Intelligent Call Admission Control Using Fuzzy Logic in Wireless Networks

Transcription

1 25 America Cotrol Coferece Jue 8-1, 25. Portlad, OR, USA FrA18.6 Itelliget Call Admissio Cotrol Usig Fuzzy Logic i Wireless Networks Yufeg Ma, Xiuli Hu, Yuyu Zag, ad Yimei Si Abstract Scarcity of te spectrum resource ad mobility of users make Quality-of-Service (QoS) provisio a critical issue i wireless etworks. Tis paper presets a fuzzy call admissio cotrol sceme to meet te requiremet of te QoS. It searces automatically te optimal umber of te guard caels i a base statio to make a effective use of resource ad guaratee te QoS provisio. Simulatio compares te proposed fuzzy sceme wit a adaptive cael reservatio sceme. Simulatio results sow tat fuzzy sceme as a better robust performace i terms of call droppig probability, call blockig probability, ad cael utilizatio. I. INTRODUCTION IN te preset ad te ext geeratio wireless etworks, cellular system is still a maor part i telecommuicatio ifrastructure. Cellular system exploits frequecy reuse to acieve ig capacity by limitig te coverage of eac base statio witi a small geograpic area called a cell. We te users move from oe cell to aoter, adoff operatio will occur. Mobile users may cage cells may times durig te lifetime of teir coectios. Specially, i future micro/pico-cellular arcitecture, adoff operatio may occur more frequetly ta i preset macro-cellular arcitecture. Sice te user s itierary ad te availability of resources i various cells is ot kow i advace, it makes QoS provisio a critical issue i cellular radio systems. Call Admissio Cotrol (CAC) is oe of te importat Mauscript received September 14, 24. Yufeg Ma is wit te Departmet of Electroic ad Iformatio, Huazog Uiversity of Sciece ad Tecology, Wua, Postal Code 4374, Cia (Telepoe: (86) ; 594@ 163.com). Xiuli Hu is wit te Departmet of Electroic ad Iformatio, Huazog Uiversity of Sciece ad Tecology, Wua, Postal Code 4374, Cia ( xli@ public.w.b.c). Yuyu Zag is wit te Departmet of Electroic ad Iformatio, Huazog Uiversity of Sciece ad Tecology, Wua, Postal Code 4374, Cia ( xli@ public.w.b.c). Yimei Si is i 188 Cuiu Road, 393 Yisasi Garde, Sogiag District, Sagai, Postal Code 216, Cia ( siyimei28@ 163.com). mecaisms i guaratyig te QoS. It ca be defied as te procedure of decidig weter or ot to accept a ew coectio. If te etwork ca ot meet te coectio eed, te coectio request will be deied. I cellular wireless etworks, oe importat parameter of te QoS is call blockig probability (CBP), wic idicates te likeliood of te ew coectio beig deied. Te oter importat parameter is call droppig probability (CDP), wic expresses te likeliood of te existig coectio beig deied durig adoff process due to isufficiet resource i target cell. From te user s poit of view, avig a call abruptly termiated i te duratio of te coectio is more aoyig ta beig blocked occasioally o a ew call attempt. It is acceptable to give iger priority to adoff call. Te metods for prioritizig adoff are te guard cael ad queuig of adoff requests. Wit te guard cael metod, a fractio of te total available caels i a cell is reserved exclusively for adoff requests from ogoig calls wic may be aded off ito te cell. Te fixed reservatio strategy wastes valuable spectrum resource wile te dyamic reservatio strategy ca offer efficiet spectrum utilizatio by miimizig umber of te required guard caels. I te preset, tere are may scemes about call admissio cotrol aimig at keepig te CDP ad CBP low wile maximizig te resource utilizatio to meet te system demad at te same time [1-6]. We te CDP decreases, te CBP will icrease accordigly. So, it is ard to guaratee te miimum CDP ad te miimum CBP at te same time. Miimizig te CDP is oe of te mai goals of QoS provisioig i wireless etworks. Te mai cotributio of tis paper is to propose a itelliget call admissio cotrol sceme based o fuzzy logic for cellular wireless etworks. Te sceme is built upo te cocept of te guard caels. We desig a fuzzy cotroller i terms of te importat QoS target. Te sceme adusts te umber of te guard caels to its optimum i time accordig to te CDP ad curret umber of te guard cael. It tries to make a effective use of resource ad keep te CDP ad CBP low at te same time. Te remaider of tis paper is orgaized as follows. Sectio 2 aalyzes te metod ad model for call admissio cotrol. Sectio 3 gives te details of te proposed fuzzy call /5/$ AACC 3981

2 admissio cotrol sceme. Sectio 4 rus simulatio to compare te fuzzy sceme wit a adaptive cael reservatio sceme. Te, it discusses te simulatio results. Fially, sectio 5 gives coclusio of te paper. II. CAC METHOD AND MODEL ANALYSIS Te total umber of available caels (deoted by C) i a cell cosists of two parts. Oe part (deoted by C ) is reserved exclusively for adoff calls. Te remaiig C- C caels are sared by bot ew calls ad adoff calls. A ew call will be admitted ito te etwork if te umber of busy caels i te cell is less ta C- C we te call is origiated. A adoff request will be admitted if te umber of busy caels i te target cell is less ta C. We assume tat te arrivals of ew call ad adoff call are idepedet Poisso processes. Te ew call arrival rate is ad te adoff call arrival rate is. Cael oldig times are assumed to follow a egative expoetial distributio wit mea 1/. For a cell capacity of C, queuig model is cosidered as M/M/C i wic te C available caels i te cell represeted as C servers. Te system state space is a fiite set E={,1,2,,C}. Te state trasitio diagram is sow i Fig Fig.1. State-trasitio diagram Let P represet te steady-state probability tat te base statio is i state. Durig te aalysis of birt-deat process, te probability P ca be obtaied like tat [1,4]: P CC k C k k k! kcc 1 1 CC kcc ( ) ( ) k (1) k! ( ) P C C,1! P CC CC ( ) P, C C! 1 C Te ew call will be blocked if te umber of busy caels is more ta C- C. Hece C CBP= P. (3) CC 2 (C-C) (C-C+1) C Te adoff request will be deied if te umber of busy caels is equal to C. Tus CDP=P C. (4) 1 C-C C-C +1 C (2) III. FUZZY LOGIC CALL ADMISSION CONTROL SCHEME I te guard cael metod, te umber of te guard caels is importat to te performace of wireless etwork. It affects te QoS ad resource utilizatio of te etwork. Te proposed fuzzy sceme tries to adust dyamically te umber of guard caels to its optimum to meet te requiremets of te QoS ad resource utilizatio. A. Structure of Fuzzy Logic Cotroller (FLC) Te cocept of fuzzy set is a extesio of classical set. For a classical set X, a elemet may belog to set X or ot. But for a fuzzy set, a elemet is related to a set by a membersip fuctio. Te membersip fuctio usually take o a value betwee ad 1. A FLC ca provide algoritms wic covert te liguistic cotrol strategies based o ituitio, euristic learigs ad export kowledge ito a automatic cotrol strategy. Te FLC is made of fuzzifier, iferece egie, Fuzzy Rule Base ad defuzzifier. Te structure of FLC is sow i Fig.2. Tis paper desigs a FLC of two iput parameters ad oe output parameter based o Mamdai fuzzy model. Te iput liguistic parameters of te FLC are set as call droppig probability (CDP) ad umber of te guard caels (C ). Te output liguistic parameter is set as te tuig umber of te guard caels ( C ). call droppig probability umber of te guard caels Fuzzifier Iferece Egie Fuzzy Rule Base B. Membersip Fuctios Fig.2. FLC structure Defuzzifier tuig umber of te guard caels We coose tresolds for CDP ad umber of te guard caels as.1 ad 12% of te total umber of caels i a cell respectively. Te tuig umber of te guard caels is cose to vary witi te rage from 12% to +12% of te total cael capacity of a cell. Te term sets of CDP, C, ad C are defied as follows: T (CDP)={Z, VS, S, M, B, VB} T (C )={VS, S, M, B, VB} T ( C )={NB, NM, NS, NVS, Z, PVS, PS, PM, PB} We coose triagular fuctios as membersip fuctios because tey are simple ad practical. Te membersip fuctios for iput ad output liguistic parameters are sow i Fig.3, Fig.4, ad Fig

3 1 Z VS S M B VB Small, ad umber of te guard caels is Very Small, te it triggers te 11 t rule ad makes tuig umber of te guard caels Positive Small. Tus, fuzzy cotroller ca compute te tuig umber of te guard caels accordig to te CDP ad curret umber of te guard caels. Te fuzzified output parameter ca be coverted to a crisp value by te maximum membersip iferece metod Z: Zero VS: Very Small S: Small M: Middle B: Big VB: Very Big Fig.3. Membersip fuctios for call droppig probability 1 VS S M B VB 3C 6C 9C VS: Very Small S: Small M: Middle B: Big VB: Very Big 12C 1 Fig.4. Membersip fuctios for umber of te guard caels 1 NB NM NS NVS Z PVS PS PM PB Table 1 Fuzzy cotrol rules Rule Number IF Call droppig probability AND Number of te guard caels THEN Tuig umber of te guard caels R1 Zero Very Small Zero R2 Zero Small Negative Very Small R3 Zero Middle Negative Small R4 Zero Big Negative Middle R5 Zero Very Big Negative Big R6 Very Small Very Small Zero R7 Very Small Small Zero R8 Very Small Middle Zero R9 Very Small Big Zero R1 Very Small Very Big Zero R11 Small Very Small Positive Small R12 Small Small Positive Very Small R13 Small Middle Zero R14 Small Big Zero R15 Small Very Big Zero R16 Middle Very Small Positive Middle R17 Middle Small Positive Small R18 Middle Middle Positive Very Small R19 Middle Big Positive Very Small R2 Middle Very Big Zero R21 Big Very Small Positive Big R22 Big Small Positive Middle R23 Big Middle Positive Small R24 Big Big Positive Very Small R25 Big Very Big Zero R26 Very Big Very Small Positive Big R27 Very Big Small Positive Middle R28 Very Big Middle Positive Small R29 Very Big Big Positive Very Small R3 Very Big Very Big Zero 12C 9C 6C 3C 3C 6C 9C 12C NB: Negative Big NM: Negative Middle NS: Negative Small NVS: Negative Very Small Z: Zero PVS: Positive Very Small PS: Positive Small PM: Positive Middle PB: Positive Big Fig.5. Membersip fuctios for tuig umber of te guard caels C. Fuzzy Rule Base Te Fuzzy Rule Base cosists a series of 3 fuzzy rules, sow i Table 1. Te cotrol rules ave te followig form: IF coditios, THEN actio. For example, if te CDP is IV. SIMULATION A. Simulatio Parameters I order to evaluate te performace of our fuzzy sceme, we implemet ad simulate a adaptive cael reservatio sceme [3] for compariso. To fairly cotrast our sceme to te adaptive algoritm, we used te traffic model ad parameters give i [3]. We assume total cael capacity of a cell is 5. We also assume tat te arrival processes of ew call ad adoff call are Poisso wit mea arrival rates of ad respectively. Cael oldig times of bot types of calls are assumed to follow a egative expoetial distributio wit mea 1/. I te simulatio, we set tat 3983

4 / 5/1 ad 1/ =18 secods. Te total simulatio time is cose to be 24 ours. B. Simulatio Results Te performace measures obtaied troug te simulatio are te CDP, te CBP ad cael utilizatio. We simulatio we ew call arrival rate cages from 1 calls per miute to 35 calls per miute. Tese performace measures are plotted as a fuctio of te ew call arrival rate. Simulatio curves of te CDP of te two scemes are sow i Fig.6. From te figure, we ca see we traffic load cages, te values of te CDP of two scemes are lower ta tresold.1. We te traffic load is low (e.g., ew call arrival rate equals to 1 calls/miute), te values of te CDP of two scemes are equal. As te traffic load icreases, te CDP of te fuzzy sceme is lower ta te adaptive sceme obviously. It idicates tat te fuzzy sceme as a better robust performace. It ca adapt to cages i te etwork load. Fig.8. Cael utilizatio Simulatio curves of te CBP are sow i Fig.7. We ca otice tat te values of te CBP icrease as te traffic load icreases for bot scemes. We te traffic load is low (e.g., ew call arrival rate equals to 1 calls/miute), te values of te CBP of bot scemes are equal. Tey bot adust C to te miimum to reduce te CBP. As te traffic load icreases, te CBP of te fuzzy sceme is lower ta te adaptive sceme. It idicates tat our proposed algoritm ca adust C better to reduce te CBP ta te adaptive algoritm. Simulatio curves of te cael utilizatio are sow i Fig.8. We ca see from te figure, cael utilizatio of te fuzzy sceme is iger ta te adaptive sceme. It sows tat te fuzzy sceme ca utilize te resource more efficietly. Fig.6. Call droppig probability V. CONCLUSION Te paper presets a itelliget call admissio cotrol sceme based o fuzzy logic i wireless etworks. It searces automatically te optimal umber of te guard caels. Simulatio results sow te proposed sceme outperforms te adaptive cael reservatio sceme. Te fuzzy sceme as a better robust performace. It keeps te CDP ad CBP low at te same time. It guaratees te QoS provisio ad icreases te resource utilizatio of te etwork. Fig.7. Call blockig probability REFERENCES [1] D. Hog ad S. Rappaport, Traffic Model ad Performace Aalysis for Cellular Mobile Radio Telepoe Systems wit Prioritized ad No-Prioritized Hadoff Procedures, IEEE Tras. Veicular Tecology, vol.35, o.3, pp.77-92, [2] Carlos Oliveira, Jaime Bae Kim, ad Tatsuya Suda, A Adaptive Badwidt Reservatio Sceme for Hig-Speed Multimedia Wireless Networks, IEEE Joural o Selected Areas i Commuicatios, vol.16, o.6, pp , Aug

5 [3] Yi Zag ad Derog Liu, A Adaptive Algoritm for Call Admissio Cotrol i Wireless Network, i 21 Proc. IEEE GLOBECOM Cof., pp [4] YouCag Ko, CoogHo Co, Adaptive Hadoff Guard Cael Allocatio Sceme Usig Fuzzy Logic i Persoal Commuicatios Service, i 1997 Proc. IEEE ICUPC Cof., pp [5] Sugwook Kim, Pramod K. Varsey, A Adaptive Badwidt Reservatio Algoritm for Qos Sesitive Multimedia Cellular Networks, i 22 Proc. IEEE Veicular Tecology Cof., pp [6] S.Boumerdassi, A Efficiet Reservatio-based Dyamic Cael Assigmet Strategy, i 2 Proc. IEE 1 t 3G Moblie Commuicatio Tecoligies Cof., pp [7] Teodore S. Rappaport, Wireless Commuicatios Priciples ad Practice. New York: Pretice-Hall, 1996, c. 2. [8] Timoty J. Ross, Fuzzy Logic wit Egieerig Applicatios. New York: McGraw-Hill, 1995, c