Cognitive Radio Survey
Cognitive Radio Survey
Cognitive Radios
Spectrum Sensing and Allocation Techniques
Farrukh Javed
7/28/2008
F‐05‐020/07‐UET_PHD‐CASE‐CP‐40, Electrical and Computer Engineering Department
Centre for Advanced Studies in Engineering, Islamabad, Pakistan
Advisor: Dr Riaz Inayat
Abstract
The un‐precedented and exponential expansion in field of telecommunications is yet to face a challenge that
cannot be amicably resolved. But the lightning pace of development itself poses a question which if not effectively
addressed will bring this development to a dead halt. The ultimate human limitation in the field of
telecommunication: “The available spectrum is but finite”. Latest advancements in research have offered various
solutions but definitely none as viable as “Cognitive Radios”. The term next generation Radios Networks or
dynamic spectrum access is also used for the same paradigm. This paper is formulated to fulfill the requirements of
doctoral qualification exam and encompasses, “the introduction to the spectrum sensing and allocation techniques
used in cognitive radios”.
2 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
Contents
Section I – Cognitive Radios
1. Introduction .......................................................................................................................................... 4
2. Next Generation (xG) networks ............................................................................................................ 5
3. Cognitive Radio ..................................................................................................................................... 6
3.1 Cognitive Capability ...................................................................................................................... 7
3.1.1 Spectrum Sensing .................................................................................................................. 7
3.1.2 Spectrum allocation .............................................................................................................. 7
3.2 Re‐Configurability ......................................................................................................................... 8
Section II ‐ Spectrum Sensing
4. Transmitter detection ......................................................................................................................... 10
4.1 Matched filler detection ............................................................................................................. 11
4.1.1 Opportunities ...................................................................................................................... 11
4.1.2 Challenges ........................................................................................................................... 11
4.2 Energy detection ......................................................................................................................... 11
4.2.1 Opportunities ...................................................................................................................... 11
4.2.2 Challenges ........................................................................................................................... 12
4.3 Cyclo‐Stationary feature detection ............................................................................................. 12
4.3.1 Opportunities ...................................................................................................................... 13
4.3.2 Challenges ........................................................................................................................... 13
5. Cooperative detection ........................................................................................................................ 13
5.1 Opportunities .............................................................................................................................. 14
5.2 Challenges ................................................................................................................................... 14
6. Interference based detection ............................................................................................................. 14
6.1 Opportunities .............................................................................................................................. 15
6.2 Challenges ................................................................................................................................... 15
7. Spectrum Sensing Challenges ............................................................................................................. 15
7.1 Interference Temperature .......................................................................................................... 15
7.2 Spectrum Sensing In Multi‐User Networks ................................................................................. 16
7.3 Speed of detection ...................................................................................................................... 16
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 3
Section III ‐ Spectrum Allocation
8. Spectrum Analysis ............................................................................................................................... 17
8.1 Channel capacity ......................................................................................................................... 17
8.2 Primary and xG network user info .............................................................................................. 17
8.3 Channel Capacity ......................................................................................................................... 19
8.4 Spectrum analysis Challenges ..................................................................................................... 19
8.4.1 Opportunities ...................................................................................................................... 20
9. Spectrum Decision .............................................................................................................................. 20
9.1 Spectrum Mgmt: ......................................................................................................................... 21
9.2 Spectrum Mobility....................................................................................................................... 21
9.2.1 Spectrum mobility challenges ............................................................................................. 22
9.2.2 Opportunities for spectrum mobility .................................................................................. 23
9.3 Spectrum Sharing ........................................................................................................................ 24
9.3.1 Architecture based classification ........................................................................................ 24
9.3.2 Challenges and Opportunities ............................................................................................. 24
9.3.3 Spectrum Sharing based on the access behaviour ............................................................. 24
9.3.4 Challenges and Opportunities ............................................................................................. 25
9.3.5 Spectrum sharing based on access technology .................................................................. 25
9.3.6 Challenges and Opportunities ............................................................................................. 25
9.4 Spectrum Sharing Challenges ...................................................................................................... 26
10. Conclusion ....................................................................................................................................... 27
Section I – Cognitive Radios
1. Introduction
The finite nature of available spectrum is undoubtedly the biggest question mark on the
phenomenal expansion in the field of telecommunication. The question that “Are we headed for a dead
stop?” has been answered in negative in very mean ways but none is apparently more suitable than the
idea of cognitive radios. The human limitation of finites is answered by human nature to cognitivity. The
term “Cognitive” as elaborated in encyclopaedia Encarta as the ability of acquiring knowledge by the use
of reasoning, intuition or perception. The concept of cognitive radios is based on humanising the
communication networks by giving them ability to sense, analyse, decide and adjust. Though the barrier
of finite nature of spectrum cannot be crossed but it sure can be bypassed by the most optimum
utilization of available spectrum.
Presently the spectrum is accessed using fixed spectrum access techniques FSA. Government
agencies allocate spectrum bands to different users and vendors based on policies and monetary
agreements. The fixed allocation, due to it’s convenience of use has survived for so many years and is
sure to last for many more. But the biggest drawback of the technique is its non‐flexibility. The mass
variations in spectrum concentration at diff time, space and freq bands results in wastage of major
portion of spectrum at most of the time instances. Federal communication commission (FCC) [1] places
the spectrum usage between the ranges 15% ‐ 85% at all times. This implies that if some means can be
adopted to effectively utilize, unused spectrum slots in different freq bands the spectrum can be used
much more efficiently and economically.
Fig. 1. Spectrum Utilisation [1]
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 5
Fig. 2. Spectrum Concentration [2]
Another estimation of spectrum concentration is given by Jean‐Pierre Hubaux in [2]. He has
observed spectrum allocation and usage at six locations(Locations: New York city; Riverbend Park, Great
Falls, VA; Tysons Corner, VANSF Roof, Arlington, VA; NRAO, Greenbank, WV; SSC Roof, Vienna, VA). He
has come up with the results shown in fig. 2. He goes on to comment that in Europe, cellular operators
have spent nearly 100 billion Euros to buy spectrum for the 3rd generation.
2. Next Generation (xG) networks
The xG Networks are heterogeneous networks which provide dynamic spectrum access by using
cognitive radios as nodal points in the network. The xG Networks can be used in parallel to licensed user
networks and can opportunistically utilize the under‐utilised spectrum without disturbing the licensed
user network.
The inherent flexibility and adaptability in the xG networks necessitates elaborate protocols that
can address a multitude of situation that are un‐encountered in fixed spectrum access scenarios. The
situation is further complicated by the fact that the xG network will almost work not only in parallel to
the licensed user network but also subservient in the regards that it will always be the xG network that
has to adept to the changing needs of the licensed user band. There might also be situations in which
various xG networks are to work in conjunction with each other.
Though the protocols defining xG networks are still a popular area for study and have not
evolved into a very refined form however the generally suggested architecture are of multiple cross
6 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
Fig. 3. xG Network Functionalities [3]
layers. Akyildiz in [7] suggested a cross‐layer architecture which is shown in fig. 3. The architecture
exhibits the cooperative and complex functionalities of xG Network.
3. Cognitive Radio
Joseph Mitolla and Gerald Maguire [4] in their ground breaking work described the novel idea as
the situation in which wireless nodes and the related networks are sufficiently computationally
intelligent about radio resources and related computer to computer communication to detect the user
communication needs as a function of use context and to provide resources and wireless resources most
required. In simpler words, a cognitive radio is a radio that can interact with its radio environments by
sensing its parameters and adapting to them. The definition indicates two basis characteristics of
cognitive radios.
• Cognitive capability: Interaction with environment in order to detect the spectrum
parameters. The spectrum needs to be analysed for spectrum concentration, power level,
extent and nature of temporal and spatial variations, modulation scheme and existence of any
other xG network operating in the neighbourhood.
• Reconfigurability: The beauty of cognitive radio is its flexibility. The radio is capable to
adopt itself so as to meet the spectrum needs in the most optional method. This flexibility in
design has been rendered only recently due to the advancements in concept of software radios
DSP techniques, antenna technology etc.
The paper is restricted to the spectrum sensing and allocation of cognitive radios which constitute
the cognitive capability of a radio. Hence the re‐configurability is only briefly discussed and then the
spectrum sensing and allocation techniques and concepts have been discussed in detail.
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 7
3.1 Cognitive Capability
Cognitive capability of a radio enables it to interact with its radio environment. The interaction is
a closed loop function called as cognitive cycle. The spectrum is analysed for its various parameters and
decisions made on the basis of this analysis are used for reconfiguration of cognitive radio. This is a
continual process and the spectrum sensing and analysis is not ceased at any instant. The cycle is
depicted as in fig. 4 in [3]
The implementation of this cycle is still an open area for research and multiple options have
been suggested. However, a broad categorization can be made as under:
• Spectrum Sensing
• Spectrum allocation
o Spectrum analysis
o Spectrum decision
3.1.1 Spectrum Sensing
Spectrum sensing implies the data collection from the radio neighbourhood not only for the
identification of temporally unused slots in the spectrum but also for all other relevant details. These
unused slots are called as spectrum holes or white space [5]. The monitoring is continuous and over the
complete spectrum.
Fig. 4. Cognitive cycle [3]
3.1.2 Spectrum allocation
Spectrum allocation is made based not only on the sensed spectrum parameters but various
other functions. The spectrum allocation follows two steps.
[Link] Spectrum analysis
The analysis of spectrum parameters and identifying the spectrum holes is carried out in order
to determine various parameters that are required to choose a suitable band for transmission.
8 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
[Link] Spectrum decision
The decision about the selection of a suitable band for transmission is not only depending on
the spectrum information acquired by spectrum analysis but also by various other consideration equally
important such as:‐
• Transmission characteristics
• Spectrum management
• Spectrum mobility
• Spectrum sharing
3.2 Re‐Configurability
After selection of a suitable band and transmission parameters for transmission the next biggest
design challenge is to reconfigure or adapt the radio to suit these parameters. The recent technological
breakthroughs in software radios can be best utilised for giving this flexibility. There are various
parameters that can be considered for reconfiguration. Incorporation of each factor adds to the design
flexibility.
• Operating frequency: This is the ability to transmit at different operating frequencies in order
to use a spectrum hole. This is the primal requirement for a dynamic spectrum access network.
• Modulation Scheme: A cognitive radio should reconfigure for adaptation to the channel and
user requirements. For, example for a delay sensitive transmission the modulation scheme
should be chosen whose delay characteristic are better than the error rate. Conversely, a low
sensitive transmission would be more suited to a low error rate channel. Another example can
be the use of CDMA for security sensitive transmissions.
• Transmission Power: The cognitive nature of a cognitive radio enables it to use a transmission
most suited to the requirement. The transmission power level can be adjusted based on the
information of the intended receiver. The adjustment of transmission power to minimum gives
two big advantages. Firstly, it enables reduced power consumption and secondly it enhances the
numbers of users linked to the network by decreasing interference.
• Communication Technology: The inherent design flexibility enables a cognitive radio to be
interoperable between different communication systems.
• Directivity of transmission: The directivity of transmission gives another dimension of expansion
to cognitive radios. The spectrum holes cannot only be in temporal sense but also spatial
distribution of holes can be considered and put to optimum utilisation.
Farrukh Ja
aved F‐05‐020. Spectrum Sensing and Alloca
ation in Cognittive Radios (DQ
QE ‐ II) 9
Power
Frequency
Time
Fig. 5. SSpectrum Hole C
Concept
10 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
Section II – Spectrum Sensing
As already introduced, the cognitive nature of a cognitive radio is manifested by its interaction
with the spectrum and the very first step in this regard is the identification of spectrum holes. Actual
detection of a channel between the transmitter and receiver is very difficult. However, the recent work
utilises the interactive nature of a cognitive radio by detecting the transmitter based observation of
spectrum users. The broad classification of cognitive radio sensing is as follows
Spectrum Sensing
Matched Filter Cyclostationary
Energy Detection
Detection Feature Detection
4. Transmitter detection
Spectrum concentration in a spatial domain can be determined by receiving transmission from
all the perceived transmitters in that domain at any time instant. The approach depends on observation
of signals received at various cognitive radios in an xG network and analysing the context for the
transmission by the primary transmission in that vicinity the hypothesis mode would be
,
,
Or
,
,
Where x(t) is the signal received at the cognitive radio, n(t)is the AWGN noise present in the
channel and h(t)is the impulse response of the channel. The second model is the frequency domain
translation of the same hypothetical model. Both hypotheses are checked for correctness based on
probabilistic models and if hypothesis H1 is found true then the presence of transmitted signal is
concluded and further analyzed.
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 11
As indicated already, the techniques used for detection can be broadly categorized as matched
filter detection, energy detection and cyclo‐stationary feature detection.
4.1 Matched filler detection
The common sense means to detect a transmission is to design an inverse filler to the impulse
response of the primary transmitter. The additional requirement would be to make a precise noise
model of the channel which can be used to decode the signal optimally.
4.1.1 Opportunities
• Matched filter detection is a widely used technique in communication and not much effort is
needed to adapt it to the requirements of a cognitive radio.
• The filter takes very little time to achieve high processing gains.
4.1.2 Challenges
• Even If inter symbol interference is completely eliminated from a transmission even than the best
performance of a matched filler is bounded by a theoretical bound known as matched filler bound.
• The biggest challenge for a matched filter is that it requires apriory knowledge of the transmission in
order to decode a signal. In case of an xG network this information can only be obtained if the
licensed spectrum user can extend the leverage. If the licensed user intends to transmit to a
matched filter detector than some information in the form of pilots, preambles, synchronization
work or spreading code will be transmitted by the transmitter which might be used by the cognitive
radio. Otherwise, the technique can only be used if licensed user intends cooperating.
4.2 Energy detection
If the transmission information is not available with the cognitive radio then the optimum
receiver is the energy detector. The detector is simply the integrated output Y of a band pass filter with
bandwidth W over a time period T. The output is compared with a threshold λ to decide whether some
signal is present in the band or not. The threshold selection can be made as a fixed value or a flexible
choice, that is though complicated but more suitable for a cognitive radio.
In [6] it is deduced that the probability of detections Pd and probability of false alarm Pf is
Pd = P [Y > λ / H1] = Qm (√2γ, √λ),
Pf = P [Y > λ / H0] = Г (m, λ/2) / Г (m)
Where λ is SNR, n = TW (time bandwidth product) and Г ( . ) , Г (. , .) are complete and
incomplete Г (gamma) functions. Qm is generalized Marcum Q function.
4.2.1 Opportunities
• As it is the most easily implement‐able detector hence it is the most widely considered detector in
research.
12 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
Fig. 6. (a) Receiver Uncertainty (b) Shadowing Uncertainty [3]
• The detector has been sufficiently studied for multi‐path and fading channels and satisfactory
results have been obtained. In [7] it is given that Pf in this case is independent of λ. When the
amplitude gain of channel h, varies due to the shadowing and fading, Pd gives the probability of
detection conditioned on instantaneous SNR as follows
Pd = Q
m (√2γ, √λ) fr(x) dx
Where fr(x) is the probability distribution function of SNR under fading conditions.
4.2.2 Challenges
• If the Pd is very low then the failure to detect a licensed primary user will erroneously create a false
hole. Consequently the interference for primary user will increase. Conversely, if the Pf is very high,
it will result in under‐utilized spectrum. This necessitates a very careful selection of comparison
threshold λ.
• Pilot detector is susceptible to noise power variations. A pilot tone from transmitter can be
considered to address this problem but again, the detectors major benefit that no transmitter
information is required at the receiver is compromised.
• An energy detector cannot differentiate communication types, hence is prone to false indication if
some un‐desirable signal is present in the considered band.
• The lack of information about the type of transmission, reduces the re‐configurable parameters,
hence results in reduced flexibility of the cognitive radio.
4.3 Cyclo‐Stationary feature detection
Cyclo‐stationary detector is based on the inherent redundancy in the transmission signals.
Modulated signals in general are associated with sine wave carriers, digital sequences in the farm of
pulse trains, repeating spreading or having cyclic prefixes. These all periodicities in these communication
signals results in an inherent autocorrelation. A cyclo‐stationary detector detects the presence of a
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 13
signal on basis of presence of periodicity in the transmission by using a spectral correlation function. The
beauty of this detector is that it can differentiate the signal easily from the noise because the noise
pattern in general, is wide‐sense stationary and indicates no periodicity.
4.3.1 Opportunities
• A cyclo‐stationary feature detector can perform better than the energy detector because of its
robustness in presence of uncertain noise powers.
• No transmitter information is required at the cognitive radio to obtain the communication
information.
• Neural network based cyclo‐stationary detectors are studied in [8] and are found extremely useful.
4.3.2 Challenges
• Cyclo‐stationary detector is computationally very complex to implement.
• Cyclo‐stationary detectors cannot detect type of communication and hence is prone to erroneous
detection from an un‐intended transmission source.
• The lack of information about communication types renders reduced flexibility to the cognitive
radio.
Though the challenges of each transmission detector technique are discussed separately but
one common problem that transmitter detection has to face is its frequent isolation from primary
network. This isolation results in what are called as “receiver uncertainty” and “shadowing uncertainty”
[3]. Receiver uncertainty exists when the receiver is unable to detect a primary transmitter due to
weakness of its signal but is adversely affecting the reception of primary receiver. Shadowing un‐
certainty is similar to receiver uncertainty except that the cause of weak signal is some obstruction in
the transmission and not the distance involved.
5. Cooperative detection
The cooperative detection is based on the cooperation between various xG users for contention
of spectrum by sharing information about their radio environment. This not only enhances the
optimality of spectrum utilization but also reduces the chances of interference to primary uses. The two
approaches to implement the cooperative detection are in a centralised or distributed manner. In
centralised approach a xG base station or hub is responsible for acquiring information from all xG users,
makes a broad picture of primary users and xG users in the complete radio environment and
disseminate the information to xG users on as required basis. In distributed approach each xG users acts
as a node and shares its neighbourhood information with all others users of xG network.
14 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
5.1 Opportunities
• Biggest advantage of cooperative detection is the manifold reduction in the uncertainties of
transmitter detection, due to sharing of information at various xG users.
• Effects of degrading factors such as multiple paths fading and shadowing are mitigated.
• Primary users’ interference is appreciably reduced.
5.2 Challenges
• In [9] the problems created due to co‐location of spectrum sensing (cooperative method) and
transmission functions are discussed and it is suggested that two separate networks for sensing and
transmission be adopted. Though problem might be resolved but at the cost of added complexity.
• In case of resource constrained network, cooperative detection might be a difficult option
considering the additional operations and overhead traffic.
• Primary receiver un‐certainty due to the passive nature of primary receiver is still un‐resolved.
Fig. 7. Interference Temperature Model [10]
6. Interference based detection
All the detection techniques discussed thus far have focused on reducing the interference to the
primary transmitter. This is because of the difficulty to detect primary receivers due to their passive
nature. However the aim remains to reduce the interference to the primary receiver irrespective of the
transmitter.
Interference based detection model has been recently proposed by FCC [10] which introduces
the idea of interference temperature. The fig 7 [10] shows an interference temp model. The power
received at primary receiver reduces exponentially with distance, until it falls to the level of noise floor.
At this point the receiver treats this communication as simply noise and not transmission. The noise
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 15
floor, though a theoretical threshold has spikes in it which are to be treated as noise by the receiver. If
an interference cap is introduced above the maximum noise level, all noise will be completely removed.
This increase in threshold results in slightly reduced range for transmitter but makes a very useful
corridor for xG networks to operate. A cognitive radio aware of the existing noise level can use the
spectrum of its choice unless the transmission power does not exceed the interference cap. Below this
level the primary receiver will treat this transmission as simply noise and no interference will occur.
The interference base detection is a new concept and is being widely researched. Novel
approaches for its implementation are still being suggested. In [11] a direct receiver detection method is
presented which uses the leakage power of LO (local oscillator) by the RF front‐end of a primary
receiver. The idea proposed is to have low cast sensor nodes which detect the leakages from primary
receiver and feed to the cognitive radios in the xG networks. This information is used by un‐licensed
users to deduce the spectrum band of choice.
6.1 Opportunities
• The biggest advantage of interference temperature based detector is the shift of focus from primary
transmitter to primary receiver. It implies that no undue effort will be made to reduce interference
e.g. If no primary receiver is around, a cognitive radio may use any frequency band of its choice
irrespective of primary transmitters in the vicinity.
• If the transmission power of a cognitive radio remains below the interference cap, it may utilise any
frequency parameters of its choice.
6.2 Challenges
The proposal is still in embryonic stage of research but is sure to break ground for a vast
improvement. The biggest challenge is the receiver interference temperature detection. The complexity
of the problem is compounded considering multiple primary and xG users.
7. Spectrum Sensing Challenges
The challenges and opportunities offered by various spectrum sensing techniques have been
discussed above. But there are few spectrum sensing challenges that are not specific to any specific
technique but are an open area for research.
7.1 Interference Temperature
As already discussed the major research done so far is to reduce interference to the primary
transmitter rather than the primary receiver. This is due to the problems faced in finding interference
temperature at the receiver due to its passive nature. A cognitive radio is generally aware of its own
transmission power levels, its location and surrounding noise level. But in order to cause minimum
disturbance to the primary users, it is mandatory to gather information about their transmission
characteristics and parameters. The main focus in this regard should be the primary receiver but
currently there exists no feasible method of detecting the interference temp at the primary receivers.
16 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
This is solely due to the passive nature of these receivers. Hence, major work till to date is focused on
reducing interference to primary transmitters. Furthermore, even if temp interference at a precise loc is
deduced, it’ll remain useless for a cognitive radio unless it can conclude the effect of its transmission on
all primary receivers.
7.2 Spectrum Sensing In Multi‐User Networks
Most of the cognitive radios will have to operate in multi‐user environments, requiring not only
minimal interference to primary users but also competing with other xG users for the same spectrum.
Most of the research done does not cater for multi‐user scenarios. However, a cooperative scheme
which is subsequently discussed can be considered. This will greatly minimize the problems arising due
to multi‐user interactions.
7.3 Speed of detection
The complete cycle of sensing, analyzing and adapting is happening in real time. Hence the
speed of acquiring spectrum information is extremely essential to avoid interference and / or missed
opportunities for spectrum utilization. This is greatly dependent on the modulation scheme of the
transmission. In [12, 13, and 14] it has been deduced that OFDM is the most suited transmission scheme
for cognitive radios. Once a primary user is detected by a single carrier, detecting all other carriers is not
necessary. This at present is also another area open for reach.
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 17
Section III – Spectrum Allocation
Spectrum sensing gives a broad‐based information to proceed on to the more important
functionality of spectrum allocation. Spectrum allocation not only comprises of the process of spectrum
analysis and spectrum decision but encompasses a variety of issues such as spectrum mobility, spectrum
sharing, spectrum access etc. Whereas the spectrum sensing primarily deals with physical layer,
spectrum allocation deals generally with the higher layers. The spectrum allocation is a widely
researched topic and major issues need to be compromised as yet. A broad of spectrum allocation
functionality is as follows:
8. Spectrum Analysis
The physical layer of spectrum sensing provides all the raw information to the spectrum analysis
functionality. The analysis not only examines the time varying radio environment but also measures
various parameters of primary and co‐existing xG users of the spectrum. The information is collected on
basis of various matrices. Generally the spectrum is analysed for three things:
• Channel capacity
• Primary user related information
• xG user information
8.1 Channel capacity
The following factors might be considered for channel capacity analysis.
• Path Loss: Path loss is directly proportional to the operating frequency. Hence it is a major
consideration for transmission power. In order to avoid interference to other users the transmission
power should be kept to the minimum but it should remain high enough to be of use. Hence the
transmission power is to be adjusted as per operating freq.
• Wireless Link Error: The error rate of the channel changes with the transmission scheme and
interference level of the spectrum band. Hence related information is of importance.
• Link Layer Delay: Due to the inherent flexibility in the design all the proposed networks are multi‐
layer concepts. This brings into consideration the delay involved in various protocol layers.
• Noise Info: The noise existent in the neighbouring radio environment must be considered in
all analyses. Especially the interference temperature model which appears to be the most feasible
spectrum sensing option utilises the noise floor as the basis of all calculations.
8.2 Primary and xG network user info
Similar information is required about xG and primary users. The following information will be
feasible in this regard.
18 Farrukh Ja
aved F‐05‐020. Spectrum Sensing and Alloca QE ‐ II)
ation in Cognittive Radios (DQ
• Interfference: Some specctrum bandss are more congested than others. The spectru um is
analyssed for inform mation aboutt the xG and p primary userss utilising the spectrum. Th his info is anaalysed
to dettect the specttrum holes.
• Holdin ng Time: Holding timme is the esstimated tim
me the vacan b occupied by a
nt slot can be
cognittive radio. Th he greater the holding tim me, the betteer is the perfo ormance of tthe cognitive radio
becau use it will red
duce the hanndoffs a cognitive radio has
h to underrgo due to primary
p userss. The
inform mation can be e acquired baased on statisstical models or a cooperaative networkk can also gen nerate
this innformation fo or all the userrs.
• User transmission Parameters:: The transmission characcteristics of the t primary and a co‐existin ng xG
user aare useful to decide the suitable specttrum for transmission. Datta such as mo odulation sch heme,
error rate, entropyy etc are usefful for the coggnitive radio tto predict thee behaviour o of these userss.
Fig . 9. Spectru
um Sensing and Allocation
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 19
8.3 Channel Capacity
Channel capacity though theoretically defined as mean average entropy but here it is considered
as a term encompassing all channel characteristics.
The spectrum characterised by channel capacity have generally made the SNR parameters as the
basis. Various models have been considered to utilise the SNR as a measure of channel capacity but
channel characteristics can never be fully encompassed utilising SNR only. All the above mentioned
factors form matrices that must be considered for channel analysis. Various models have been
suggested in this regard which take additional factors into consideration e.g. in [3] the interference
temperature model discussed in previous section is suggested for measures of channel capacity. In [14]
bandwidth and permissible transmission power are suggested as the basis of system capacity
measurement C.
C = B log ( 1 + S/ (N+I) )
Where B is the bandwidth, S is the received signal power from cognitive radio, N is the cognitive radio
noise power and I is the interference power received at the cognitive radio due to primary users. In [15]
OFDM based cognitive radio channel capacity is defined as
1
log 1
2
Where Ω is the collection of unused spectrum segments, G (f) is the channel power gain at freq
f. So & No are the signal and noise power per unit frequency respectively.
8.4 Spectrum analysis Challenges
• Heterogeneous Spectrum Sensing: All un‐conventional and unconventional analysis methods are
designed to operate on a limited frequency band. Even in software radios the general approach is to
follow a frequency filtration mechanism immediately after the RF front end and then the analogue
to digital conversion stage. However, in case of a cognitive radio the complete spectrum is to be
analysed for various parameters. The more elaborate is the analysis the more flexible and
interference free performance will be acquired.
• Non Cooperative Primary and xG users: The biggest analytical challenge is the analysis of
primary and co‐existing xG users. The data is extremely important for deciding a suitable band for
transmission. All decisions for switching from one spectrum hole to another are also based on the
same data. In a non‐cooperative xG network this poses to be a big design challenge.
• Varying Transmission Parameters: The spectrum holds multiple users at all time utilising
different modulation schemes, data, error rate etc. In order to accommodate the requisite flexibility
to the channel, cognitive radio should be able to adapt to all the varying conditions.
20 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
• Real Time Analysis: The Cognitive nature of a cognitive radio demands continuous real time
analysis. The broad spectrum time varying parameter over which the cognitive radio is operating
make the real time analysis very difficult to accomplish.
• Delays in Processing: In order to cause minimum disturbance to the primary network
(licensed users) and have better spectrum sharing with co‐existing cognitive radios it is primal to
undergo a handoff. Another situation for handoff may arise, if the conditions in the occupied slot
deteriorate below an acceptable level. Or a situation in which a more feasible slot is offered for
occupation for which the handoff losses to performance are a worthy bargain. The handoffs cause
processing delays which might be very taxing for the cognitive radio performance.
Special considerations are to be made to minimise the no of handoffs and the performance
degradation so that the cognitive radio must be able to adapt to its surroundings in minimum
possible time. This necessitates the minimisation of processing delays in the analysis stage. Though,
the advanced DSP techniques and processors have evolved to a level that makes this task possible
but still highly cumbersome.
8.4.1 Opportunities
As discussed spectrum sensing proves to be a design challenge at physical layer but the
spectrum analysis is a software design problem. The evolution of software radios has for the time made
it possible to visualise the existence a level of flexibility primal for a cognitive radio. This has rendered
many discussions to describe cognitive radio as an extension of software radios. In any case, as the
technology for a cognitive radio matures, the software radio concepts have already evolved to a level
where they can be effectively utilised for spectrum analysis and decision.
9. Spectrum Decision
The spectrum once analysed for all relevant parameters is chosen for transmission at a
particular slot. The decision also includes the transmission characteristics. This is the most
computationally extensive and complex block of cognitive radio. The decision not only caters for the
spectrum characteristics but also the protocols involved for spectrum sharing and spectrum mobility.
Based on this information few holes are identified for transmission. These are than compared with
transmission requirements to decide which slot will most suitably fulfil the user’s requirements of data
rate, modulation scheme, bandwidth etc. A suitable slot is finally selected which can provide the
required QoS for the transmission. The decision is made in real time and has to be changed with the
changing parameters. The complexity of decision block is evident from multifarious challengers it poses.
The research issues that need to be investigated to implementation of spectrum decision functionality
are:
• Spectrum management
• Spectrum mobility
• Spectrum sharing
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 21
9.1 Spectrum Mgmt:
The term is being utilised to indicate those functions of the spectrum decision block directly
related with spectrum characteristics. These include all the elements discussed in section 8.1. Based on
the discussed factors, conclusions about the spectrum utilisation are to be made. The challengers faced
in spectrum management functionality include.
• Decision Model: As mentioned in the previous section, the criterion for spectrum selection is not
restricted to SNR parameters, various models are discussed e.g. [14] and [15] included additional
spectrum characteristics for choosing the optimum spectrum characteristics. However, no proposed
criterion as yet includes all relevant requirements for measuring true channel capacity.
• Multiple Spectrum decision: A very interesting opportunistic idea discussed in [3] and [16] is
to use multiple non‐contiguous spectrum bands for transmission. This has various advantages. First,
the limitation on bandwidth to transmission is lifted, as multiple slots are simultaneously being
utilized for transmission. This will enable to use wide band transmission techniques that greatly
enhance the performance of the transmission. It is because of the fact that even if some of the slots
are compromised due to interference, still complete data will not be lost. The temperature
interference model previously discussed is also suited for wide band transmission.
• Reduced Transmission Power: Another major advantage is that transmission power required
for each slot will be much lesser. How to determine number of transmission slots and the set of
appropriate bands are still open research issues in xG networks.
• Cooperation with reconfiguration: The spectrum allocation stage is succeeded by the re‐
configurability functionality of a cognitive radio. The decisions made must be able to consider the re‐
configurability limitations in order to make as correct decision about the transmission slot and
parameters as possible.
• Heterogeneous Spectrum: The heterogeneity of spectrum creates un‐conventional and varying
problems. The spectrum contains primary users with licensed bands with greater priority use and xG
networks co‐existing and competing for spectrum resources. Spectrum sharing is accomplishing by
protocols which are at times well decided or in some cases evolving or amorphous. These all
considerations make the spectrum management problems a complex one. So spectrum slots avail as
holes, still need to be prioritized for occupation.
9.2 Spectrum Mobility
A cognitive radio is design to switch its operating frequency on the run and switch from are
white space to another, all the time. The aim is to always occupy the most suitable transmission slot.
This “Get the best channel” strategy results in spectrum mobility. The switching at a channel from one
hole to another is called “Spectrum Handoff” [3].
22 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
A spectrum handoff may be initiated in various situations. If a primary user appears, the cognitive
radio has to undergo a handoff. Another situation may be, if the conditions in the occupied slot
deteriorate below an acceptable level or a situation in which a more feasible slot is available for
occupation for which the handoff losses to performance are a worthy bargain.
Special considerations are to be made to minimise the no of handoffs and the performance
degradation that is involved in each hand off. Well defined network protocols are required for regulating
the process of handoffs. These handoffs should be smooth and arrangements should be made in the
protocols to cater for the latency involved in each had off.
[7] Suggests a multi‐layered mobility management protocol required to accomplish the mobility
functionalities. The examples suggested include use of “Wait Status” in a TCP network during the
process of handoff. The switching of transmission parameters has to be accomplished during the same
period smoothly. For a data transmission e.g. in FTP the protocol is suggested to store some packets to
transmit in the handoff period. For a real time application this packets storage would not be practical.
9.2.1 Spectrum mobility challenges
There are many open research issues in spectrum mobility. A few may be:
• Latency: The delay is one major problem. Reduction of this delay is a design challenge and refining
the process of handoff is a standardized protocol requirement. The process should be smooth and
arrangements for minimised performance degradation are to be made.
• Suitable Algorithms for mobility: Sophisticate algorithms are required to be devised to decide
suitability of channels for mobility. Thong the spectrum management functionality decides the
suitability of a slot but spectrum mobility block should indicate if the switch off is suitable in terms
of hand off losses.
• Appearance of a primary user: Appearance of a primary user creates a situation in which
handoff is to be made immediately with minimum losses. Special algorithms are to be defined that
minimise the performance loss in this situation. An example is to keep a “Most suitable next slot(s)”
always pre‐decided and in case a handoff is to be made, it may be accomplished without delay.
• Vertical and inter‐cell handoff schemes: Inter‐cell handoff is simply a switch from one slot to
another due to performance considerations while vertical handoffs are the switching at frequencies
between two different networks. In such a diverse environment it is highly necessary to devise
special procedures and mechanisms to implement these handoffs.
• Suited threshold for inter‐cell handoffs: As a handoff may be carried out if conditions at a
certain freq slot have deteriorated below an acceptable level or if another slot promises must
enhanced performance. Deciding an optimum threshold or level is highly important for both
Farrukh Ja
aved F‐05‐020. Spectrum Sensing and Alloca
ation in Cognittive Radios (DQ
QE ‐ II) 23
9.2.2 O
Opportunities for spectrum
m mobility
Th
he spectrum mobility is an
n open area for research aand not manyy algorithms aand protocolss have
been prop
posed on the subject. Som
me areas openn for research
h are indicated below.
• Prioritised white sppace: Instead of waiting for a situaation where aa handoff beccomes necesssary it
can be
b considered d to allow th
he spectrum mobility mod dule to makee a queue off suitable slots for
occuppation. Similaar queues shhould be maade by the spectrum
s shaaring, spectrrum managem ment,
spectrum sharing aand user requ uirements blo ocks. The firstt intersectionn of these queeues should b be the
optimmum slot read dy for occupation. Whenevver a situatio on arises that necessitate aa handoff, the slot
may b be immediate ely occupied w without delayy. The spectrrum decision block should also be carryying a
FIFO qqueue of suittable slots eaach having a “duration of suitability” ttag with it aftter which it leaves
the queues.
q The sequence
s in which the decision
d blockk consults itss predecesso ors should alsso be
flexible such that tthe block with maximum rejections is aalways first cconsulted. This will reduceed the
proceessing delay.
• Soft and
a hard speectrum hando off: The terrm handoff iss more suited to cellular networks buut the
adapttation to cognnitive radio iss a useful onee because it b brings some reeadily made ssolutions with h it. A
soft h
handoff would be the onee in which the next slot iss occupied beefore leaving the previouss slot.
Fig. 10. Spectru
um Decision Fun
nctionality
24 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
This results in almost no loss in performance. There might be a standby arrangement which is ready
to take over at any time having the next most feasible slot as its operating frequency. Whenever the
situation arises the handoff is made with little or no delay. This might results in enhanced
complication and increase cell occupancy but major issues related to spectrum handoff performance
degradation might be resolved.
9.3 Spectrum Sharing
The ultimate goal of a cognitive radio is to share the spectrum with the primary and co‐existing
xG users. The term “spectrum sharing” has been variedly used in different works. In [7] the spectrum
sharing is described as the complete process including spectrum sensing, spectrum allocation, spectrum
access, transmitter receiver hand‐shake and transmitter mobility. In this paper the terminology is used
to indicate the functionality of a cognitive radio that enables it contest with other xG users and interact
with primary users for sharing the spectrum.
The spectrum sharing techniques can be classified on the basis of there different aspects i.e.
According to their architectural assumption, spectrums allocation behaviour and/or spectrum access
techniques. Each approach must also elaborate the inter‐network or intra‐network sharing.
9.3.1 Architecture based classification
• Centralised spectrum sharing: As indicated by the name, there is a central entity that controls the
spectrum sharing process. In order to aid the process a network of sensor nodes is suggested which
feed the controlling central entity. Based on the information of theses sensors, a spectrum
allocation map is formulated.
• Distributed spectrum sharing: In distributed spectrum sharing each node is self sufficient with its
own sensing mechanism. The sharing is done on basis of local (or if possible global) policies.
9.3.2 Challenges and Opportunities
Centralised spectrum sharing is by far the better option as it accommodates not only the
spectrum sharing but also as a hub for control / Coordination of various cognitive radios in the xG
network. Sharing of interference information and spectrum concentration can also be communicated to
each user through the central hub. The implementation of protocols will be much easier and less delay
will be involved.
Distributed spectrum sharing is considered in situations where development of an infrastructure
is not possible due to limited resources. This results in reduced spectrum utilisation but economic use of
resources.
9.3.3 Spectrum Sharing based on the access behaviour
The spectrum sharing technique may also be classified on basis of access behaviour. The
spectrum access can be co‐operative or non‐cooperative.
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 25
9.3.4 Challenges and Opportunities
The compromise like previous case is again between the simplicity and performance. A co‐
operative scheme is of course more optimal than the non‐cooperative scheme. In [20] it is studied that
the spectrum utilization in case of a cooperative spectrum sharing is much higher than a non‐
cooperative spectrum sharing scenario even approaching the global optimum. The study is carried at on
basis of spectrum utilization, fairness, throughput, channel allocation and potential neighbours.
Similar results are obtained in [21]. However, limitation on these works is the assumption that
location and transmission power at primary users is known which may not generally be the case. The
non‐cooperative sharing can be considered in restrained resources situation where they are beneficial
due to the minimal communication requirements.
9.3.5 Spectrum sharing based on access technology
• Overlay spectrum sharing: This is the situation where a cognitive radio occupies only the vacant
slots (holes) in the spectrum and vacates them on appearance of primary users. As a result
interference to primary system is minimised.
9.3.6 Challenges and Opportunities
Spectrum sharing requirement based on access technology are a direct dictate of the spectrum access
technique being utilized. Only from a spectrum sharing point of view, the underlay technique is more
feasible as there is no requirement of information about the co‐existing primary and xG users. The only
consideration is the noise level. However, performance improvement can be acquired by having a
26 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
centralised or cooperative network which renders each node capable of differentiating between noise
and another xG user. Another advantage is the availability of much wider bandwidth which makes the
completion for bandwidth less fierce. An overlay spectrum sharing techniques has the primal advantage
of creating less interference for the primary network.
9.4 Spectrum Sharing Challenges
Spectrum sensing challenges because of the type of technique being utilised have been
discussed but some well recognized common spectrum sharing challengers are discussed below:
Fig. 11. Spectrum Unit [3]
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 27
• Spectrum Unit: The complete operation of xG networks focuses on switching of channel. Hence
the definition of channel or spectrum unit is highly critical. Different modulation schemes use
different spectrum units. This aspect has not been discussed in detail in most work done so far. In
[28] a spectrum space is introduced for xG networks with power, freq, time, space and signal as its
possible dimensions. Although not orthogonal, but these dimensions can be used to distinguish
signals. In [3] a virtual cube model is discussed. The resource is modelled in a 3‐dimensional
resource space with time, rate and power/code as dimensions. The rate dimension models the data
rate of the network the time dimension models the time required to transfer information.
10. Conclusion
It is not an over‐statement to declare the cognitive radio as the future of telecommunication
rather than the part of the future. An environment aware, intelligent network with cognitive radios as its
nodes is the only available option for sustaining the enhanced and varying needs that continue to
appear with every passing day. In this paper cognitive radios are discussed with special emphasis on the
spectrum sensing and allocation functionalities. Various techniques used are discussed along with the
challenges and opportunities offered by them. The sub‐functionalities of each major block are also
explained in detail. The discussion on spectrum allocation also includes the discussion on spectrum
sharing, spectrum mobility and spectrum management as part of spectrum allocation. The paper is a
survey of latest trends in spectrum sensing and allocation functionalities of cognitive radios and also
highlights open areas for research.
28 Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II)
References
[1] FCC, ET Docket No 03‐222 Notice of proposed rule making and order, December 2003.
[2] Jean‐Pierre Hubaux, M. H. Manshaei, M. Felegyhazi, J. Freudiger, and P. Marbach: Spectrum Sharing
Games of Network Operators and Cognitive Radios. EPFL, August 2005.
[3] Ian F. Akyildiz, Won‐Yeol Lee, Mehmet C. Vuran, Shantidev Mohanty: NeXt generation/ dynamic spectrum
access / cognitive radio wireless networks; A survey. School of Electrical and Computer Engineering, Georgia
Institute of Technology, United States. January 2006.
[4] J. Mitola III, Cognitive radio: an integrated agent architecture for software defined radio, Ph.D Thesis, KTH
Royal Institute of Technology, 2000.
[5] S. Haykin, Cognitive radio: brain‐empowered wireless communications, IEEE Journal on Selected Areas in
Communications 23 (2) (2005) 201–220.
[6] F. Digham, M. Alouini, M. Simon, On the energy detection of unknown signals over fading channels, in:
Proc. IEEE ICC 2005, vol. 5, May 2003, pp. 3575–3579.
[7] A. Ghasemi, E.S. Sousa, Collaborative spectrum sensing for opportunistic access in fading environment, in:
Proc. IEEE DySPAN 2005, November 2005, pp. 131–136.
[8] A. Fehske, J.D. Gaeddert, J.H. Reed, A new approach to signal classification using spectral correlation and
neural networks, in: Proc. IEEE DySPAN 2005, November 2005, pp. 144–150.
[9] S. Shankar, Spectrum agile radios: utilization and sensing architecture, in: Proc. IEEE DySPAN 2005,
November 2005, pp. 160–169.
[10] FCC, ET Docket No 03‐237 Notice of inquiry and notice of proposed Rulemaking, November 2003. ET
Docket No. 03‐ 237.
[11] B. Wild, K. Ramchandran, Detecting primary receivers for cognitive radio applications, in: Proc. IEEE
DySPAN 2005, November 2005, pp. 124–130.
[12] I.F. Akyildiz, Y. Li, OCRA: OFDM‐based cognitive radio networks, Broadband and Wireless Networking
Laboratory Technical Report, March 2006.
[13] H. Tang, Some physical layer issues of wide‐band cognitive radio system, in: Proc. IEEE DySPAN 2005,
November 2005, pp. 151–159.
[14] T.A. Weiss, F.K. Jondral, Spectrum pooling: an innovative strategy for the enhancement of spectrum
efficiency, IEEE Radio Communication Magazine 42 (March) (2004) 8–14.
[15] H. Tang, Some physical layer issues of wide‐band cognitive radio system, in: Proc. IEEE DySPAN 2005,
November 2005, pp. 151–159.
[16] I.F. Akyildiz, Y. Li, OCRA: OFDM‐based cognitive radio networks, Broadband and Wireless Networking
Laboratory Technical Report, March 2006.
Farrukh Javed F‐05‐020. Spectrum Sensing and Allocation in Cognitive Radios (DQE ‐ II) 29
[17] V. Brik, E. Rozner, S. Banarjee, P. Bahl, DSAP: a protocol for coordinated spectrum access, in: Proc. IEEE
DySPAN 2005, November 2005, pp. 611–614.
[18] C. Raman, R.D. Yates, N.B. Mandayam, Scheduling variable rate links via a spectrum server, in: Proc. IEEE
DySPAN 2005, November 2005, pp. 110–118.
[19] S.A. Zekavat, X. Li, User‐central wireless system: ultimate dynamic channel allocation, in: Proc. IEEE
DySPAN 2005, November 2005, pp. 82–87.
[20] C. Peng, H. Zheng, B.Y. Zhao, Utilization and fairness in spectrum assignment for opportunistic spectrum
access, in: ACM Mobile Networks and Applications (MONET), 2006.
[21] H. Zheng, C. Peng, Collaboration and fairness in opportunistic spectrum access, in: Proc. IEEE ICC 2005,
vol. 5, May 2005, pp. 3132–3136.
[22] V. Brik, E. Rozner, S. Banarjee, P. Bahl, DSAP: a protocol for coordinated spectrum access, in: Proc. IEEE
DySPAN 2005, November 2005, pp. 611–614.
[23] L. Cao, H. Zheng, Distributed spectrum allocation via local bargaining, in: Proc. IEEE Sensor and Ad Hoc
Communications and Networks (SECON) 2005, September 2005, pp. 475–486.
[24] L. Ma, X. Han, C.‐C. Shen, Dynamic open spectrum sharing MAC protocol for wireless ad hoc network, in:
Proc. IEEE DySPAN 2005, November 2005, pp. 203–213.
[25] S. Sankaranarayanan, P. Papadimitratos, A. Mishra, S. ershey, A bandwidth sharing approach to improve
licensed spectrum utilization, in: Proc. IEEE DySPAN 2005, November 2005, pp. 279–288.
[26] J. Zhao, H. Zheng, G.‐H. Yang, Distributed coordination in dynamic spectrum allocation networks, in: Proc.
IEEE DySPAN 2005, November 2005, pp. 259–268.
[27] H. Zheng, L. Cao, Device‐centric spectrum management, in: Proc. IEEE DySPAN 2005, November 2005, pp.
56–65.