A Mobility Scheme for Cognitive Radio Networks Emna Trigui, Moez Esseghir, Leila Merghem-Boulahia

ICD/ERA, CNRS UMR STMR 6279, University of Technology of Troyes, 12, rue Marie Curie, 10004 Troyes Cedex, France

{emna.trigui, moez.esseghir, leila.merghem_boulahia} @utt.fr

Abstract— Cognitive radio network is a promising wireless network where smart devices are able to opportunistically exploit the spectrum holes and optimize the overall radio spectrum use. The provision of quality of service and mobility management is imperative to provi ding efficient cognitive radio systems. In this paper, we present a mobility scheme that enables cognitive radio users to seamlessly switch towards the best available spectrum band when moving from one network to another one. On the one hand, our proposal conserves the frequency and the bandwidth needed for users’ application. On the other hand, it allows cognitive radio users to have the best band with a good price and a large use duration based on a trading mechanism established between primary and cognitive radio users. S imulation results prove that our approach preserves users’ continuity of service during their mobility and ensures high utility for both primary and cognitive radio users.

Keywords—Cognitive radio, wireless networks, mobility, quality of services, continuity of service, trading, multi-agent systems.

I. INTRODUCTION Mobility management traditionally involves two tasks:

location and handoff management. Location management tracks the location of the network nodes, while handoff management is responsible for keep ing end to end connections active when nodes move from one network to another one [1]. In this paper, we will study the handoff management for a cognitive radio terminal while ensuring its application requirements. Cognitive radio main goal is to optimize the radio spectrum assignment by dynamically and efficiently exploit ing the spectrum white spaces [2].

Cognitive rad io (CR) [3] is a new technology that allows terminals to sense the nearby spectrum and utilize detected spectrum holes opportunistically.

CR networks define two types of users: primary and secondary users. Primary users (PU) are licensed users that can access the spectrum thanks to their licenses. However, secondary users (SUs) are unlicensed users (or CR terminals) having CR capabilit ies to opportunistically access the unused spectrum.

User’s mobility p resents an important challenge in CR wireless network (CRWN) [4] since it can significantly affect its performance by causing services interruption or degrading quality of services (QoS) for example.

In this paper, we provide a multi-agent based solution for user’s mobility to alleviate service interruption issue. Especially, we adopt a specified trading mechanis m that we define between SUs and PUs in order to optimize the spectrum attribution during the handoff. We propose hence a real pricing and sharing system between CR networks’ users.

The remainder of the paper is organized as follows. Section II briefly presents some related work. In section III, we expose our mult i-agent systems (MAS) based approach for handoff management in mobile CRWN. Simulation results are given in section IV. Finally, section V concludes the paper.

II. RELATED WORK Studies handling mobility issue in cognitive radio networks

are restricted and most of them dress only some challenges and opportunities for CR networks mobility design without giving an effective solution to the problem [5- 6].

Authors in [5] analyse in detail the main open research issues and propose some ideas for efficient spectrum handoff and mobility management, which include the need to divide and mark location areas , the need to include a connection admission control (CAC) during the handoff process and finally the necessity to optimize the spectrum handoff performance in o rder to minimize delay and packet loss .

The underlined architecture presented in [6] is handling spectrum access on network layer. It is also concerned with the involvement of different CR layers for the control of the connectivity. However, this architecture does not consider the spectrum handoff delay and the service continuity.

Besides, in [4], authors discuss the impact of mobility at different layers of the protocol stack. Through a case study focusing on routing in cognitive Ultra Wide Band (UW B) networks, they prove that any solution for CR network can actually lead to poor performances results if devices’ mobility is not taken into account at the design step. Furthermore, this study confirms that specific solutions for mobile CRW N are still scarce in the literature and most of them target layer 3 (network) [7] and layer 4 (transport) [8]. In this work, we use IEEE 802.11 standard for the physical and MAC Layers, we consider the Internet Protocol (IP) at the network layer and we handle the spectrum management at the application layer.

Some recent studies have addressed user’s mobility when solving cognitive radio issues such as the congestion control [9] and the opportunistic spectrum scheduling [10] without solving the effective handoff problem.

In [10], authors’ studies focus on scheduling the spectrum allocation for mobile CR devices. Authors define the related problem as the Maximum Throughput Channel Scheduling problem (MTCS) based on the mobility information. Then, they propose a general scheduling framework to resolve the MTCS prob lem, which seeks a channel assignment schedule for each user such that the maximum expected throughput can be achieved.

Recently, QoSMos [11] a new European project has been brought to deal with the quality of service and mobility driven

978-1-4799-1004-5/13/$31.00 ©2013 IEEE

2013 12th Annual Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET)

97

cognitive radio systems. The main objective of this project is to provide a framework for cognitive radio systems using test-bed. This project focuses on opportunistic use of spectrum combined with managed QoS and seamless mobility applied to an early example being TV White Spaces in the UHF bands.

Mobility in CRWN is an open research area that needs to be more investigated. Accordingly, we focus in this work on providing an effective handoff solution for CR users.

III. COGNITIVE RADIO MOBILITY SCHEME In this section, we briefly describe the scenario we use, and

then we present our proposed cognitive radio mobility scheme. We consider a set of mobile CR nodes (MCNs) moving

from a zone i towards a zone j through a set of intermediate zones. Zones are considered as ad-hoc or cellular networks and are deployed with a set of primary users (PUs). PUs are operating in different frequency bands . Each zone has its own characteristics such as available frequencies, PUs number and the amount of unused spectrum resources.

The challenge, in this scenario, consists in allowing MCNs to move from one zone to another one seamlessly without causing service interruption while respecting node requirements and new environment conditions.

We deploy agents on MCN and PU nodes, respectively. Whenever an MCN comes close to a new zone (i.e . its Received Signal Strength RSS becomes low), its agent starts collecting informat ion about PUs conditions in the new zone. The MCN broadcasts first a band request message to all PUs. According to PUs replies, MCN agent updates its knowledge base with spectrum and PUs conditions such as PU’s used frequency, the amount of free bands and the proposed band price and use duration. Then, it chooses the suitable band to share through a specified spectrum handoff decision process that we have developed. To ensure efficient handoff decision that takes into account users requirement, our proposal uses a multi-agent trading mechanis m.

In the following, we will depict our algorithm operation dealing with its different processes (Fig. 1).

As shown in Fig.1, the first step performed by the MCN

when it comes close to a new zone is Collecting PUs information. In this step, the MCN broadcasts a band request message to all PUs that are in the new zone in o rder to co llect informat ion about PUs spectrum band conditions. Each PU replies with a band response message in which it specifies the used frequency, the amount of free bands and the proposed price and use duration. Using the collected data about the new zone, the MCN builds its PUs information database.

Secondly, the MCN removes from recorded list of detected PUs the ones that have not enough free bands regarding MCN’s need. If none of the PUs has sufficient bands, the MCN waits for a given time (∆t) and then retries again to access the spectrum. Otherwise, the MCN proceeds to the selection of the most appropriate PU’s spectrum band.

The PU selection Mechanism allows the MCN to select the PU having the most appropriate band according to its requirements.

Once the appropriate PU is selected, an agreement between MCN and PU is established and the PU attributes needed bands to the requesting MCN. Two cases are possible:

The selected PU presents a frequency different from the frequency currently used by the MCN. In this case, the MCN has to perform a Spectrum Handoff, i.e. The MCN changes its frequency and switch to the new frequency of the selected PU band.

The selected PU offers the same frequency the MCN is using. In this case, the MCN has no need to perform a spectrum handoff and it starts the Band Sharing process.

The Band Sharing process allows the MCN to share the band with the selected PU for the agreed use duration. The MCN proceeds sharing the PU’s band until its agreed use duration exp ires or it comes close to a new zone where it starts again collecting PUs in formation. In the following subsection, we present the algorithm of the PU selection mechanis m in detail.

A. PU selection Algorithm Basically, the MCN searches for a PU providing the

frequency it is currently using to avoid unnecessary spectrum handoffs and to conserve session continuity. Unless this PU is found, the MCN preselects the PU that has the lowest price per use duration ratio. This rat io is denoted PPS. PPS stands for Price Per Second as we assume that the use duration is expressed in seconds. However, in both cases, the MCN cannot base its handoff acceptance decision only on the frequency or on the PPS indicators since it may pay a very high price that exceeds its payment capacity. Let Pmax be the maximum acceptable price that the MCN can pay for spectrum use. If the init ial o ffer of the preselected PU (with same frequency or with the best price per use duration ratio) specifies a price higher than Pmax, the MCN initiates the Trading process in order to find with the PU a common agreement for its spectrum access. The trading process is described in detail in the following subsection.

In the rest of the paper, we use the following notations. Free_PU_List: The list of PUs having sufficient free

bands. PUsf: The PU affording the frequency that MCN is

using. Pmax: The maximum price accepted by the MCN. Pinitial: The price in itially proposed by the PU. Dinitial: The duration initially proposed by the PU. PPS: Price per use duration proposed by the PU.

initial

initial

DP

PPS (1)

Fig1.Handoff steps

2013 12th Annual Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET)

98

B. Trading Process Once the trading process is activated, the MCN agent tries to cheapen PU’s price by proposing to pay a new price PNew equal to its Pmax for spectrum use. If the PU refuses its offer, the MCN tries to negotiate a use duration extension and proposes a new duration DNew as given by equation (2)

max

*PP

DD initialinitialNew

(2)

where Pinitial and Dinitial represent the price and the duration initially proposed by the preselected PU, respectively.

The PU ability to concede depends on its willingness to reduce the proposed price or to prolong the spectrum use duration. This negotiation ability is modeled using two parameters: α and β. α represents the threshold for the PU price acceptance and β corresponds to its threshold for the use duration acceptance. In other words, PU accepts a negotiated price at least equal to α mult iplied by its init ial proposed price and a duration at maximum equal to (1+β) multiplied by its initial proposed duration. α and β are parameters that can be specified by the primary user designer. Fig. 2 shows PU and MCN behaviors during the trading process in detail.

Let us consider the following example. During the trading

process the MCN proposes to the PU a new price equal to 3$ (or any payment unit). Assuming that Pinitial=5$ and α=0.7, the PU rejects this new price as it is lower than α* Pinitial. Knowing that Dinitial=170 seconds (or any time unit), the MCN proposes a new duration DNew=283.3s calculated by equation (2). If we suppose that β is equal to 0.8, the PU accepts the new duration 283.3s as DNew is lower than (1+β) * Dinitial, with a price equal to 5$.

In case of disagreement between the MCN and the preselected PU, i.e . failed trading as the preselected PU rejects the new use duration; the MCN starts a new trading process with the fo llowing PU in its sorted list based on PPS ratio. When an agreement is reached, i.e. successful trading as the PU accepts the new negotiated price or use duration, the preselected PU attribute the requesting band to the MCN.

It is worth noting that the negotiation mechanis m that we propose in this paper (Fig. 2) is different from the one proposed in [12] and is performed, if necessary, with any preselected PUs and not only with PUsf.

Besides, our proposal in this paper is conceived for a class of MCNs, which have a payment price constraint Pmax not to be exceeded. However, in [12] we have proposed another selection and negotiation processes for a second class of users, which are requiring specified use duration in addition to the price (i.e . price per second constraint). In the following section, we expose our approach results .

IV. RESULTS Thorough simulat ions have been performed focusing on

the influence of our trading based approach on user’s continuity of service and users’ satisfaction. First, we expose our simulations setup and then we discuss the obtained results.

We perform our tests under OMNETPP simulator [13], which is a discrete event simulat ion tool. We perform the simulation by considering the parameters given in table 1.

TABLE I SIMULATION PARAMETERS

Algorithm: PU Selection BEGIN The MCN searches for a PUsf in the Free_PU_list. If (No PUsf is found) Then PUpreselected PU having the lowest PPS Else PUpreselected PUsf End If //The MCN verifies the Price proposed by the PU preselected

A: If ( Pinit ial(PUpreselected) ≤ Pmax(MCN) ) Then PUselected PUpreselected Else //The MCN starts a Trading process with the PU preselected

Price_Negotiation() If (Price Negotiation succeed) Then trading result= successful Else Duration_Negotiation() If (duration Negotiation succeed) Then trading result successful Else trading result failed End If End If If (trading result== successful) Then PUselected PUpreselected Else //choice of another PU PUselected Next PU having the lowest PPS Go to A End If End If END

Fig. 2.Trading Process

2013 12th Annual Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET)

99

Parameters Values MCN distribution ( ) 5 PU number 100 Size of spectrum band 4 MHz Size of spectrum sub-band 1 MHz MCN speed 10mps MCN Mobility type Linear Simulation time 600s Simulation runs number 10

To evaluate our approach performance, we first study α and β impact on the successful trading rate, then we measure the handoff blocking rate and we analyse both MCNs and PUs utility. Besides, we measure the PUs bands utilization and the frequency handoff rate. Finally, to gauge the efficiency of our proposal, we compare it to the case where no trading is possible with PUs and the MCN refuses an offer with a price higher than its Pmax. This case is referred as PCWT standing for “price consideration without trading”.

The successful trading rate represents the negotiations that are followed by an agreement between the MCN and the PU.

In this subsection, we will study α and β impact by considering all possible combinations. For border cases, when α=1, PU will always reject MCN proposed price as PNew is always lower than α * Pinitial (condition to refuse price). Likewise, when β=0, PU will always reject MCN proposed use duration. So, with α=1and β=0, no trad ing is possible with PU. On the contrary, with α=0 and β=1, PU accepts any proposed price and at most the double of the in itial duration.

In Fig . 3, we plot the successful trading rate as a function of α and β thresholds . PU number is fixed to 100 and the number of MCN is fixed to 120. Fig. 3 shows that we obtain for around 100% of successful trading rate when α is near to 0 and when β comes close to 1. These results are obviously expected. Furthermore, it is important to emphasize that the successful trading rate remains high even with intermediate values of α and β thresholds.

For the rest of the paper, we consider three combinations of

α and β that present reasonable values for trading and we

show their impact on the handoff blocking rate, the spectrum utilisation and on the users’ utility.

The handoff blocking rate represents the percentage of requests that does not lead to band attribution. Lower level of handoff blocking rate reflects the service continuity and the system efficiency.

In Fig. 4, we plot the handoff b locking rate as a function of MCNs number. PU number is fixed to 100 and the number of MCNs varies from 100 to 160. The comparison results exposed in Fig. 4 show that our approach ensures a handoff blocking rate lower than the PCWT one with any corresponding values of α and β. In addition, we prove that we can guarantee low handoff blocking rate which consequently minimizes session interruption during user’s mobility.

One of the most important concerns of the cognitive radio

researches is to maximize the exp loitation of the spectrum bands. Thus, we present in Fig. 5 the spectrum utilizat ion rate as a function of MCNs number.

As shown in Fig. 5, our proposal improves significantly

the spectrum utilization compared to the PCWT approach. Besides, the bands utilization remains very important with three combinations of α and β and the spectrum explo itation still very important when the number of MCNs increases. We observe also that the combination α=0.5 and β=0.6, that results in a high successful trading rate (for around 96%) may reach 97% of bands utilizat ion with 140 MCNs.

Fig. 5 PUs’ bands utilization

Fig. 4 Handoff blocking rate

Fig. 3 Successful trading rate as a function of α and β thresholds

2013 12th Annual Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET)

100

To reveal user’s satisfaction, we have analyzed the utility noted U for primary and CR users . MCN and PU utility are given by equations (3) and (4), respectively.

)()(

)(

min

maxPUp

PUd

PD

MCNU paid

attributed

(3)

Where Ppaid is the price paid by the MCN after allocation agreement and Dattributed is the agreed spectrum use duration. dmax(PU) is the maximum use duration that can be provided to an MCN and pmin(PU)is the minimum price the MCN can pay for spectrum access. pmin and dmax are init ially set and are the same for all PUs.

)())(()(

maxSUd

SUpMaxD

P

PUU

favorite

attributed

paid (4)

Where Max(pmax(SU)) is the maximum acceptable price that any MCN can pay for spectrum use and dfavorite(SU) is the spectrum use duration that an MCN prefers to have. In this scenario, we consider that all MCNs have the same dfavorite. U(PU) is highly dependent on the price paid by the MCN divided by the assigned spectrum use duration. Whereas U(MCN) is inversely proportional to the price it pays per the agreed use duration. Fig 6 and Fig 7 plot MCNs and PUs utility as function of MCNs number, respectively.

Fig. 6 shows that our solution achieves almost the same utility results as the PCWT approach. This proves the

efficiency of our contribution as in the PCWT approach, the MCN accepts only PU’s price lower than its Pmax resulting in the best utility for MCNs. Likewise, Fig. 7 clears that our approach ensures higher utility for PUs.

Furthermore, results show that α and β thresholds have a significant impact on both handoff blocking and spectrum utilizat ion rate but not on user’s utility. Th is proves that our approach can provide good utility for MCNs with α and β values that give interesting utility for PUs and that minimize the handoff blocking rate and maximize the spectrum bands utilisation at the same time.

Another important metric characterizing the spectrum handoff performances in mobile CR networks is the frequency handoff rate that reflects the average number of spectrum handoffs when the MCN changes from one zone to another one. We compare the frequency handoff rate of our approach with the one of the algorithm NoF (No Frequency consideration) that does not consider the PU having the same frequency band as the MCN in the selection process. Fig. 8 presents the comparison results.

Fig. 8 proves that our approach decreases to the half the

frequency handoff rate when the MCN moves from one network to another one. Consequently, we confirm that our approach improves the handoff performances in a cognitive radio network and preserves the continuity of service for mobile nodes.

To summarize, our trading based approach outperforms significantly the PCWT one in terms of handoff blocking rate and bands exploitation and guarantees interesting utility for both primary and secondary users.

V. CONCLUSION In this paper, we presented a new scheme for mobility

management in CR network. Our proposal takes into account users preferences and spectrum conditions. Simulat ion results show that using the proposed trading mechanism improves overall PUs and MCNs utility and preserves service continuity.

In our future work, we will carry out studies handling the handoff delay and the impact of the speed and the mobility model of cognitive radio users on the system performance.

Fig. 8 Frequency Handoff rate

Fig. 7 PUs’ utility

Fig. 6 MCNs’ utility

2013 12th Annual Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET)

101

The deployment of a learning algorithm throughout the MCNs seems to be also an interesting extension to improve the PU selection process, by promoting, for example those with whom agreements have been concluded.

ACKNOWLEDGMENT This work is partly supported by the Ministry of Higher

Education and Research of France.

REFERENCES [1] I.F. Akyildiz, J. Xie a,nd S. Mohanty, “ A survey of mobility

management in next-generation all-IP-based wireless systems”, IEEE Wireless Communication, 2004, pp. 16–28.

[2] I.F. Akyildiz, W-Y. Lee, M.C. Vuran and S. Mohanty, “Next generation/dynamic spectrum access/cognitive radio wireless networks: a survey”, Computer Networks, vol.50, n°13, pp. 2127–2159, 2006.

[3] J. Mitola, “Cognitive radio architecture : The Engineering Foundations of Radio XML Link”, John Wiley and Sons, 2006.

[4] L. De Nardis and M.-D.P Guirao, “Mobility aware design of cognitive radio networks: challenges and opportunities”, Cognitive Radio Oriented Wireless Network and opportuinities, 2010, pp. 1-5.

[5] X. Fu, W. Zhou, J. Xu, J. Song, “Extended mobility management challenges over cellular networks combined with cognitive radio by using multi-hop network”, 8th ACIS international Conference on software engineering, artificial intelligence, networking and parallel/distributed computing, 2007, pp. 683–688.

[6] Damljanovic, Z., “Mobility management strategies in heterogeneous cognitive radio networks”, Journal of Network and Systems Management, vol. 18, 2010, pp. 4-22.

[7] K. Chowdhury and M. Felice, “SEARCH: A Routing Protocol for Mobile Cognitive Radio Ad-Hoc Networks,” Computer Communications, vol. 32, no. 18, pp. 1983 –1997, 2009.

[8] K.R Chowdhury, M. Di Felice, and I.F. Akyildiz, “Tp-crahn: a transport protocol for cognitive radio ad-hoc networks,” IEEE INFOCOM 2009, pp. 2482–2490.

[9] A. Al-Dulaimi, S. Al-Rubaye and J. Cosmas, “Adaptive congestion control for mobility in cognitive radio networks”, Wireless Advanced, 2011, pp. 273-277.

[10] L. Zhang, K. Zeng, P. Mohapatra, “Opportunistic spectrum scheduling for mobile cognitive radio networks in white space”, IEEE Wireless Communications and Networking Conerence 2011, pp. 844-849.

[11] I Karla, J. Bitó, L. Csurgai-Horváth, B. Bochow, U. Celentano, P. Grønsund, M. López-Benítez, R. Samano-Robles, “Cognitive Spectrum Portfolio Optimisation, Approaches and Exploitation”, 19th European Wireless Conference’13, Guildford, UK Mars 2013.

[12] E. Trigui, M. Esseghir and L. Merghem boulahia, “ Multi-Agent Systems Negotiation Approach for Handoff in Mobile Cognitive Radio Networks”, IEEE International Conference on New Technologies, Mobility abd Security (NTMS’12), Istanbul, 7- 10 Mai 2012, pp 1-5.

[13] G. Pongor, “OMNeT: Objective Modular network Testbed”, International Workshop on Medeling, Analysis and Simulation On Computer and Telecommunication Systems 1993, pp. 323-326.

2013 12th Annual Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET)

102