Performance Analysis of Channel Switching with Various Bandwidths in Cognitive Radio

Size: px

Start display at page:

Download "Performance Analysis of Channel Switching with Various Bandwidths in Cognitive Radio"

Dina Doyle
5 years ago
Views:

1 Performace Aalysis of Chael Switchig with Various Badwidths i Cogitive Radio Po-Hao Chag, Keg-Fu Chag, Yu-Che Che, ad Li-Kai Ye Departmet of Electrical Egieerig, Natioal Dog Hwa Uiversity, 1,Sec.2, Da-Hsueh Rd., Shou-Feg, Hualie, Taiwa, R.O.C. po@mail.dhu.edu.tw, m @ems.dhu.edu.tw, m @ems.dhu.edu.tw, d @ems.dhu.edu.tw Abstract Dyamic spectrum access is a key techique i cogitive radio. Wheever primary users appear, secodary users must evacuate the primary chael rapidly, ad switch the appropriate chael to primary users. There are two types of chael access methods, amely reactive ad proactive chael access. I reactive method, cogitive radio does ot eed to switch chael util primary users appear, while i proactive method, secodary users predict future chael traffic by usig chael history ad switch chael before primary users appear. Most of the previous researches assume that all primary users possess the same chael badwidth. I this paper, we take various chael badwidths ito cosideratio to make spectrum hadoff decisio with the method proposed before uder real cogitive radio eviromet, ad use this method to aalyze the performace. Chael utilizatio rate i both methods is ehaced while cosiderig various badwidths. Keywords- reactive chael access; proactive chael access; Badwidth. I. INTRODUCTION Spectrum is a precious ad limited atural resource, ad most portios of it have bee authorized. With the developmet of etwork techiques, the demad for spectrum is icreasig. The biggest problem of spectrum applicatio is ot scarcity but ieffective, accordig to [1]. Not oly do we have to solve the ieffective problem, but we also eed to make good use of the spectrum. With "spectrum sesig capability, cogitive radio techology has bee proposed [2], hopig that through time ad space cofiguratio, full use of those uused spectrum resources ca effectively solve the problem of the spectrum cogestio or the uequal distributio. The first step of developig cogitive radio is dyamic spectrum access [3]. Cogitive radio users must moitor the idle spectrum periodically, aalyze the surroudig wireless messages at the same time to adapt themselves to the eviromet, ad use these messages to lear ad adjust the trasmissio state ad parameters i radio. As soo as the primary users appear, cogitive radio users ca seamlessly switch to other idle chaels, makig trasmissio cotiue i spite of the appearace of the primary users. Liu, et al. [4] discusses the cocept of spectrum mobility, ad develops the probability models of spectrum holes ad spectrum hadoff accordig to the characteristics whe primary users appear. The probability of hadoff is a importat idicator to affect spectrum mobility, ad it is also a importat part of spectrum maagemet. There are two types of chael access methods, amely proactive chael access ad reactive chael access. It is still uclear uder what coditio we shall use which of these hadoff methods. Wag, et al. [5] itroduces these two methods, ad uses a PRP M/G/1 model to derive the formula to aalyze which of the two spectrum hadoff methods ca achieve the best efficiecy with the variatio of spectrum sesig time. Without iterferece to primary users, secodary users are allowed to temporarily use the chael for trasmissio. Aravida, et al. [6] uses proactive chael access method without iterferece uder TV broadcast coditio. Secodary users ca use the iformatio of chael history to build a predictio model, ad use it to make the spectrum hadoff decisio. Not oly ca it ehace the chael utilizatio, but it ca also reduce the iterferig time produced by primary users. Höyhtyä, et al. [7] proposes a simple method to classify the chaels to either periodic or stochastic patters, which may i tur help the secodary users to schedule the chael. Yag, et al. [8] sets the period whe primary users appear as a alterative expoetial ON/OFF model, ad predicts future primary chael state accordig to previous primary users traffic. Secodary users ca try ot to iterfere with the primary users by quickly switchig to aother uused primary chael to cotiue its trasmissio. Xue, et al. [9] emphasizes the importace of hadoff delay i real life so that the secodary users ca take it ito cosideratio to make the spectrum hadoff decisio. Most of the previous researches assume that all primary users possess the same chael badwidth. I reality, however, cogitive radio eeds to adapt to the heterogeeous etwork architecture i which ot all primary users possess the same chael badwidths. Accordig to IEEE stadard [10], cogitive radio is applied for 54 ~ 862 MHz UHF/VHF TV bads ad must adapt to 6MHz, 7MHz, ad 8MHz TV bads. This will lead us to make differet spectrum hadoff decisio. Hece, we take various chael badwidths ito cosideratio to make spectrum hadoff decisio with the method proposed before uder real cogitive radio eviromet, ad use this method to aalyze the performace. The rest of the paper is orgaized as follows: I Sectio II, we itroduce the system model. I Sectio III, we discuss the strategy of hadoff decisio. I Sectio IV, we show our simulatio results. I Sectio V, we provide our coclusio. 29

II. SYSTEM MODEL A. System predictio model At the begiig of chael predictio model show i Fig. 1, secodary users will perform spectrum sesig, ad the save the sesig iformatio ito chael history database.

If it eeds spectrum hadoff, the chaels idle time i the ext slot subtracts chael switch cost ad spectrum sesig time to get the chaels remaiig time.

Fially, i the hadoff decisio sectio, cogitive radio will look for the chael that has the maximum packet capacity to be the hadoff chael, ad use it to trasmit data.

At the period of spectrum sesig, cogitive radio will use the spectrum sesig result ad the iformatio from chael history database to predict the probability of chael idle time i ext slot. Figure 3.

2 II. SYSTEM MODEL A. System predictio model At the begiig of chael predictio model show i Fig. 1, secodary users will perform spectrum sesig, ad the save the sesig iformatio ito chael history database. Primary users chael s traffic iformatio ca be obtaied from the historical database. Cogitive radio ca use spectrum sesig results ad chael history to predict chaels idle time i the ext slot. If it eeds spectrum hadoff, the chaels idle time i the ext slot subtracts chael switch cost ad spectrum sesig time to get the chaels remaiig time. The packet capacity ca be derived from the product of chaels remaiig time ad the chael badwidth. Fially, i the hadoff decisio sectio, cogitive radio will look for the chael that has the maximum packet capacity to be the hadoff chael, ad use it to trasmit data. I our simulatio eviromet, we assume that cogitive radio is operated i a slotted model, ad has N primary chaels for access as show i Fig. 3. At the period of spectrum sesig, cogitive radio will use the spectrum sesig result ad the iformatio from chael history database to predict the probability of chael idle time i ext slot. Figure 3. The slotted structure of the secodary user Figure 1. Flowchart of proactive spectrum access B. Chael model We use the commo alterative expoetial ON-OFF model to be our chael model as show i Fig. 2. The differece with previous research is that our badwidth is variable. At first, all primary chaels will be radom i ON or OFF state. Whe the chael is i the ON state, which meas that the chael is BUSY, the secodary users caot access the chael; whe the chael is i the OFF state, which meas that the chael is IDLE, the secodary users ca access the chael for trasmissio. The duratios that the primary user passage shows ON (BUSY) ad OFF (IDLE) are idepedetly expoetially distributed. For chael, the period of ON, B, follows a expoetial distributio with mea 1/λ B. O the other had, the period of OFF, I, also follows a expoetial distributio with mea 1/λ I. III. THE STRATEGY OF HANDOFF DECISION A. Reactive v.s. Proactive chael access Spectrum hadoff occurs whe the primary user appears i the licesed chael that is temporarily used by the secodary users. The mai sigificace of the spectrum hadoff is to help the secodary users switch to the suitable idle chaels to resume trasmissios. The types of spectrum hadoff are divided ito two types: Reactive chael access: Wheever a primary user occurs, secodary users have to hadoff by followig steps: first, after detectig ay primary user, secodary users iterrupt trasmissio ad do spectrum sesig. Secod, accordig to the real-time spectrum sesig, fid idle chaels ad switch to the idle chael to resume trasmissios. Not oly does it have to sped real time spectrum sesig, some real time applicatios (for example, watchig movies olie) will also cause great harm. This is show i Fig. 4. Figure 2. The alterative expoetial ON/OFF chael model f(b )= λ B e B λ B, B 0 0, B < 0 (1) Figure 4. The reactive chael access model Proactive chael access: Secodary users use chael history to predict future chael traffic, ad schedule the 30

3 chael access itelligetly. They may use the predicable method with the result of spectrum sesig to predict future chael traffic before primary users appear i order to avoid disruptio by primary users, ad to maitai reliable commuicatio. The advatage of usig proactive chael access is the reductio of commuicatio disruptio by primary user, for secodary users ca switch to other chaels before primary users appear ad resume trasmissio faster. This is show i Fig. 5. CH = arg max [ (T IDLE ΓT switch )] (4) 1, if CH(Prev) CH(Curret) Γ = 0, else D. Badwidth cosideratio Whe we make a switch decisio i the process of proactive chael access, we take the chael badwidth ito cosideratio to achieve the best QoS. PACKET = T remai * BW (5) T remai = T IDLE - α ( Tswitch Tsese ) (6) 1, if CH(Prev) CH(Curret) α = 0, else CH=arg max ( PACKET ) (7) Figure 5. The proactive chael access model B. Chael predictio We assume that the primary chaels are a series of ON/OFF model; whe the chael is i ON state, the chael is beig used; whe the chael is i OFF state, the chael is IDLE ad secodary users ca access this chael. The frequecy of primary users' appearace ca be built i a alterative expoet ON/OFF model. The average chael IDLE time ca be set as E[T IDLE ]= 1 λ, ad the average I chael IDLE time ca be set as E[T ON ]= 1 λ. We ca use B the built model to predict future chael traffic. We assume that the idle probability of chael N i ext slot is P which ca be derived based o reewal theory [11]: PACKET meas the packet capacity of the chael which performs trasmissio oce ad is equal to the product of chael remaiig idle time ad chael badwidth. T remai meas the chael remaiig idle time, ad BW meas chael badwidth. If it eeds to do chael switch, chael idle time has to subtract switch cost ad spectrum sesig to calculate the chael remaiig idle time. Fially, we will fid the chael which has the maximum packet capacity to be our spectrum switch chael. The flowchart of this proactive chael access model is show i Fig. 6. P = λ B λ B + λ I e (λ B +λ I ) t, s = 0 λ B e (λ B +λ I ) t, s = 1 (2) We ca use the followig formula to predict the chael idle time i ext slot: T IDLE = P λ (3) I C. Itelligece chael switchig Whe we make the spectrum switch chael decisio, we cosider [9] which metioed switchig cost to achieve more realistic simulatio. I real life, spectrum switch decisio makig must cosider more situatios such as packet-loss-ratio, sychroizatio ad delay. It must produce o-egligible hadoff delay. The hadoff delay eeds to be take ito cosideratio, ad it must affect the hadoff decisio as follows: Figure 6. The flowchart of proactive chael access model The available amout of the badwidth is the key to determie the chael switch. Whe the remaiig chael idle time is the same, the chael with bigger badwidth possesses higher data trasmissio rate so that secodary users ca fiish their trasmissio faster. Example: I Fig. 7, the chael remaiig idle time of CH Y is larger tha that of CH X. It is the best choice to switch to CH X accordig to covetioal chael switch method which 31

Selectig CH X ot oly reduces the cost of chael switch ad the slots eeded for further spectrum sesig so that more slots ca be used for data trasmissio, but also requires less time to fiish trasmissio.

8, cogitive radio will first select a chael which is i IDLE state with maximum badwidth to be the First chael, ad start data trasmissio for 180ms.

4 does ot cosider the chael badwidth. After takig it ito cosideratio, eve if the chael remaiig idle time of CH Y is larger tha that of CH X, CH X ca achieve faster data trasmissio sice the chael badwidth of CH X is twice that of CH Y. Selectig CH X ot oly reduces the cost of chael switch ad the slots eeded for further spectrum sesig so that more slots ca be used for data trasmissio, but also requires less time to fiish trasmissio. Figure 7. Chael selectio with badwidth cosideratio As for the process of reactive chael access show i Fig. 8, cogitive radio will first select a chael which is i IDLE state with maximum badwidth to be the First chael, ad start data trasmissio for 180ms. After fiishig data trasmissio for 180ms, cogitive radio will perform spectrum sesig for 20ms periodically. Oce primary user appears, cogitive radio will switch to the chael which has maximum badwidth to resume commuicatio if idle chaels exist; otherwise, cogitive radio will keep spectrum sesig util the first idle chael appears. IV. SIMULATION RESULTS We take various chael badwidths ito cosideratio to make spectrum hadoff decisio with the method proposed before uder real cogitive radio eviromet, ad use this method to aalyze the performace. I our simulatio eviromet, there are a secodary user ad te primary users. The badwidths of te primary chaels are uiformly distributed from 1 to 10 MHz ad the mea idle ad busy periods are also uiformly distributed i mi=0.5 ad max=0.6~2.0. Sesig period ad switch cost is 20ms. The period of data trasmissio is 180ms. Simulatio time is set to sec. Fig. 9 shows the umber of chael switch i proactive ad reactive methods. Note that proactive chael access method will process chael switch before primary users appear i order ot to iterfere with the primary users. O the other had, reactive method just switch after primary users appear. Therefore, the umber of chael switch i proactive method will be larger tha that i reactive method. Fig. 10 illustrates the umber of iterruptio by primary users i these two methods. We fid that the umber of iterruptio by primary users i proactive method is always smaller tha that i reactive method sice proactive method predicts future primary chael state accordig to previous primary users traffic. Chael utilizatio rate i both methods is show i Fig.11. Istead of choosig the logest trasmissio time, proactive method always chooses the chael with maximum data trasmissio rate to switch. Comparig with reactive method which switches to the first chael with the idle state, the chael utilizatio rate of proactive method is lower tha that of reactive method. Proactive method ca derive chael remaiig time accordig to chael predictio model, ad use the product of chael remaiig idle time ad the chael badwidth to switch to chael with maximum data trasmissio rate, while i reactive method, cogitive radio will switch to the chael which has the maximum badwidth without cosiderig the probability of appearace of primary users, as show i Fig. 12. Figure 8. The flowchart of reactive chael access model Figure 9. Number of chael switches 32

5 badwidths ito cosideratio for that purpose. Uder these circumstaces, the rule of chael switch should be chaged i order to fid the chael which meets the requiremets of the users. By comparig the Proactive with the Reactive methods, we coclude that the former has better performace i terms of total trasmissio data rate ad the umber of disruptios by primary users, while the latter ejoys less chael switches ad higher chael utilizatio rate. Figure 10. Number of disruptios by primary users Figure 11. Chael utilizatio rate Figure 12. Total data trasmissio rate V. CONCLUSION Previous methods of chael switch just apply to chaels with the same badwidth. I this paper, we take differet REFERENCES [1] Federal Commuicatios Commissio, Spectrum policy task force report, FCC , Nov [2] Simo Hayki, Cogitive Radio: Brai-Empowered Wireless Commuicatios IEEE Joural o Selected Areas i Commuicatios, vol.23, o.2, pp , Feb [3] Z. Tabakovic, S. Grgic, ad M. Grgic, Dyamic spectrum access i cogitive radio, ELMAR, ELMAR '09. Iteratioal Symposium, pp , Sept , [4] Hog-jie Liu, Zhog-xu Wag, Shu-fag Li, ad Mi Yi, Study o the Performace of Spectrum Mobility i Cogitive Wireless Network IEEE Sigapore Iteratioal Coferece o Commuicatio Systems, Nov , [5] Li-Chu Wag ad Chug-Wei Wag, Spectrum Hadoff for Cogitive Radio Networks: Reactive-Sesig or Proactive-Sesig? Performace, Computig ad Commuicatios Coferece, IPCCC IEEE Iteratioal, vol., o., pp , 7-9 Dec [6] Prashath Aravida, Kumar Acharya, Sumit Sigh, ad Haitao Zheg, Reliable Ope Spectrum Commuicatios Through Proactive Spectrum Access First Iteratioal Workshop o Techology ad Policy for Accessig Spectrum, August 5, 2006, Bosto, MA, Uited States. [7] Marko Höyhtyä, Sofie Polli, ad Aare Mämmelä, Performace Improvemet with Predictive Chael Selectio for Cogitive Radios First Iteratioal Workshop o Cogitive Radio ad Advaced Spectrum Maagemet, 2008 (CogART 2008), pp.1-5, Feb. 14, [8] Lei Yag, Lili Cao, ad Haitao Zheg, Proactive Cael Access i Dyamic Spectrum Networks, 2d Iteratioal Coferece o Cogitive Radio Orieted Wireless Networks ad Commuicatios, 2007 (CrowCom 2007), pp , Aug. 1-3, [9] Xue Feg, Qu Daimig, Zhu Guagxi, ad Li Yachu, Smart Chael Switchig i Cogitive Radio Networks 2d Iteratioal Cogress o Image ad Sigal Processig, 2009 (CISP '09), pp.1-5, Oct , [10] C. Cordeiro, K. Challapali, D. Birru, ad N. Sai Shakar, IEEE : the first worldwide wireless stadard based o cogitive radios, First IEEE Iteratioal Symposium o New Frotiers i Dyamic Spectrum Access Networks, 2005 (DySPAN 2005), pp , Nov. 8-11, [11] Cox, D. Reewal Theory, Spottiswoode Ballatye ad Co. Ltd,

Radar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1

Radar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1 Advaces i Egieerig Research, volume 8 d Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 7) Radar emitter recogitio method based o AdaBoost ad decisio tree Tag Xiaojig,