Fuzzy Logc ower Control n Cogntve Rado Zeljko Tabakovc Croatan ost and Electronc Communcatons Agency Jursceva 3 Zagreb, Croata zeljkotabakovc@telekomhr Sonja Grgc, Mslav Grgc Unversty of Zagreb, Faculty of EE and Comp Department of Wreless Communcaton Unska 3/XII, Zagreb, Croata sgrgc@eeeorg; mgrgc@eeeorg Abstract Opportunstc rado spectrum access has the possblty to mprove spectrum utlzaton needed for next generaton moble networks The man challenge to opportunstc rado spectrum access les n fndng balance n conflctng goals of satsfyng performance requrements whle mnmzng nterference In ths paper we propose new strategy for fuzzy logc transmt power control whch enables cogntve secondary user to acheve ts requred transmsson rate and qualty, whle mnmzng nterference to the prmary users and other concurrent secondary users Keywords - cogntve rado; opportunstc rado spectrum access; transmt power control; fuzzy logc systems I INTRODUCTION In recent years, demand for wreless communcaton servces has grown far beyond earler predctons Furthermore, n order to satsfy future market demand for moble and broadband servces, we can envsage deployment of next generaton moble networks and servces whch wll need rapd and more flexble access to rado spectrum Due to polcy of exclusve frequency assgnment, rado spectrum has become congested and scarce resource Nevertheless, related surveys have proved that most of the allocated spectrum s underutlzed [, ] To deal wth ncreasng conflct of spectrum congeston and spectrum underutlzaton, cogntve rado technque has been proposed as a flexble method whch allows secondary users to utlze already lcensed bands opportunstcally [3, 4] Opportunstc rado spectrum access has the possblty to mprove spectrum utlzaton and n perspectve allowng next generaton moble networks access to the attractve rado spectrum bands The man challenge to opportunstc rado spectrum access les n fndng balance n conflctng goals of satsfyng performance requrements for secondary user ) whle mnmzng nterference to the actve prmary users U) and other secondary users Secondary user should not degrade performance statstcs of lcensed prmary users In order to acheve these tasks, secondary user s requred to recognze prmary users, determne envronment characterstcs and quckly adapt ts system parameters correspondng to the operatng envronment Man abltes of cogntve rado CR) wth opportunstc rado spectrum access capabltes are spectrum sensng, dynamc frequency selecton and adaptve transmt power control In recent years, studes on transmt power control TC) are progressng n order to nvestgate dfferent TC strateges for opportunstc rado spectrum access systems [5-9] resented TC strateges dffer dependng on settngs of prmary goals for TC, presumptons about avalable nput data and on methodology used for transmt power control parameter determnaton In [5], opportunstc TC s presented whch enables cogntve user to maxmze ts transmsson rate e power, whle guaranteeng U outage probablty The authors n [6] proposed fuzzy logc TC scheme whch dynamcally adjust transmt power relatng to nterference observed at U, dstance between U and and receved power dfference at the base staton In order to avod nterference at U, exchange of sensng nformaton between U and s requred In [7], authors propose dstrbuted cogntve network access scheme wth the am of provdng best QoS wth respect of combnaton of rado lnk and core network performance Fuzzy logc decson has been used to choose the most sutable access opportunty even n mult-technology scenaros A power control approach based on spectrum sensng sde nformaton n order to mtgate nterference to the U s presented n [8] Cogntve rado transmt power s calculated n three step procedure usng mssng probablty of energy detecton dependence on dstance between U and In [9], the authors nvestgate the optmal power control wth and wthout nterference temperature constrants based on observed Shannon capacty The optmal power control n cogntve rado network s modeled as a concave mnmzaton problem [] In ths paper we propose alternatve transmt power control strategy whch enables cogntve secondary user to acheve ts requred transmsson rate and qualty, whle mnmzng nterference to the prmary users and other concurrent secondary users roposed TC ensures that each n the network receves and transmts just enough energy to convey necessary nformaton Spectrum sensng data and regulatory requrements defnes maxmum acceptable transmt power Dependng on the qualty of servce, recever sets requred sgnal-to-nterference-plus-nose rato ) Comparson of measured and requred at recever determnes transmt power control rato and mnmum requred transmt power transmt power s determned by balancng these two requrements of maxmum acceptable transmt power to satsfy nterference constrants and mnmum requred transmt power to satsfy determned level of servce 978--444-453-/9/$5 9 IEEE
roposed TC strategy s mplemented usng fuzzy logc system FLS) [, ] Fuzzy logc systems have been successfully appled n felds such as automatc control, data classfcaton, decson analyss, expert systems and computer vson Advantage of FLS s that t merge objectve analytc and experence based subjectve knowledge FLS formalze control algorthms whch can tolerate mprecson and uncertanty of nput data lke spectrum sensng data and measurements n ths case Addtonally, proposed TC strategy can be mplemented wth low cost and easy to mplement fuzzy logc controllers The advantage of proposed TC strategy s n mnmzng mutual nterference and reducton of frequency reuse dstance for U and other Ths leads to rado spectrum utlzaton mprovement and ncreasng overall networks capacty wth avalable rado spectrum Addtonally, reducton of transmtter power results n mnmzng battery consumpton of moble termnals for next generaton moble networks and servces The outlne of the paper s as follows Secton descrbes system model of prmary lcensed system and co-exstence wth opportunstc spectrum access secondary system n the same geographc area In secton 3 outlne of the proposed TC strategy s presented The desgn of cogntve rado fuzzy logc TC s elaborated n secton 4 Smulaton results and performance evaluaton of proposed TC strategy are presented n secton 5 Conclusons are gven n Secton 6 II SYSTEM MODEL In ths paper we consder scenaro n whch a prmary system s lcensed servce wth a coherence tme T C and actvty probablty α followng block statc model rmary transmtter s operatng wth average transmt power U _ Tx Cogntve secondary system co-exsts n the same area wth prmary system usng opportunstc rado spectrum access and should not ncrease level of nterference observed by prmary system We assume that prmary system s not aware of the presence of cogntve secondary system and that there s no actve communcaton or cooperatve behavor between U and Channel model for a par of prmary users and cogntve secondary users s shown n Fg We assume that the quasstatc Raylegh fadng s present, and the channel coeffcents between communcaton users are consdered to be ndependent Raylegh dstrbuted varables The receve sgnal model can be presented as: Y Y U U + U t t) = h t) + h t) Z ) ) t) = h t) + h t) Z ) ) U + t where Y U and Y denotes U and receved sgnal, X U and X denotes U and transmtted sgnal, t s tme, h and h are random varables representng fadng channel coeffcents between prmary and secondary transmtter and recever, h and h are random varables representng nterference fadng channel coeffcents Z U t) and Z t) are addtve whte Gaussan nose at U and wth varances σ U and σ Observed at the prmary recever U _ A can be represented as: h A U _ A U _ A = N K h + h _ j U _ j = j= j A + σ U _ A and at the cogntve secondary recever _ B s calculated as: h B _ B = N K h _ = j= B + h _ B j U _ j + σ _ B where K and N represent number of co-channel prmary and cogntve secondary users, h j s power gan of fadng channel coeffcents, U _ A, U _ j are transmtted powers of co-channel prmary users, and _ B, _ are transmtted powers of co-channel cogntve secondary users at the cogntve secondary recever s used for determnaton of mnmum requred transmt power of n order to mnmze nterference potental of cogntve secondary user U Tx h sensor TC Tx U Rx Rx Fgure Channel model for scenaro of co-exstng par of U and III OWER CONTROL STRATEGY In order to fully explot potental of opportunstc spectrum access, nterference control s a crucal ssue It s essental to keep the transmsson power of the cogntve at a mnmum level whle ensurng adequate sgnal qualty at the recevng end roposed power control strategy s based on balancng transmt power level between mnmum requred and maxmum acceptable Mnmum requred transmt power s obtaned by adjustng t to satsfy targeted at the recever Maxmum acceptable transmt power s establshed consderng permssble nterference at the prmary user's recever It s determned by spectrum sensng data and h h h h Z U Z 3) 4)
predefned regulatory constrants As a result, proposed TC strategy tres to mantan requred at the cogntve recever, whle not causng excessve nterference to the lcensed prmary users Intal power of cogntve transmtter s calculated usng path loss estmaton determned by measurement of common plot channel n open-loop power control cycle Common plot channel s broadcast sgnal wth known transmtted power Intal power s calculated as: Tx = Tx _CC Rx _ CC Ladd + Req + N + I 5) where Tx s ntal transmt power, Tx _ CC s predetermned transmtted power of common plot channel, Rx _ CC s measured power of common plot channel at recever, L add are addtonal gan, loss and tolerances, Re q s requred, and N + I s measured nose plus nterference at Intal power of cogntve transmtter s used for ntal communcaton between transmtter and recever of next generaton moble network It s startng pont for process of adaptve adjustng of transmtter power whch s done by teratve self adaptng power control strategy descrbed bellow Instant transmt power of the secondary user Tx _ s determned by balancng maxmum acceptable transmt power Tx MAX and mnmum requred transmt power Tx Re q as follows: Tx Re q f Tx MAX Tx Re q Tx _ = 6) f Tx MAX < Tx Re q Maxmum acceptable transmt power MAX represent maxmum allowed power of the cogntve n order to satsfy nterference constrants determned by spectrum sensng of the prmary user at the secondary user locaton Tx sens and requrements mposed by regulatory regme Tx rr Maxmum acceptable transmt power s determned as: Tx { } Tx MAX mn Tx sens, = 7) Tx rr Determnaton of transmt power nterference constrant Tx sens used for calculaton of maxmum acceptable transmt power s llustrated n Fg Based on the sensng power Rx _ sens measured at the secondary user transmtter we can determne transmt power nterference constrant accordng to the followng three cases: Tx sens CASE : Sensng power s bellow U threshold level Rx _ sens < U _ th ) When lcensed prmary user s undetectable, cogntve secondary user transmtter can transmt wth ts peak power If sensng controller s determnng prmary lnk wth low power level, cogntve has to reduce ts maxmum transmt power n order to avod harmful nterference to the U recever at unknown locaton Transmt power nterference constrant s proportonal to the addtonal path loss beyond threshold level as llustrated n Fg Tx sens MAX Fgure Determnaton of transmt power nterference constrant CASE : Sensng power s above U threshold level n the U workng area U _ th Rx _ sens < U _ th + U _ Re q + M ) where U _ Re q s requred for U and M s addtonal fadng and multple nterference margn In ths case transmtter of the cogntve secondary user s stuated near the edge of the U coverage zone, therefore all transmssons should be avoded n order no to cause addtonal nterference to the prmary user recever n the same area referably, secondary user should delay transmssons for later of transfer ts transmssons to dfferent channel usng channel mtgaton technques for opportunstc rado spectrum access CASE 3: Sensng power s well above U threshold level + + M ) U _ th U _ Re q Rx _ sens 3 N U_th U_th+ U_Req+M Snce sensng controller s determnng prmary lnk wth relatvely hgh power, cogntve s stuated near to the U transmtter In ths case cogntve can transmt wth lower power level wthout causng measurable dsturbng nterference to the U recever Ths case s approprate for short range communcatons between transmtter and recever dstant from U recever Transmt power nterference constrant of cogntve should be adjusted n such way that requred for the U recever s guaranteed Mnmum requred transmt power Tx Re q Rx sens determnes cogntve transmt power just enough to satsfy requred at recever Comparson of measured at recever wth requred determnes transmt power control rato for transmt power adjustment The new mnmum requred transmt power s obtaned by multplyng present mnmum requred transmt power wth transmt power control rato R TC obtaned from the output of the fuzzy logc system as follows:
Tx Re q t + ) = t ) R 8) Tx Re q TC ossble mplementaton of proposed adaptve transmt power control strategy usng fuzzy logc s descrbed n next secton IV FUZZY LOGIC TC DESIGN Fuzzy logc transmt power controller for opportunstc rado spectrum access contans fuzzy logc processor, performance evaluator and regulatory database as shown n Fg 3 Regulatory database contans regulatory rules defnng maxmum permssble radated power or power spectral densty of cogntve transmtter n geographc area of nterest Spectrum sensng data collected at cogntve determnes level of spectrum actvty and potental nfluence of cogntve to the prmary network These two elements defne maxmum acceptable transmt power of cogntve as descrbed n secton 3 erformance evaluator compares measured at recever wth requred Based on dfference fuzzy logc processor determnes transmt power adjustment for estmatng mnmum requred transmt power of cogntve Req erformance evaluator Dff Knowledge repostory Rules Fuzzy sets Regulatory database Tx rr Fuzzy logc processor Spectrum sensng data Fuzzfer Rx_sens Inference system Fuzzy logc transmt power controller supply three nput varables to fuzzy logc processor: dfference Dff ) from performance evaluator, maxmum permssble radated power Tx rr from regulatory database and spectrum sensng data Rx_sens from Tx sensor Output varables are: maxmum acceptable transmt power Tx MAX and transmt power control rato R TC For membershp functon for nput and output varables a trapezodal functon g x; x, x, a a s chosen gven by: ), x x + for x a < x x a g x) = for x < x x 9) x x + for x < x x + a a where x x ) s left rght) edge of the trapezodal functon and a a ) s left rght) wdth of the trapezodal functon V ERFORMANCE EVALUATION In ths secton we gve some numercal results of proposed fuzzy logc transmt power control strategy We have randomly generated 3 transmtter and recever pars and smulated 3 of Raylegh faded receved sgnal for each par To evaluate transmt power control strategy maxmum power of U transmtter s set to 4 dbm and maxmum power of transmtter s set to 33 dbm U s mplemented wthout transmt power control functon Transmt power control rato R TC s obtaned as crsp db value from dfference of observed and requred at recever Fg 4 shows Sugeno type fuzzy nference dagram for determnng transmt power control rato Defuzzfer Tx MAX R TC Fgure 3 Archtecture of fuzzy logc transmt power controller Fuzzy logc processor s n general non-lnear mappng parallel processor of nput data vector nto output data scalar or vector Fuzzy logc processor contans four man components: fuzzfer, nference system engne, knowledge repostory and deffuzzfer as shown n Fg 3 Input varables are crsp values whch are transformed n approprate fuzzy sets degree of membershp va membershp functon Implcaton method s performed by the set of rules whch have general form: f X s X and Y s Y THEN Z s Z Knowledge repostory contans nference rules used for connectng nput and output space of lngustc varables and fuzzy sets of nput and output varables Aggregaton and deffuzfcaton of results s done by the weghted average of all rule outputs obtaned by nference system A detaled overvew of the fuzzy logc systems s out of the scope of ths paper and can be fnd n the lterature [, ] Fgure 4 Fuzzy nference dagram of R TC The maxmum acceptable transmt power of secondary user s determned usng regulatory constrant set to 3 dbm and smulated U spectrum sensng data Spectrum sensng data of prmary user transmtter s smulated wth 3 randomly dstrbuted Raylegh faded sgnal for three characterstc cases of spectrum sensng data as descrbed n secton 3 Fg 5 shows results of adaptve regulaton of transmt power usng proposed cogntve rado fuzzy logc controller Frst dagram shows calculated dfference at recever whch s nput parameter for determnaton of
transmt power control rato parameter transmt power s adjusted n order to acheve requred at recever and satsfy nterference constrants defned by maxmum acceptable transmt power as shown n second dagram It can be seen that nstant power of the transmtter s sgnfcantly bellow maxmum power of transmtter used n lterature for capacty optmal power adaptaton [5, 9] Fnally, resultng measured at recever s presented n thrd dagram Fuzzy logc transmt power controller mantans constant recever wth ± db devaton comparng to requred dfference db) transmt power dbm) db) - - -3 3 5 5 5 3 4 5 3 4 5 3 - - -3-4 3 4 5 Fgure 5 Fuzzy logc transmt power controller smulaton results Average transmt power n analyzed scenaro s dbm Comparng ths to average transmt power for capacty optmal power adaptaton we can calculate co-channel frequency reuse dstance for proposed TC strategy We can conclude that co-channel frequency reuse dstance s reduced by 335 tmes n moble envronments usng dstance attenuaton factor of 38 Therefore, nterference potental of transmtter s reduced and same rado frequency channel can be used more frequently leadng to mprovng spectrum utlzaton and ncreasng overall networks capacty wth avalable rado spectrum As a consequence of mnmzng the transmtter power n proposed TC strategy battery consumpton of moble termnals for next generaton moble networks and servces s reduced VI CONCLUSIONS In ths paper, we have presented alternatve transmt power control strategy for cogntve secondary users applyng opportunstc spectrum access Transmt power control s realzed usng fuzzy logc system whch enables smple and low cost mplementaton of TC functon resented TC strategy allows cogntve secondary user to acheve ts requred transmsson rate and qualty, whle mnmzng nterference to the prmary users and other concurrent secondary users Advantage of presented cogntve TC scheme s that t s followng altrustc approach whch results n smaller nterference potental and reducton of frequency reuse dstance Ths allows also other next generaton moble network users to beneft from avalable rado spectrum, leadng to the mprovement of overall spectrum utlzaton and maxmzng overall U and networks capacty REFERENCES [] Federal Communcatons Commsson, "Facltatng opportuntes for flexble, effcent and relable spectrum use employng cogntve rado technologes", notce of proposed rulemakng and order, FCC 3-3, December 3 [] KN Steadman, AD Rose, and TTN Nguyen, "Dynamc Spectrum Sharng Detectors", nd IEEE Internatonal Symposum on New Fronters n Dynamc Spectrum Access Networks DySAN 7), Dubln, Ireland, Aprl 7, pp 76-8 [3] J Mtola and GQ Magure, "Cogntve rados: Makng software rados more personal", IEEE ers Commun, Vol 6, No 4, August 999, pp 3-8 [4] S Haykn, "Cogntve rado: Bran-empowered wreless communcatons", IEEE J Select Areas Commun, Vol 3, No, February 5, pp - [5] Y Chen, G Yu, Z Zhang, H-H Chen, and Qu, "On Cogntve Rado Networks wth Opportunstc ower Control Strateges n Fadng Channels", IEEE Trans On Wreless Communcatons, Vol 7, No 7, July 8, pp 75-76 [6] H-S T Le and Q Lang, "An Effcent ower Control Scheme for Cogntve Rados", IEEE WCNC 7 roceedngs, pp 559-563 [7] N Baldo and M Zorz, "Cogntve Network Access usng Fuzzy Decson Makng", IEEE ICC 7 roceedngs, pp 654-65 [8] K Hamd, W Zhang, and KB Letaef, "ower Control n Cogntve Rado Systems Based on Spectrum Sensng Sde Informaton", IEEE ICC 7 roceedngs, pp 56-565 [9] We Wang, T eng, and W Wang, "Optmal ower Control under Interference Temperature Constrants n Cogntve Rado Network", IEEE WCNC 7 roceedngs, pp 6- [] R Horst, M ardalos, and NV Thoa, "Introducton to Global Optmzaton", Kluwer Academc ublshers, [] JM Mendel, "Fuzzy Logc Systems for Engneerng: A Tutoral", IEEE roceedngs, Vol 83, No 3, March 995, pp 345-377 [] E Cox, "Fuzzy Fundamentals", IEEE Spectrum, Vol 9, No, October 99, pp 58-6