take out useful
information from unprocessed data. Once the knowledge it makes, a decision is made
based on the available knowledge and present state and input data. Machine learning
techniques can be implemented to solve problems correlated to pattern recognition,
natural language processing, and robotics [Dynamic Spectrum Access and
Management in Cognitive Radio Networks, EKRAM HOSSAIN].
2.5. Cognitive Radio scheme based on Artificial intelligence
Artificial intelligence (AI) techniques designs learning and decision-making process
that can be implemented in knowledgeable CR systems [Dynamic Spectrum Access and
Management in Cognitive Radio Networks, EKRAM HOSSAIN and C. Clancy et al. pp. 47–
52]. One example has been demonstrated in [C. Clancy et al. pp. 47–52], the idea of
machine learning useful to maximization of capacity and DSA for SU. The proposed
system architecture is shown in Figure 2.7. Here, the knowledge base maintains the
conditions of the system and the accessible actions [P. Jackson].
Figure 2.7 Cognitive radio architecture with machine learning.
Page 22
2 Cognitive Radio overview 37
The reasoning engine employs the knowledge base to select the optimal action. The
learning engine carries knowledge exploitation based on the observed information
(e.g. channel error rate and information on channel availability). In the knowledge
base, two data structures i.e. inference rule and action are defined. The inference rule
is used to represent the environmental state. According to this condition, an action can
be executed to change the condition so that the system objectives can be achieved. For
a example, inference rule can be defined as “SNR equal to 5dB and modulation equal
to QPSK ”, whereas the action defined as “decrease the modulation mode” with
condition “SNR less than and equal to 8dB” and post-condition “modulation equal to
BPSK”. The input, which is obtained from measurement, the reasoning engine
matches the existing state (modulation and SNR in this case) with the predicates and
find out the predicate results either true or false. Then, from the predicate results set, a
suitable action is taken. In this example, if the present SNR = = 5 dB and present
modulation = = QPSK, the precondition will be true and the predicate will be active.
As a result, the CR engine will choose to reduce the modulation mode. Here, the
modulation will be changed to BPSK, as declared in the post condition. A learning
algorithm is used to update the condition of the system as well as the available actions
according to the environment of radio. This update can be made using an objective
function (e.g. minimize the BER) with an objective to find out the optimal action
given the input (quality of channel) and the knowledge availability. Various learning
algorithms can be used in a CR network (hidden Markov model, expert system
[P. Jackson], neural network [S. Haykin, Neural Networks], genetic algorithm, particle
swarm optimization, DNA inspired algorithm). The two main components in the
architecture as shown in Figure 2.7 are the action and the reasoning. The action
component is used to observe the input from the environment and the condition of the
Page 23
2 Cognitive Radio overview 38
system, and to take the suitable decisions for the CR. The reasoning part stores the
knowledge and the rules. This is used by the reasoning engine to obtain the optimal
decision according to the defined objective. The full details of the realization of CR
based on the base-10 GA, PSO, DNA inspired algorithm used in this thesis, along
with the all parameter settings are explored in chapter 3,4 and 5.
2.6. Designing Cognitive Radio based on location-aware
Geo-location is an important CR-enabling technology because the large range of
usage that may result from a radio being aware of its present location and probably
being aware of its intended path and target. The global positioning system (GPS) is a
satellite-based system that utilizes the time difference of arrival (TDoA) to locate a
receiver [Cognitive Radio Technology by Bruce Fette]. GPS receivers typically consist of
a one-pulse-per-second signal as it appears at each radio from each source of satellite,
resulting in a computing of propagation delay from each source in spite of position. In
the nonexistence of GPS signals, triangulation method can be used to locate a radio
from non-cooperative or even cooperative emitters. Other approaches are time of
arrival (ToA), angle of arrival (AoA) and Received Signal Strength (RSS) explored in
[Cognitive Radio Technology by Bruce Fette, Wireless communications by Anderea
Goldsmith]. In the case of RSS, if the transmit power on a signal is precisely known,
the patterns of the antenna radiation gains are known accurately, and the receiver is
capable to measure receive signal strength accurately, then a propagation model may
be used to compute the distance to the transmitter and receiver as a function of RSS.
But propagation channels are varying dynamically, thus this approach is challenging.
This location finding approach is analogous to the ToA approach. If a process of
Page 24
2 Cognitive Radio overview 39
correlation based on a PU transmitter’s database, an RSS-based receiver application
can find out in which regulatory area it is located. For example, if a CR is receiving
particular TV channels and particular AM and FM stations all at the same time, it may
conclude its city location. If the location of the transmitters is built-in the database
along with levels of transmission, the RSS process might improve this computation
due to the fairly large number of measurements. The quality of RSS-based location
estimates is somewhat low. It is helpful to CRs for a few applications but not for
others.
There is an alternative method like radio frequency fingerprinting, which is
widely explained in the literature as a technique for identification for transmitter. The
received signal is extremely site specific because of its dependence on the terrine and
intervening obstacles. So the multipath structure of the channel is unique to every
location and can be considered as a fingerprint or signature of the location if same RF
signal is transmitted from location [Wireless communications by Anderea Goldsmith,
O. Leon et al.]. This property has been exploited in system to develop a “signature
database” of a location grid in specific service areas. The received signal is measured
as a CR moves along this grid and recorded in signature database. When another CR
moves in the same area, the signal received from it compared with the entry in the
database, thus its location is determined. Such a scheme may also be useful for indoor
application where the multipath structure in an area can be exploited. This work
exploits this principle for detection of legitimate PU by SUs in order to prevent
adversary attacks (denial of service (Dos)). The full details of the implementation of
the fingerprint method used in this research, along with the all parameters are
presented in chapter 6.
Page 25
2 Cognitive Radio overview 40
2.7. Summary
Due to the instruction-based approach used in traditional spectrum licensing, the radio
spectrum cannot be efficiently exploited. Thus, a new spectrum-licensing idea is
being developed that will improve the flexibility of spectrum access. This flexibility
will be obtained through the use of CR implemented as SDRs. In CR, a wireless
system can change the transmission parameters dynamically according to the
environmental change. A cognitive radio transceiver should have the ability to
observe, orient, plan, decide and act to improve the performance of wireless network.
With this capability, SUs can utilize the frequency spectrum that unused by PU.
However, an SU have to guarantee that the interference caused to the PU due to its
communication remains under the limit of interference temperature. This spectrum
sensing can be doing either in a cooperative or a non-cooperative manner. In the
cooperative spectrum sensing, multiple SUs cooperate by exchanging sensing results
among each other. In case of non-cooperative spectrum sensing, each SU senses the
radio spectrum separately. A number of approaches in designing CR based on AI have
been discussed. Also, several approaches for CR location awareness schemes are
explained.