Content uploaded by Usama Mir
Author content
All content in this area was uploaded by Usama Mir on Mar 08, 2016
Content may be subject to copyright.
203
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
Dynamic Spectrum Sharing in Cognitive Radio Networks: a Solution based on Multiagent
Systems
Usama Mir, Leila Merghem-Boulahia, Dominique Gaïti
ICD/ERA, UMR 6279,
Université de Technologie de Troyes,
12 rue Marie Curie, 10010 Troyes Cedex, France
{usama.mir, leila.merghem_boulahia, dominique.gaiti}@utt.fr
Abstract— In modern day wireless networks, spectrum utilization
and allocation are static. Generally, static spectrum allocation is not
a feasible solution considering the distributed and dynamic nature
of wireless devices, thus some alternatives must be ensured in order
to allocate spectrum dynamically and to mitigate the current
spectrum scarcity. An effective technology to ensure dynamic
spectrum usage is cognitive radio, which seeks the unutilized
spectrum holes opportunistically and shares them with the
neighboring devices. Using cognitive radio capabilities, the nodes
are not restrained to static spectrum utilization, rather they can
choose it on demand. However, dynamic spectrum usage raises
several challenges, which need to be addressed in detail. These
challenges include efficient allocation of spectrum between licensed
(or primary) and cognitive radio (or secondary) users in order to
maximize spectrum utilization and to avoid device level
interferences. To this extend, we develop a novel solution for
spectrum allocation using multiagent system cooperation that
enables secondary user devices to utilize the amount of available
spectrum, dynamically and cooperatively. The key aspect of our
design is the deployment of agents on each of the primary and
secondary user devices that cooperate in order to have a better use
of the spectrum. For cooperation, contract net protocol is used,
allowing spectrum to be dynamically allocated by having a series of
message exchanges amongst the devices. Simulation results show
that our solution achieves up to 80% of the whole utility within the
span of few messages, and provides an effective mechanism for
dynamic spectrum allocation.
Keywords- Multiagent Systems; Cognitive Radio; Dynamic Spectrum
Sharing; Contract Net Protocol; Cooperation; Ad hoc Networks.
I. INTRODUCTION
In most of the modern day applications, radio spectrum
allocation and sharing is a static function, in which the spectrum
is assigned to a particular dominant primary (or licensed) user
[3], for a long period of time in order to avoid interferences and
collisions. Parallel to this, to deal with increasing user demands,
dynamic spectrum allocation for new wireless networks is
necessary. However, since existing wireless networks occupy
extensive parts of the radio spectrum, there is no sufficient
spectrum available to all the new unlicensed wireless networks
[1] [25]. Thus, research has to be done to address this problem
via dynamic sharing and assignment of spectrum. For example,
in USA, Federal Communication Commission (FCC) considers
to allow sharing of unused portions of TV bands to promote
dynamic use of spectrum [2] [4].
One effective technology to alleviate the problem of static
spectrum assignment and to maximize dynamic spectrum usage
is cognitive radio (CR) [17], a radio in modern wireless systems,
in which a CR (or a secondary user) node changes its parameters
(transmission or reception) to share the spectrum dynamically
and to avoid the interference with the other primary or secondary
users. The parameter alteration is done by having some
knowledge about the radio environment factors such as radio
frequency (RF) signals, device level interferences, etc. To
achieve efficient and dynamic allocation of spectrum between
highly distributed CR devices, a balanced, simple and
cooperative approach is necessary. Research is therefore in
progress on exploring the cooperative spectrum sharing
techniques in CR networks. Similar to CR network, a multiagent
system (MAS) [21] [27] is a system composed of multiple
autonomous agents, working individually or in groups (through
interaction) to solve particular tasks. Like CR nodes, agents
work dynamically to fulfill their user needs and no single agent
has a global view of the network. Each agent maintains its local
view and shares its knowledge (when needed) with other agents
to solve the assigned tasks.
Recent advances in technology (especially in the domain of
programmable integrated circuits and distributed artificial
intelligence) have created an opportunity for us to develop a new
class of intelligent, autonomous, and interactive CR devices [8].
These devices can then be used in a wide variety of network
domains (WLAN [48], WRAN [49], MANETs [23]). In
addition, an efficiently designed CR with a software agent
deployed on it would be capable of interacting with neighboring
radios to form a dynamic, loosely-coupled and infrastructure-
less collaborative network. While CR physical architecture and
its sensing capabilities have received considerable attention [5]
[28], the question of how to share radio resources in cooperative
scenarios is also an important research issue for current
researchers [3] [8] [22].
Therefore, in this paper, a MAS based strategy is proposed for
dynamic spectrum allocation. Specifically, we consider a
204
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
cooperative MAS [29] [36], in which the agents are deployed
over primary and secondary1 user devices. By cooperative MAS
we mean that the primary user agents exchange a tuple of
messages and help neighboring secondary user agents to
improve their spectrum usage. Moreover, the cooperation
mechanism we develop is similar to that of contract net protocol
(CNP) [10] [30], in which the individual secondary user (SU)
agent should send messages to the appropriate neighboring
primary user (PU) agents whenever needed and, subsequently,
the related PU agents should reply to these agents in order to
make spectrum sharing agreements. We propose that the SU
agents should take their decisions based on the amount of
spectrum, time and price proposed by the PU agents and should
start spectrum sharing whenever they find an appropriate offer
(without waiting until the reception of all the neighboring PU
agents’ responses [14]). Then, after completely utilizing the
desired spectrum, SU agents should pay the agreed price to the
respected PU agents.
In fact, this work is divided into following four parts:
First, we present a brief state of the art on various
available approaches for spectrum sharing using
multiagent systems, game-theoretical approaches and
medium access control solutions.
Second, we detail four different scenarios, in which
spectrum sharing challenges need to be addressed in
details. We also propose some initiative measures,
which are necessary to be taken for efficient utilization
of the available spectrum in the mentioned scenarios.
Third, we present a cooperative framework with the
related spectrum sharing algorithms. The proposed
MAS is cooperative where PU agents exchange a series
of messages to share their spectrum with the requesting
SU agents. The more complex scenarios with agents’
competitive behaviors will be examined as a part of our
future study.
Finally, we conduct extensive simulations to verify the
working of the proposed cooperative algorithms for
dynamic spectrum sharing in the context of cognitive
radio networks.
The rest of the paper is organized as follows. The following
section briefly presents related works. Section III presents four
scenarios, in which dynamic spectrum sharing is a vital issue. In
Section IV, we describe spectrum allocation problem with the
help of an example. In Section V, we propose our model with
the interlinked working of various modules and their related
algorithms. The experimental setup, some results and
discussions are given in Section VI. Section VII concludes our
work with the future perspectives.
1 The words cognitive, secondary and unlicensed user will be used
interchangeably throughout the article.
II. PRIOR WORK
Research has been going around for several years in order to
apply multiagent systems for decision making process and
resource sharing. A rather new application of multiagent systems
is for efficient allocation of spectral resources in CR networks.
In TABLE I, we give the similarities between an agent and a
CR. Basically, both of them are aware of their surrounding
environments through interactions, sensing, monitoring and they
have autonomy and control over their actions and states. They
can solve the assigned tasks independently based on their
individual capabilities or can work with their neighbors by
having frequent information exchanges.
TABLE I. COMPARISON BETWEEN AN AGENT AND A CR
Agent Cognitive radio
Environment awareness via past
observations Sensing empty spectrum portions and
primary user signals
Acting through actuators Deciding the bands/channels to be
selected
Interaction via cooperation Interaction via beaconing
Autonomy Autonomy
Working together to achieve shared
goals Working together for efficient
spectrum sharing
Contains a knowledge base with local
and neighboring agents’ information Maintains certain models of
neighboring primary users’ spectrum
usage
In literature, few strands of work have focused on spectrum
sharing using MAS [13] [37]; but in these works, several
limitations exist. For example, in [37], a MAS is used for
information sharing and spectrum assignments. All the
participating agents deployed over access points (APs), form an
interacting MAS, which is responsible for managing radio
resources across collocated WLANs. The authors have not
provided any of the algorithms and results for their approach.
The work in [13] considers a distributed and dynamic MAS
based billing, pricing and resource allocation mechanism where
the agents work as the auctioneers and the bidders to share the
spectrum dynamically. The protocol used for radio resource
allocation between the CR devices and operators is termed as
multi-unit sealed-bid auction, which is based on the concept of
bidding and assigning resources. The ultimate aim of using
auctions is to provide an incentive to CR users to maximize their
spectrum usage (and hence the utility), while allowing network
to achieve Nash Equilibrium (a solution concept, where each
user is assumed to know the equilibrium strategies of the other
users, and no user has anything to gain by changing its own
strategy). Auctions have traditional drawbacks of users’
untruthful behaviors, which can cause serious drawbacks to the
working of loyal users.
Game-theory has also been exploited for spectrum allocations
in CR networks [6] [11] [18] [19] [39]. In game-theoretical
approaches, each SU has one individual goal i.e., to maximize its
spectrum usage and the Nash equilibrium is considered to be the
205
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
optimal solution for the whole network (or game). Furthermore,
it incorporates two basic assumptions: first, the rationality
assumption, that is, the participating primary and secondary
users are rational so that they always choose strategies that
maximize their individual gain. And, second, the users’ common
knowledge assumption, which includes the definitions of their
preference relationship. These assumptions may behave well by
allowing each user (or player) to rationally decide on its best
action, although in most of the competitive games, sometimes
users can provide false information in order to maximize their
profits and thus can affect the whole network performance.
According to some current research works, spectrum sharing
problems are similar to medium access control (MAC) issues [9]
[32], where several users try to access the same channel and
their access should be shared with the neighboring users to avoid
the interferences. Generally, in MAC-based spectrum sharing
solutions, when a CR user uses a channel, it sends a busy signal
to the neighboring users through a control channel in order to
avoid the interference. To estimate control signals, the authors of
[20] suggest a fast fourrier transform-based radio design, which
enables CR users to detect the carrier frequency of a control
signal without causing any harmful collisions to the neighboring
users. Others [23] suggest the use of a global plan to exchange
the control information between CR devices. However,
maintaining global plans needs a large amount of frequent
information to be exchanged between CR users causing complex
device level architectural overheads.
III. SPECTRUM SHARING SCENARIOS
Here, we provide some of the possible scenarios, which need
the development of new solutions for dynamic spectrum sharing.
These scenarios are addressed as a part of a Franco-German
project TEROPP [46]. This project aims at developing various
efficient spectrum management solutions. Up till now, our
contribution to this project is the development of a cooperative
approach for opportunistic spectrum allocation. In these
scenarios, the current spectrum assignments are static and inter-
device collision is a big issue. Therefore, efficient solutions are
needed in order to enable dynamic spectrum usage and to avoid
interferences. The scenarios are divided into four different
domains as follows: (1) Spectrum sharing and interference
avoidance in ISM bands, (2) Spectrum sharing in cellular
networks, (3) Opportunistic spectrum utilization in TV bands,
and (4) Spectrum allocation in ad hoc networks. After detailing
and suggesting possible initiatives towards dynamic spectrum
access, we will describe our cooperative framework as a solution
to enable spectrum sharing under ad hoc network domain.
Precisely, multi-hop architectures, topology changes and arrival
and departure of nodes at any time are the reasons for
developing a cooperative solution for dynamic spectrum sharing
under ad hoc network setting.
A. Spectrum Sharing and Interference Avoidance in ISM Bands
Recently, WLAN [26] has been adopted as a common
technology by internet home users and companies. Characterized
by cheap devices and reasonable data-rates, WLANs can be
deployed anywhere. Designed to operate over license-free ISM
(Industrial, Scientific and Medical) bands, WLANs are restricted
to employ only few orthogonal channels, which is more than
enough to provide wireless access in a residential area.
However, the huge increment in the number of WLANs
operating in the same location introduced a new interference
level that could be anarchic. This interference is considered to be
the main limitation for WLANs performance and it introduces
new challenges to all the neighboring technologies that operate
in the ISM bands [26]. Similar problems may arise with the
deployment of LTE femto-cells [40]. These small cells, located
at a home or a building, can provide better coverage and higher
capacity in indoor environments. However, they suffer from
interference caused by the neighboring femto-cells. The
common point of introducing these two cases is that the
interference is most of the times unwanted and it needs to be
avoided.
As an interference avoidance solution, we foresee a
cooperative environment where the devices in a WLAN or LTE
cell can have CR capabilities, which allow them to optimize
frequency reuse. They can also select an alternative spectrum
portion, in case of any interference. Then, they can send the
newly searched spectrum portion information to the neighboring
devices in order to avoid the possible collisions.
B. Spectrum Sharing in Cellular Networks
This scenario explains the spectrum sharing issues for
cellular networks where the area is administrated by a central
entity (such as a base station) and it is able to impose basic
etiquettes to the users [7]. The mobile users (having CR
capabilities), can perform signal measurements and can apply
the etiquettes in order to contribute to an efficient use of the
available spectrum. These etiquettes may be in the form of
behavioral rules, such as, using correct MAC address, switching
to a convenient base station and transmitting measurement
reports. In such a context, distributed operational modes will be
privileged and different overlay functions may be implemented
such as rendezvous facilities [16], in order to optimize frequency
reuse and to enable efficient usage of available bands.
A hospital can be considered as an application example of
this scenario, where the number of users cannot be determined in
a precise way. With the CR capability, a given terminal (a
doctor’s iPhone or a PDA) might be able to sense the best
possible spectrum band. This band can then be shared and
coordinated with the neighboring devices by having a series of
interactions using multiagent systems and by taking into account
the number of current users and their priorities. The shared band
information can then be sent to the BS agent for the
administrative purposes.
206
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
C. Opportunistic Spectrum Utilization in TV Bands
The European countries are working on improving their TV
services by stopping the broadcast of PAL (phase alternative
line) signals and using DVB-T (digital video broadcasting-
transmitter) standards instead [41]. This process will create a
sufficient amount of unused spectrum resources especially in the
case of digital dividend [45]. Let us explain the exploitation of
ultra high frequency (UHF) bands to understand the concept of
numerical dividend. Generally, UHF bands are split in channels,
where channels 21 to 69 were originally assigned to TV
services. These channels are 8MHz in width, and the channel 21
corresponds to the bands 470-478 MHz. A DVB-T covers a city
and its neighboring sectors, and uses 6 UHF channels to
broadcast 36 TV programs. For example, in France, nearly 100
DVB-transmitters are used for broadcasting TV programs [42].
In a given place, we can expect that the TV services use only 6
among 49 UHF channels, leaving 43 channels as unutilized.
These huge amounts of empty spectral resources justify the
world interest for TV bands.
In a conference held under WRC’07 (World
Radiocommunication Conference) [44], discussions about the
utilization of the TV bands have already been started. The
researchers have decided to assign the UHF channels 60 to 69 to
IMT (International mobile telecommunication) services. Another
initiative is taken by the European countries with the creation of
the task group 4 (TG4) [43]. TG4 is responsible for measuring
the performance of DVB-transmitters in order to utilize the
unused TV bands opportunistically. These measurements will
then be compared with the results obtained from mobile devices
working in WiMAX.
To summarize, we provide here a few steps to be taken for
the opportunistic utilization of spectral resources in digital
dividend as following:
At first, DVB-transmitters must have the capabilities of
cognitive radios for sensing, characterizing and
monitoring the unutilized TV bands. This is possible
with the development of efficient signal processing
techniques.
Then, because DVB-transmitters normally share their
spectral resources with the radio microphones,
therefore more precise spectrum sharing techniques
must be deployed.
Finally, some techniques must be ensured in order to
differentiate between a DVB-T and a microphone
signal.
D. Spectrum Allocation in Ad hoc Networks
Here, we give an example of an SU equipped with CR
capabilities and agent functionalities. The user is in a remote or
an emergency situation, where it does not have direct access to
radio resources and its access technology requires an energy that
the user does not own. In this situation (as shown in Figure 1),
SU senses the nearby signals of primary users PU1 and PU2
(step 1) and cooperates with the agents deployed on them (step
2). This cooperation process allows SU to act on primary users’
responses by utilizing their available spectrum (step 3). Thus,
the cognitive radio capability of an SU plays the role of
interoperability, such that it can receive the information about
neighboring users’ spectrum bands and their access technology.
Likewise, the role of the deployed agent is to cooperate and
modify SU’s software configuration by loading the necessary
algorithms that fit the best to the current state (step 4).
Figure 1. Description of an ad hoc scenario
Figure 2. Ad hoc WLAN with three primary and six secondary users
IV. PROBLEM STATEMENT
In the above scenario, we have presented the role of an agent
and a CR in an ad hoc emergency situation. However,
considering a more general and practical perspective, we address
the spectrum allocation challenges in a private ad hoc area or a
well identified administrated perimeter such as a campus, a
conference center or a hospital. Note that our proposed
algorithms can also be easily applied for the emergency ad hoc
network scenario.
. In our proposed scenario (Figure 2), there is an ad hoc
WLAN [15], deployed in the area with sets of primary PU =
(PU1, PU2 ….. PUn) and secondary SU = (SU1, S2 …… SUm)
users. To allow nodes to communicate, the agents are deployed
207
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
at each of them Whenever an SU device detects an empty
portion of the spectrum as needed by its user, its agent starts
communicating with the relative PU agent (having that empty
spectrum part), until a spectrum sharing agreement is been
made.
A. Formalization
Let G = (N, A) be a directed network consisting of a set of
mobile nodes N such that (SU
PU)
N and a set of directed
arcs A. Each directed arc (i, j)
A connects a secondary user
SUi to a primary user PUj. Similarly, we can denote the directed
arc (j, i)
A to show the direction of connection from PUj to
SUi. The secondary users are cooperating with the neighboring
primary users to have a spectrum sharing deal. We assume that
sij is the amount of spectrum a secondary user ‘i’ is desiring to
get from a primary user ‘j’. Similarly, tij is the amount of time,
for which ‘i’ wants to utilize the spectrum and pij is the price it is
willing to pay to ‘j’. For the primary user ‘j’ on the other hand,
sji is the amount of spectrum it is willing to share with ‘i’, tji is
the respected time limit and pji is the price it is expecting to get
after sharing its spectrum. We can formulate the above model
for each secondary user ‘i’ as:
Maximize
Aji ijijts
),(
(1)
Subject to
Minimize
Aji ij
p
),(
SU
N (2)
Similarly for primary users:
Maximize
Aij ji
p
),(
(3)
Subject to
Minimize ji
Aij jits
),(
PU
N (4)
And lji ≤ sji ≤ uji
where lji and uji are the lower and upper bounds of available
spectrum of primary user ‘j’. This means that the secondary user
‘i’ cannot ask for an amount of spectrum above this limit.
B. An Example
In static circumstances, the spectrum portions are assigned to
primary users and in response the internet service providers get
their spectrum price. As an example consider a primary user
PUj, who has bought a portion of a spectrum of the size of 8MB
(Figure 3). During the peak office timings (t0-t1), the assigned
portion may remain busy (or used) due to high user traffic such
as for video conferencing and lecturing, but most of the other
times (t1-t2 and t3-tn) the spectrum can remain unused. Obviously
at free timings, PUj can utilize its spectrum portion for other
activities (e.g., watching video songs) but generally people
prefer these kinds of activities to be performed on week-ends.
With our proposed solution, a given secondary user SUi will be
able to choose the best spectrum band/channel dynamically. This
choice is made in cooperation with the agent embarked on PUj
[35], by taking into account the amount of spectrum needed, the
respected time limit and the related price.
Figure 3. An example of a primary user’s spectrum utilization during a day
V. COOPERATIVE SOLUTION FOR DYNAMIC SPECTRUM
SHARING
In this section, we explain the proposed cooperative spectrum
sharing scheme, with primary and secondary user’s internal
architectures and their algorithmic behaviors.
A. Agent
We start here by defining an agent as a dynamic and loosely
coupled unit, having the capabilities of performing a task
autonomously, based on the knowledge received from its
environment and/or through other agents’ interactions. These
loosely-coupled units then work together to form a multi-
agent system [21] [27]. Generally, an agent is appropriate and
relevant for an SU node in a sense that it allows the introduction
of various artificial intelligence (AI) techniques [12] to CR
networks and helps an SU node to behave more efficiently by
having frequent interactions with its neighboring devices. Once
in place, cooperative multiagent systems have the potential of
increasing the SU capabilities in a variety of ways. For example,
a single SU agent is limited in its knowledge (and information)
about spectrum access, but a bundle of SU agents can
collectively identify spectrum holes and can communicate them
to other nodes.
B. Contract Net Protocol
In multiagent literature, several approaches exist for
cooperation [12]. Amongst these approaches, contract net
protocol (CNP) [30] [34] is the most simple way for agents’
cooperation and decision making. In CNP (Figure 4), the
collection of agents is called contract net and several agents can
form these nets in order to solve the assigned tasks. Each agent
can either be a manager or a contractor. Basically, the manager
agent initiates a task to the contractor agents by sending Call for
Proposals (CfPs) messages. As a result, various eligible
208
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
contractors show their interest (in solving the task) by sending
their proposals. The manager selects the best proposal (via
accept) and the contract is then awarded to the selected
contractor. The selected contractor solves the assigned task as
agreed with the corresponding manager. Due to its simple and
efficient nature, our proposed approach is based on bilateral
message exchange and task allocation mechanisms of CNP.
Figure 4. Message exchange in CNP
C. Working of the Proposed Solution
The SU based design (Figure 5 and algorithm. 1) consists of
the following five different interlinked modules.
First, the dynamic spectrum sensor (DSS) is used to sense
the empty spectrum portions (or spectrum holes). Several
techniques exist for spectrum sensing such as PU’s weak
signal and its energy detection [28], cooperative centralized
detection [5], etc. For DSS, it is necessary that the sensing is
performed by considering a real-time dynamic environment,
because it is not obvious at what time a spectrum band is
occupied or when it is free.
The second module spectrum characterizer (SC)
characterizes the spectrum holes based on the Shannon’s
theorem [33] to create a capacity based descending ordered
list of all available PUs.
Secondary user interface (SUI), which is the third part sends
a request message to the SU device agent, whenever a user
wants to have a portion of spectrum (for internet surfing,
watching high quality videos, etc).
The fourth part, agent’s knowledge module (AKM) gets PU
characterization information from SC, which serves as a
motivation for agents that subsets of PUs having vacant
spectrum spaces are available. This list is not permanent
rather it is updated and maintained on regular time intervals
based on the information provided by SC module.
Moreover, AKM creates a CfP message based on the inputs
from SUI and SC: CfP (SUID, s, t, d)
where SUID is the secondary user ID (or the secondary
user’s agent identification) and it is used to help PU to reply
back to the corresponding SU, s is the amount of spectrum
needed by the SU, t is the desired time limit (or holding
time) for the spectrum utilization, and d is the deadline to
receive the primary users’ proposals.
Finally, agent coordination module (ACM) geo-casts the
CfP to the neighboring (and currently available) PU agents.
By available PUs, we mean that the PU agents have not yet
left the one-hop neighborhood and they have some unused
spectrum to share. Moreover, ACM is also responsible for
selecting the most suitable received proposal.
Having received the CfPs, the interested PU agents send their
proposals to the corresponding SU agents. The proposal is in the
following form:
Proposal (PUID, s, t, p)
where PUID is the primary user’s agent identification, s is the
amount of spectrum PU is willing to give to the respected SU, t
is the proposed spectrum holding time, and p is the price PU is
willing to receive. Note that the PU agent only contains AKM
and ACM modules, where AKM manages the neighborhood
information and ACM selects the most suitable CfP via
cooperation.
Figure 5. Working of CR and agent modules
Each PU maintains an ordered list of CfPs in its cache based
on the values of s and t for the purposes of future cooperation
(algorithm 2). At the same time, the receiving SU locally sorts
fetched proposals and an accept message is sent to the most
suitable proposal. The information of selected PU is also sent to
AKM (of SU) for future interactions. In case of an accept
message from the selected SU, the spectrum sharing is started
based on agreed parameters from both the sides. PU can still
respond to further CfPs if it wants its other unused spectrum
portions to be shared. If the PU receives a reject message from
SU, it continues sending proposals to further available CfPs, for
which the deadline is not yet expired.
209
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
Above we have presented a cooperative framework for
spectrum allocation that can generate highly effective behavior
in dynamic environments and achieve better utility of the
participating agents. The proposed solution is based on
multiagent system cooperation with the deployment of agents
over primary and secondary users. The experimental evaluations
presented in the following section will confirm the efficiency of
our proposal for dynamic spectrum allocations.
VI. EXPERIMENTS AND RESULTS
In this section, we present some simulation results,
conducted in order to validate the working and performance of
the proposed spectrum allocation algorithms. We start by
examining the achieved utility of both primary and secondary
users and then compare the time values, for which the spectrum
is being utilized. We also present the spectrum gain and loss
with the amount of messages used for cooperation. The words
(PU, PU agent, SU, SU agent respectively) are used
interchangeably throughout the following section.
A. Simulation Setup
We perform our simulations under the assumption of a
noiseless and mobile ad hoc network. By mobile ad hoc we
mean that the nodes in the neighborhood of each of the SUs
change. We randomly place a number of primary and secondary
users in a specified area where each of the devices contains an
agent deployed over it for cooperation purposes. For simplicity,
two different fixed values of times (such as T1 and T2) are
assumed, where “Time 1” (T1) represents the short-term case
and “Time 2” (T2) is the longer period. When T1 is considered,
the SU agents can ask for an amount of spectrum within one
hour limit (i.e., 0
T1
60Minutes) and similarly this limit is
within two hours, as in case of T2 (i.e., 0
T2
120Minutes).
These two approximations capture the same amount of time
values in real wireless environments without delving into
complex situations. Our simulation starts with the total number
of 6 SUs and 4 PUs, and for each next round there is an addition
10 agents (i.e., 6 SUs and 4 PUs). The simulation is conducted
for 10 subsequent rounds, with a total of 20 hours per day, for
both T1 and T2 respectively and the average values of
parameters are taken to draw the graphs. The PU agent’s utility
is calculated as the price paid by SU agents for spectrum
utilization divided by the amount of spectrum it has shared for
the respected time period (holding time) as required by the SUs.
The SU agent’s utility is represented as its spectrum usage for
the required time divided by the corresponding price paid to the
PUs. Thus, by assigning weights or priorities to each of the
mentioned parameters, the appropriate utility values for both the
primary and secondary users are chosen.
We assume that each PU has random available spectrum
portions and the neighborhood of SUs and PUs is randomly
changing. Also, we follow the assumption that once agreed, PUs
would not be able to withdraw their commitments and they
Algorithm 1: Behavioral Algorithm for an SU
Init – Let PU be the set of primary users in secondary user agent’s
one-hop neighborhood and ℓ is the time interval base d on the
information provided by the SC module in order to maintain capacity
based ordered list of primary users.
/* SU characterizes each primary user on the basis of
capacity*/
For each i{ iЄ PU) } do
Eval (SNR(i))
/* SNR: is the primary user’s signal to noise ratio
obtained through DSS */
Eval (B(i))
/* B: bandwidth of PU given by DSS*/
C(i)= B(i) log2 [1 + SNR(i)]
/* c: capacity calculated using Shannon’s
theorem*/
End For
/*Sending of CfP message*/
If PU != {}
For each i Є PU
/*Geo-cast CfP*/
Send CfP (SUID, s, t, d) to PU(i)
End for
End If
/*L is a list for saving received proposals*/
For each received proposal ‘m’ do
Characterize m using )( )()( mp mtms
and add it in L
End For
If L={} and the deadline to receive proposals has expired
Recreate CfP
Else If L={i} where i is the only element in L and deadline
for proposal reception has expired
Send an accept message to i
Else
Send accept to primary user
corresponding to the best proposal
Send reject to all other primary users
End If
Algorithm 2: Behavioral Algorithm for a PU
While busyflag = false do
If received message = CfP
/*K is a list for saving received CfPs*/
For each received CfP ‘n’ do
Characterize n using )()( )( ntns np
and add it in K,
where p(n) is related price according to required
spectrum
End For
For best CfP in K do
Construct a proposal (PUID,s,t,p) and send it to
corresponding secondary user
End For
End If
If received message = accept
Start spectrum sharing with selected secondary user
End If
If received message = reject OR some unused
spectrum
parts are still available
Continue analyzing further
CfPs for spectrum
sharing
Else Set busyflag = true
End If
End While
210
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
should share their spectrum with the corresponding SUs for the
agreed time period. Further, the total number of cooperation
messages (CfP, proposal, accept and reject) generated in the
system, determine the cooperation cost. Thus, the cooperation
strategy that is better (both between T1 and T2) in terms of less
number of messages and, which gives good utility values is
considered as the most cost efficient. The total number of
resources successfully shared (over the number of resources
required) presents the success rate, while the number of non-
allocated spectrum portions (due to disagreements between
primary and secondary user agents) measures the overall
spectrum loss. All the experiments were realized using JADE
[47] on a PC with 3GB memory and 2.4 GHZ dual processor.
B. Results
In Figure 6, we compare the average utility of each primary
and secondary user at T1 with those at T2 for different numbers
of users (10, 20, 30…). The figure depicts that when time limit
is T2, the utilities are a bit less compared to the results obtained
at T1. This is because the environment is mobile and some of the
users are slightly hesitant to share their spectrum for longer
periods. We observe that when there are 10 agents, the average
utility values are almost identical for both T1 and T2, showing
the optimal behavior. But in other cases, the average utility
values are different, showing that the performance of agents in
terms of their average utility values has decreased slightly with
the increased number of agents.
Figure 7 illustrates the spectrum resource requirements and
utilization over time periods T1 and T2. In the beginning (with
10 required resources), all of them are completely shared;
whereas when the required spectrum resources arrives at the
middle values (such as 30 to 40), approximately 90% of them
are shared. This spectrum sharing trend continues following the
same pattern reaching bigger values (such as 50 and 60), with
achieved sum of resources comprised between 45 and 50. Thus,
the performance degradation in terms of spectrum sharing is not
high, even with large resource requirements.
Our approach is also relative to time, because in CR networks
the spectrum holding time is one of the most important factors to
be considered. Again, we run the simulation with several values
of primary and secondary user agents. Figs. 8 and 9 plot the
overall mean times (or holding time), for which the spectrum is
required and utilized for a total of 10 to 120 agents. When time
limit is T1, the results are almost fully satisfied, for 80 to 120
agents, while the time values are somewhat lesser at T2. Both
the results are super linear and coherent with those of Figure 7,
which displays that the spectrum sharing remains high even with
the larger number of agents.
Figure 6. Agents’ percentage utilities
Figure 7. Spectrum resource requirement and utilization by SUs
Figure 10 depicts the maximum number of supported SUs by
the neighboring PUs. Supported SUs are those, which have
completely gained the required spectrum. We observe that when
there are 10 to 15 PUs, the number of supported SUs is literally
the same for both T1 and T2. This means, for limited number of
agents even if the time values are high, the number of supported
SUs is almost the same. However, with large number of agents
(more than 50), the number of supported SUs at T2 are slimly
lesser than T1. Therefore, in ad hoc situations, if we increase the
time values along with an increment in number of agents, the
results will be slightly less optimal.
Figure 8. Spectrum holding time at T1
211
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
Figure 9. Spectrum holding time at T2
Figure 10. Supported SUs
The number of cooperation messages transmitted and received
in the entire system with the success rate (in percentage) is
shown in Figure 11 (and Table II). According to Figure 11, the
values of exchanged messages are almost leveled off for the
middle periods (from 30 to 70 agents). Further, Table II depicts
that the average number of messages (per agent) remains
between 4 to 5 even with the increased number of agents. We
can also see that the approach is linear in terms of messages and
success rate. Particularly when time limit is T2 (around 90 to
120 agents), the performance of the approach substantially
degrades (reaching below 80%), but nevertheless it remains
steady.
Figure 11. Number of messages with success rate
TABLE II. SUCCESS RATE AND NUMBER OF MESSAGES AT
T1 AND T2
Number of
messages Success
rate (in %)
No of
agents T1 T2 T1 T2
10 45 41 100 98.7
20 81 72 90 85
30 117 115 88.23 84
40 159 161 87.31 82
50 185 176 86 82
60 253 261 85 80
70 271 262 84.41 79.3
80 325 366 82 80
90 388 392 82 78.53
100 416 434 81 77.26
110 475 483 80.5 78.77
120 503 516 80 77.42
Another important aspect of our approach is the analysis of
how the performance varies as the amount of participating
agents increases. In this context, Figs. 12 and 13 show the
overall spectrum loss, which is the loss caused by the unused
spectrum, due to spectrum sharing disagreements. As the agents’
demands augment, the percentage of spectrum loss grows on a
steady pace. This is because some of the SUs are not able to find
non-busy PUs or due to the relative change in their
neighborhood. From the figures, it is also clear that the amount
of overall spectrum loss (for both the time limits T1 and T2) is
minimum (10 to 15%), when the number of users are at the
middle stages (i.e., around 50). Spectrum loss then reaches bit
higher values (16 to 22%), with increase number of agents, but
still there is not a rapid degradation in the overall system
performance. Note that the other factors such as collisions,
device level interferences and delays are not considered here.
C. Discussion Related to Results
The above experiments and results prove that our solution is
an effective one in order to provide dynamic spectrum sharing
for CR networks and it can provide better utility of agents with
the exchange of few cooperation messages. However, there are
some important points related to our results, which need further
discussion. First, we assume that the ad hoc environment is
interference free; however, this assumption is not always true. In
reality, the transmission power of most of the devices is so high
that they can easily interrupt the working flow of neighboring
devices, causing interferences. Thus, addressing spectrum
sharing under interference enabled ad hoc networks is still an
issue and several researchers are working on solving this issue to
the modest details [31] [38].
Next issue is related to the limited number of agents we have
used to perform our experiments. Since, JADE only allows a
maximum of 100 to 120 agents on a single machine; therefore
we have only shown the behavior of our approach with limited
number of agents. In order to prove the consistent working of
our model with large number of agents, we are working on
212
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
developing mathematical model based on Markov chain. This
model will also help us to verify other parameters such as
communication cost and agents’ utility. Though, these
mentioned issues need to be addressed in detail, still our model
is flexible enough to replicate the real-world network settings
where spectrum sharing can be performed in the similar
cooperative way.
Figure 12. Spectrum loss at T1
Figure 13. Spectrum loss at T2
VII. CONCLUSION AND FUTURE PERSPECTIVES
In this paper, we developed a cooperative framework for
spectrum allocation that can generate highly effective behavior
in dynamic environments and achieve better utility of the
participating devices. The proposed approach is based on
multiagent system cooperation and implemented by deploying
agents on cognitive radio and primary user devices.
Experimental evaluations confirm the efficiency of our
algorithms for distributed and decentralized environments. The
results show that the proposed approach can absorb the high
spectrum sharing demands by introducing the cooperation
between primary and secondary user devices. Furthermore, the
proposed approach improves the overall utility and minimizes
the spectrum loss with a minimum communication cost. The
spectrum allocation success rate is almost 80% even with large
number of agents. While we only proposed a specific
cooperation strategy to maximize system utility, the proposed
cooperation framework can be extended towards minimizing
other key problems such as inter secondary user interferences
and collisions. We intend to examine this problem as a part of
our continuing work. We are currently working on a
mathematical analysis of our approach using Markov chain. In
addition, the proposed approach assumes that nodes are highly
cooperative while in real systems, nodes can be selfish or
competitive, so more precise work is needed to explore the
competitive behaviors. We will also try to compare the results
with game-theoretical approaches to have an even better
validation of our work.
ACKNOWLEDGEMENT
This work was co-funded by ANR (French Research
Agency) via grant ER502-505E ("Technologies for terminals in
opportunistic radio applications") and by Higher Education
Commission (HEC) Pakistan.
REFERENCES
[1] “Spectrum policy task force report,” ET Docket No. 02–135, (November
2002).
[2] “Unlicensed Operation in the TV Broadcast Bands,” Federal
Communications Commission, First Report and Order and Further Notice
of Proposed Rulemaking. 06-156, October 2006.
[3] B. Canberk, I.F. Akyildiz, and S. Oktug, “Primary user activity modelling
using first-difference filter clustering and correlation in cognitive radio
networks,” Elsevier Science Journal on Ad hoc Networks, vol. 7, pp. 810-
836, 2009.
[4] C. R. Stevenson, C. Cordeiro, E. Sofer, and G. Chouinard, “Functional
requirements for the IEEE 802.22 WRAN standard,” Technical Report,
September 2005.
[5] D. Cabric, S.M. Mishra, and R.W. Brodersen, “Implementation issues in
spectrum sensing for cognitive radios,” Proc. Asilomar Conference on
Signals, Systems and Computers, pp. 772–776, 2004.
[6] D. Niyato and E. Hossain, “Competitive pricing for spectrum sharing in
cognitive radio networks: dynamic game, inefficiency of Nash equilibrium,
and collusion,” IEEE Journal on Selected Areas in Communications, vol.
308, pp. 192-202, 2008.
[7] D. Raychaudhuri and X. Jing, “A spectrum etiquette protocol for efficient
coordination of radio devices in unlicensed bands,” Proc. IEEE
International Symposium on Personal Indoor (PIMRC 03), pp. 172–176,
2003.
[8] E. Jung and X. Liu, “Opportunistic spectrum access in heterogeneous user
environments,” Proc. IEEE New Frontiers in Dynamic Spectrum Access
Networks (DySPAN 08), pp. 1-11, 2008.
[9] F. Akyildiz, W-Y. Lee, M. C. Vuran, and S. Mohanty, “NeXt
generation/dynamic spectrum access/cognitive radio wireless networks: A
survey,” International Journal of Computer and Telecommunications
Networking, vol. 50, pp. 2127-2159. 2006.
[10] F-S. Hsieh, “Developing cooperation mechanism for multi-agent systems
with Petri nets,” Engineering Applications of Artificial Intelligence
Journal, vol. 22, pp. 616-627, 2009.
[11] G. Hosseinabadi, H. Manshaei, and J-P. Hubaux, “Spectrum sharing games
of infrastructure-based cognitive radio networks,” Technical report on
LCA-REPORT-08-027, 2008.
[12] G. Weiss, “A modern approach to distributed artificial intelligence,” MIT
press, 2000, USA.
[13] H. J. Kloeck and F. Jondra, “Multi-agent radio resource allocation,” ACM
Mobile Networks and Applications, vol. 11, pp. 813-824, 2006.
[14] H. M. Kelash, H. M. Faheem, and M. Amoon, “A multiagent system for
distributed systems management,” Transactions on, Engineering,
Computing and Technology, vol. 11, 2006.
213
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
[15] I. Doghri and H.K.-B. Ayed, “Towards fair P2P auctions over MANETs,”
Proc. International Conference on Computer and Information Technology,
pp. 658-663, 2008.
[16] I. Romdhani, M. Kellil, H- Y. Lach, A. Bouabdallah, and H. Bettahar,
“Mobility-aware rendezvous point for mobile multicast sources,” Proc.
International Wired/Wireless Internet Communications conference (WWIC
04), pp. 62-73, 2004.
[17] J. Mitola, “Cognitive radio: an integrated agent architecture for software
defined radio,” PhD Thesis, KTH Royal Institute of Technology, Sweden,
2000.
[18] J. O’Neel, “Analysis and design of cognitive radio networks and
distributed radio resource management algorithms.” PhD Thesis, Virginia
Tech, USA, 2006.
[19] J. Zhang and Q. Zhang, “Stackelberg game for utility-based cooperative
cognitive radio networks,” Proc. ACM International Symposium on Mobile
Ad hoc Networking and Computing (MOBIHOC 09), pp. 23-32, 2009.
[20] K- C. Huang, X. Jing, and D. Raychaudhuri, “MAC protocol adaptation in
cognitive radio networks: an experimental study.” Proc. International
Conference on Computer Communications and Networks (ICCCN 09),
pp.1-6, 2009.
[21] K. P. Sycara, “Multiagent systems,” Artificial Intelligence Magazine, vol.
19, pp. 79-92, 1998.
[22] K.R. Chowdhury, M.D. Felice, and I.F. Akyildiz, “TP-CRAHN: A
transport protocol for cognitive radio ad hoc networks,” Proc. IEEE
Conference on Computer Communications (INFOCOM’ 09), pp. 2482-
2490, 2009.
[23] L. Ma, X. Han, and C-C. Shen, “Dynamic open spectrum sharing MAC
protocol for wireless ad hoc networks,” Proc. IEEE New frontiers Dynamic
Spectrum Access Networks (DySPAN’05), pp. 203-213, 2005.
[24] L. Panait, and S. LukeOn, “Cooperative multi-agent systems learning: state
of the art,” Proc. Autonomous Agents and Multi-Agent Systems
(AAMAS’05), pp. 387–434, 2005.
[25] M. Mchenry, “Spectrum white space measurements,” New America
Foundation Broadband Forum, June 2003.
[26] M. Sawan, H. Yamu, and J. Coulombe, “Wireless smart implants dedicated
to multichannel monitoring and microstimulation,” IEEE Circuits and
Systems Magazine, vol. 5, pp. 21-39, 2005.
[27] M. Wooldridge, “An Introduction to Multiagent Systems,” John Wiley &
Sons Press, 2002, England.
[28] N. Sahai, Hoven, and R. Tandra, “Some fundamental limits in cognitive
radio,” Proc. Allerton Conference on Commonunication, Control and
Computing, 2004.
[29] P.J. Denning and C. Martell, “Coordination,“ Springer Verlag, 1998, USA.
[30] R. G. Smith, “The contract net protocol: High-level communication and
control in a distributed problem solver,” IEEE Transactions on
Computation, vol. 29, pp. 1104–1113, 1980.
[31] S. J. Kim and G. B. Giannakis, “Optimal resource allocation for MIMO ad
hoc cognitive radio networks,” Proc. International Conference on
Communication, Control and Computation, pp. 39-45, 2008.
[32] S. Kumar, V. S. Raghavan, and J. Deng, “Medium access control protocols
for ad hoc wireless networks: a survey,” International Journal of Ad hoc
Networks, vol. 4, pp. 326-358, 2006.
[33] T. C. Clancy, “Dynamic Spectrum Access in Cognitive Radio Networks,”
PhD Thesis, University of Maryland, USA, 2006.
[34] T. Sugawara, T. Hirotsu, S. Kurihara, and K. Fukuda, “Effects of
fluctuation in manager-side controls on contract net protocol in massively
multi-agent systems,” Proc. IEEE International Conference on Distributed
Human-Machine Systems (DHMS’08), 2008.
[35] U. Mir, L. Merghem-Boulahia, and D. Gaïti, “Utilization of a cooperative
multiagent system in the context of cognitive radio networks,” Proc. IEEE
International Workshop on Modelling Autonomic Communications
Environments (MACE’09), pp. 100-104, 2009.
[36] U. Mir, L. Merghem-Boulahia, and D. Gaïti, “A cooperative multiagent
based spectrum sharing”. in Proc. Advanced International Conference on
Telecommunications (AICT’10), pp. 124-130, Barcelona, 2010.
[37] X. Jiang, H. Ivan, and R. Anita, “Cognitive radio resource management
using multi-agent systems,” Proc. Conference on Consumer
Communications and Networking, (CCNC’07), pp. 1123-1127, 2007.
[38] Y. Su and M. Schaar, M, “A new perspective on multi-user power control
games in interference channels,” IEEE Transactions on Wireless
Communications, vol. 8, pp. 2910–2919, 2009.
[39] Z. Ji and K. Liu, “Dynamic spectrum sharing: A game theoretical
overview,” IEEE Communications Magazine, vol. 45, pp. 88-94, 2007.
[40] http://mobiledevdesign.com/tutorials/lte-femtocells-0603/, Feb. 03, 2011
[41] http://www.dvb.org/ , Jan. 12, 2011
[42] http://www.dvb.org/about_dvb/dvb_worldwide/france/, Jan. 31, 2011
[43] http://www.ero.dk/TG4, June. 06, 2010
[44] http://www.itu.int/ITU-R/index.asp?category=conferences&rlink=wrc-
07&lang=en, Sept. 02, 2007
[45] http://ec.europa.eu/information_society/policy/ecomm/radio_spectrum/topi
cs/reorg/pubcons_digdiv_200907/index_en.htm, Sept. 04, 2009
[46] TEROPP,http://era.utt.fr/fr/projets_de_recherche/carnot_teropp.html, Jan.
11, 2011
[47] http://jade.tilab.com/, July. 07, 2010
[48] http://ayman.elsayed.free.fr/msc_student/wlan-tutorial.pdf, June 2002
[49] http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4085653, Feb.
08, 2007
214
International Journal o
n Advances in Telecommunications
, vol
3
no
3
&
4
, year 20
10
,
http://www.iariajournals.org/telecommunications/
2010, © Copyright by authors,
Published
under agreement with
IARIA
-
www.iaria.org
APPENDIX
Abbreviations
ACM agent coordination module, 6
AI artificial intelligence, 5
AKM agent’s knowledge module, 6
APs access points, 2
CfP call for proposal, 6
CNP contract net protocol, 2
CR cognitive radio, 1
DSS dynamic spectrum sensor, 6
DVB-T digital video broadcasting- transmitter, 4
FCC federal communication commission, 1
ISM industrial, scientific and medical, 3
JADE java application development framework, 8
LTE long term evolution, 3
MAC medium access control, 3
MANETs mobile ad hoc networks, 1
MAS multiagent system, 1
PAL phase alternative line, 4
PDA personal digital assistant, 3
PU primary user, 2
PUID primary user’s agent identification, 6
RF radio frequency, 1
SC spectrum characterizer, 6
SU secondary user, 2
SUI secondary user interface, 6
SUID secondary user’s agent identification, 6
TG4 task group 4, 4
UHF ultra high frequency, 4
WRAN wireless regional area network, 1
WLAN wireless local area network, 1
WRC world radiocommunication conference, 4
