What is Cognitive Radio?
In this section, we define cognitive radio and investigate the algorithms and types of technologies
that already exist.
9.3.1 Definitions of Cognitive Radio
As any newly emerging technology, the definition of “cognitive radio” can be seen in many different
ways. In fact, the term Cognitive Radio means different things to different audiences. The
earlier definition by Joseph Mitola in his dissertation titled “Cognitive Radio – An Integrated Agent
Architecture for Software Defined Radio” , was given as follows. The cognitive radio identifies
the point at which wireless PDAs and the related networks are sufficiently computationally
intelligent on the subject of radio resources and related computer-to-computer communications to (a)
detect user communications needs as a function of use context, and (b) to provide radio resources
and wireless services most appropriate to those needs. Cognitive radio increases the awareness that
computational entities in radios have of their locations, users, networks, and the larger environment.
Mitola included the concept of machine learning as a property of cognitive radio. Mitola’s
definition on cognitive radio includes a high level of awareness and autonomy, in a sense that cognition
tasks, that might be performed, range in difficulty from the goal driven choice of RF band,
air interface, or protocol to higher-level tasks of planning, learning, and evolving new upper layer
The FCC gave the following definition on cognitive radio . A cognitive radio is a radio that
can change its transmitter parameters based on interaction with the environment in which it operates.
At the same time, it should also note that FCC refers to a SDR as a transmitter in which the operating
parameters . . . can be altered by making a change in software that controls the operation of the device
without . . . changes in the hardware components that affect the radio frequency emissions. It went on
to claim that the majority of cognitive radios will probably be SDRs, but neither having software nor
being field reprogrammable are requirements of a cognitive radio.
To summarize from the aforementioned two versions of definitions on cognitive radio, we can
see that Mitola emphasized the level of device/network intelligence which adapts to user activity;
while the FCC seems primarily concerned with a regulatory friendly view, focused on transmitter
behavior at the moment. Therefore, the relationship between the cognitive radio and SDR from
the views of Mitola and the FCC can be seen in Figure 9.3, where cognitive radio adapts to the
spectrum environment; while SDR adapts to the network environment. They partially overlap in their
9.3.2 Basic Cognitive Algorithms
It is therefore not difficult to discern that a fully functional cognitive radio should have the ability to
do the following works: (1) Tune to any available channel in the target band. (2) Establish network
communications and operate in all or part of the channel. (3) Implement channel sharing and power control protocols which adapt to spectra occupied by multiple heterogeneous networks. (4) Implement
adaptive transmission bandwidths, data rates, and error correction schemes to obtain the best throughput
possible. (5) Implement adaptive antenna steering to focus transmitter power in the direction
required to optimize received signal strength.
The core of a cognitive radio is its inherent intelligence, which makes it different from any normal
wireless terminal available today, in either 2- or 3G systems. This intelligence will allow a cognitive
radio to scan all possible frequency spectra before it makes an intelligent decision on how and when
to make use of a particular sector of the spectrum for communications. Therefore, it is inevitable
that a cognitive radio needs great signal processing power to deal with the vast amounts of data it
captures from various radio channels. Thus, the capability to process all those enormous amounts of
data on a real-time or quasi-real-time basis is a must for any cognitive radio.
It is still too early to specify exactly the algorithms that a cognitive radio should use at the moment
of writing this book. However, we would like to provide some evidence as to how a primitive cognitive
radio may behave. Obviously, any cognitive radio has to use the following two protocols for its very
basic operation: (1) DFS, and (2) TPC.
The DFS was originally used to describe a technique to avoid radar signals by 802.11a networks
which operate in the 5 GHz U-NII band. Now, it has been generalized to refer to an automatic
frequency selection process intended to achieve some specific objective (like avoiding harmful interference
to a radio system with a higher regulatory priority). On the other hand, TPC was originally
a mechanism for 802.11a networks to lower aggregate transmit power by 3 dB from the maximum
regulatory limit to protect Earth Exploration Satellite Systems (EESS) operations. Now it has been
generalized to a mechanism that adaptively sets transmit power based on the spectrum or regulatory
environment. These two protocols will become a must for all cognitive radios.
In addition, a cognitive radio should have IPD capability , which is another key cognitive
radio behavior. The IPD is the ability to detect an incumbent user (one with regulatory priority)
based on a specific spectrum signature. The operation of IPD bears the following characteristics: (1)
DFS requires an IPD protocol to identify unoccupied, or lightly used frequencies. (2) IPD includes
detection schemes focused on the characteristics of the specific incumbents in the band, or bands, that
the cognitive radio is designed to support. (3) IPD eliminates the need for geo-location techniques
(GPS, etc.) to determine the location of the radio and, using a database, identifies unused channels.
As both TPC and IPD algorithms are intuitive, as suggested by its name, we will only explain
the implementation of the DFS cognitive algorithms in depth, in the following text.
The DFS algorithm was originally proposed in the ITU-R recommendation M.1461  to avoid
possible interference to existing radar operations in the vicinity. Many radar systems and unlicensed
devices operating co-channels in proximity could produce a scenario where mutual interference is
experienced. The DFS methodology is used to compute the received interference power levels at
the radar and unlicensed device receivers. A DFS algorithm may provide a means of mitigating this
interference by causing the unlicensed devices to migrate to another channel once a radar system
has been detected on the currently active channel. This model first considers the interference caused
by the radar to the unlicensed device at the output of the unlicensed device antenna. If the received
interference power level at the output of the unlicensed device antenna exceeds the DFS detection
threshold, the unlicensed device will cease transmissions and move to another channel. The algorithm
then computes the aggregate interference to the radar from the remaining unlicensed devices. Each of
the technical parameters used in the method and the radar interference criteria will also be described.
The received signal level from the radar at the output of the unlicensed device antenna can be
evaluated by using the following equation:
IU = PRadar + GRadar + GU − LRadar − LU − LP − LL − FDR (9.1)
where IU is the received interference power at the output of the unlicensed device antenna in dBm,
PRadar is the peak power of the radar in dBm, GRadar is the antenna gain of the radar in the direction of the unlicensed device in dBi, GU is the antenna gain of the unlicensed device in the direction of
the radar in dBi, LRadar is the radar transmit insertion loss in dB, LU is the unlicensed device receive
insertion loss in dB, LP is the propagation loss in dB, LL is the building and nonspecific terrain
losses in dB, and FDR is the frequency dependent rejection in dB.
Equation (9.1) is calculated for each unlicensed device in the distribution. The value obtained is
then compared to the DFS detection threshold under investigation. Any unlicensed device for which
the threshold has been exceeded will begin to move to another channel, and consequently is not
considered in the calculation of interference to the radar, as given by
IRADAR = PU + GU + GRadar − LU − LRadar − LP − LL − FDR (9.2)
where IRADAR is the received interference power at the input of the radar receiver in dBm, PU is
the power of the unlicensed device in dBm, GU is the antenna gain of the unlicensed device in the
direction of the radar in dBi, GRadar is the antenna gain of the radar in the direction of the unlicensed
device in dBi, LU is the unlicensed device transmit insertion loss in dB, LRadar is the radar receive
insertion loss in dB, LP is the radio-wave propagation loss in dB, LL is the building and nonspecific
terrain losses in dB, and FDR is the frequency dependent rejection in dB.
With the help of equation (9.2), we can calculate each unlicensed device being considered in the
analysis that has not detected energy from the radar in excess of the DFS detection threshold. These
values are then used in the calculation of the aggregate interference to the radar by the unlicensed
devices using the following equation:
where IAGG is the aggregate interference to the radar from the unlicensed devices in Watts, N is the
number of unlicensed devices remaining in the simulation, and IRADAR is the interference caused to
the radar from an individual unlicensed device in Watts.
It is necessary to convert the interference power calculated in Equation (9.2) from dBm to Watts
before calculating the aggregate interference seen by the radar using Equation (9.3).
The parameters used in the above DFS algorithm can be explained as follows: To obtain “radar
antenna gain” (GRadar), we need to know the azimuth and elevation antenna pattern models for the
radar considered. The models should provide the antenna gain as a function of an off-axis angle
for a given main beam antenna gain. The unlicensed device power level (PU ) in this analysis is
assumed to be 38 dBm and 6.6 dBm. The building and nonspecific terrain losses (LL) include building
blockage, terrain features, and multipath. In the above analysis, this loss has been treated as a
uniformly distributed random variable between 1 and 10 dB for each radar unlicensed device path.
When determining Radar and Unlicensed Device Transmit and Receive Insertion Losses (LRadar and
LU ), we have assumed that the analysis includes a nominal 2 dB for the insertion losses between
the transmitter and receiver antenna and the transmitter and receiver inputs for the radar and the
unlicensed device. Finally, to compute the radio-wave propagation loss (LP ), the NTIA Institute
for Telecommunication Sciences Irregular Terrain Model (ITM) was used . The ITM model
computes radio-wave propagation based on the electromagnetic theory and on the statistical analysis
of both terrain features and radio measurements to predict the median attenuation as a function of
distance and variability of the signal in time and space.
9.3.3 Conceptual Classifications of Cognitive Radios
The characteristic features of a cognitive radio have a lot to do with the spectrum facts in different
regions or countries. If we are only looking at the US market, we will see that a lot of spectra have
been assigned for licensed use by the FCC. Actual spectrum use varies dramatically from region to region: spectrum is more congested in urban areas, and hardly used in rural areas. Some licensed
services only operate in a few locations nationally (for example, Fixed Satellite Services). Even in
urban areas, only a fraction of available spectra is in continuous use. We have to admit that, in terms
of reclaiming fallow spectrum, a lot of low hanging fruit is available for harvest using cognitive
techniques. Regulatory activity is just beginning to open up opportunities to reclaim lightly used
spectra for new services.
Currently, there are two conceptual forms of cognitive radios. One is called full cognitive radio,
in which every possible parameter observed by the wireless node and/or the network is taken into
account while making a decision on the transmission and/or reception parameter change. The other is
called Spectrum Sensing Cognitive Radio, which is a special case of Full Cognitive Radio in which
only the RF spectrum is observed.
Also, depending on the parts of the spectrum available for cognitive radio, we can distinguish
“Licensed Band Cognitive Radio” and “Unlicensed Band Cognitive Radio.” When a cognitive radio
is capable of using bands assigned to licensed users, apart from the utilization of unlicensed bands
such as the U-NII band or the ISM band, it is called a Licensed Band Cognitive Radio. One of the
Licensed Band Cognitive Radio-like systems is the IEEE 802.15 Task group 2  specification.
On the other hand, if a cognitive radio can only utilize the unlicensed parts of a RF spectrum, it is
an Unlicensed Band Cognitive Radio. An example of an Unlicensed Band Cognitive Radio is IEEE
Although cognitive radio was initially thought of as an SDR extension (Full Cognitive Radio),
most of the current research work is focused on Spectrum Sensing Cognitive Radio, particularly on
the utilization of TV bands for communication. The essential problem of Spectrum Sensing Cognitive
Radio is the design of high-quality spectrum sensing devices and algorithms for exchanging spectrum
sensing data between different nodes in a cognitive radio network. It has been shown in  that
a simple energy detector cannot guarantee the accurate detection of signal presence. This calls for
more sophisticated spectrum sensing techniques and requires that information about spectrum sensing
must be regularly exchanged between nodes. In , the authors showed that the increasing number
of cooperating sensing nodes decreases the probability of false detection. To adaptively fill free RF
bands, OFDM seems to be a perfect candidate. Indeed in  T. A. Weiss and F. K. Jondral
from the University of Karlsruhe, Germany, proposed a Spectrum Pooling system in which free
bands sensed by nodes were immediately filled by OFDM subbands. Some of the applications of
Spectrum Sensing Cognitive Radio include emergency networks and WLAN higher throughput, and
transmission distance extensions.