3G CELLULAR NETWORKS TRAFFIC LOAD BALANCING BASED ON FUZZY LOGIC

Transcription

1 3G CELLULAR NETWORKS TRAFFIC LOAD BALANCING BASED ON FUZZY LOGIC Jane M. Mutua Department of Telecommuncaton Engneerng PAUSTI Narob, Kenya George N. Nyakoe Department of Telecommuncaton and Informaton Systems JKUAT Narob, Kenya Vtalce K. Oduol Department of Electrcal and Informaton Engneerng Unversty of Narob (UoN), Narob, Kenya Abstract Network congeston control remans mportant and of hgh prorty, especally gven the growng sze, demand, and speed of cellular networks. In ths paper the problem of capacty optmzaton n 3G Cellular Networks s addressed usng Rule-Based Fuzzy Logc to control the Plot power and as a result mprove the cell capacty. The autonomous operaton s amed at reducng the frequent attenton and effort requred by a rado optmzaton engneer to carry out capacty optmzaton tasks whch are currently done mostly manually. The FLC wll be nvolved n the detecton of hgh load 3G cells that do not have enough cell resources avalable and could beneft from CPICH power adjustment as a rado optmzaton engneer would normally do manually. Smulaton results show that the CPICH power control system based on fuzzy acheves a sgnfcant mprovement n the Downlnk cell utlzaton. Index Terms Capacty, Common Plot Channel (CPICH), Fuzzy Logc, Wdeband Code Dvson Multple Access (WCDMA); I. INTRODUCTION Network congeston control remans mportant and of hgh prorty, especally gven the accelerated growth n sze and demand of cellular networks [1]. Ths s where the need for an ntellgent capacty optmzaton algorthm becomes apparent. The 3G network Wdeband Code Dvson Multple Access (WCDMA) system s a self-nterferng system. As the Network load ncreases, the nterference rses thereby negatvely mpactng the qualty of servce and the coverage of cells. Therefore, the capacty, coverage, and qualty of servce of the WCDMA system are mutually dependent [2]. In other words, when one of these factors undergoes a change, the other factors are affected. The operator can trade qualty aganst cj3overage, capacty aganst qualty, but the amount of resources does not change, t s only redstrbuted. For example, to extend the coverage of a cell, t s requred to ether offer less capacty or decrease the qualty requrements, or both. In order to ncrease the capacty, t s necessary to shrnk the coverage or decrease the qualty requrements, or both. A. Capacty Plannng Capacty Plannng s essental n wdeband code dvson multple access (WCDMA) rado access networks (RAN) so as to evaluate the optmal ste confguraton n terms of plot and common control channel powers, throughput, and the soft handover parameter [3],[4]. The objectve for capacty plannng s to support the subscrber traffc wth acceptable low blockng and latency. The capacty of a cell affects the coverage of the cell. The cell breathes as the amount of users vares. To keep the qualty of servces n sutable levels, admsson control, packet schedulng, and handover mechansms are used. The mportance of capacty ncreases when the network expands and the amount of traffc grows. Each cell should be loaded relatvely equally and n a way that there s room for future growth. Automated optmzaton algorthms are requred to perform the rado network optmzaton process quckly and effcently, wth mnmal cost, tme and effort contrbuton [2]. The cell capacty requres frequent attenton by the rado optmzaton engneer due to the constantly shftng traffc patterns as subscrbers are added to the network [5]. Specfc consderatons for capacty optmzaton nclude dfference n the capacty requrements between peak and off-peak hours of the day. Therefore, ths requres more frequent changes n the network, e.g. several tmes a day to address movng patterns and varyng user concentratons, and are qute complex as they requre learnng and pattern recognton algorthms. Due to the complexty and expense of optmzng network coverage and capacty manually, partcularly as network operatons and performance management for data networks such as 3G get cumbersome, fuzzy logc s a good canddate for automated processng because of ts logcal resemblance to a human operator [6]. B. Network Parameters There exst numerous confgurable base staton parameters that the network operator can modfy to nfluence the coverage and capacty for example antenna settngs such as azmuth, heght and tlt, CPICH power and handover parameters [7]. 1) Common Plot Channel Power Control In rado access networks (RAN) usng wdeband code dvson multple access (WCDMA), the common plot channel (CPICH) s used by the user equpment for channel qualty estmaton, cell selecton, and handover [1]. The CPICH sgnal strength determnes the coverage area of the cell, affects the cell capacty, and n addton the qualty of servce, and s therefore a crucal parameter n rado network plannng and optmzaton [5]. The CPICH plot power allows for control of the strength of the CPICH sgnal such that the more power s set for the plot sgnals, the better coverage s obtaned. The optmal settng of CPICH power requres overcomng several challenges such as the coverage-capacty-qualty tradeoff, ensurng adequate handover performance, controllng the amount of nterference n the network and balancng the load among neghbor cells [8],[9]

2 A conventonal strategy s to unformly assgn a constant proporton, typcally 10-15%, of the total cell power to CPICH [2]. Although convenent, ths strategy may be neffcent n traffc varyng cells. It has been shown n prevous research that adoptng non-unform CPICH and optmzng ts power settng can save CPICH power [3],[4] and balance cell load [7],[8]. Whereas power savng on CPICH may not be a crucal aspect to the powercontrolled voce traffc, t s of great sgnfcance to data traffc. Moreover, reducng the CPICH power enables addtonal power savng on some of the other common control channels, of whch the power s typcally set n proporton to that of CPICH. Whle settng the CPICH power level the frst challenge we meet s a coverage-capacty tradeoff; ths tradeoff rses such that the hgher the CPICH power the bgger the coverage, whle the lower the CPICH power allows more power to be used by traffc channels. For ths project the CPICH power s optmzed n such a way that the Cell coverage s reduced when a hgh cell load s detected by reducng the CPICH power. When a low load cell s detected the CPICH power s ncreased causng an ncrease n the Cell coverage hence ensurng the cell s well utlzed. Due to ts sgnfcant mpact on 3G cellular networks, CPICH power optmzaton has been the subject of varous studes. In [5] for example, the authors presented a self-optmzaton based algorthm for tunng the CPICH plot power whch when runnng automatcally, the algorthm can be used to autonomously control the plot power and load balance traffc n the network and when scheduled or trggered manually, the algorthm can also be used to optmze the network capacty n clusters expectng a surge n durng a certan tme for example at a stadum durng a match. The study n [10] presented a model and algorthm for antenna tlt and plot power optmzaton for load balancng amed at maxmzng system capacty whle [11] presented a rule-based parametrc algorthm for common plot channel and antenna tlt optmzaton n UMTS FDD Networks. Whle these studes provded a good foundaton for ths, they focused on the varous mathematcal approaches wth no practcal end to end completeness and the authors dd not address challenges wth effcency and manual operatons n ther solutons. 2) Frequency Spectrum and Frequency Reuse The avalable frequency spectrum s a crucal factor n determnng the capacty of a 3G cellular system. To reduce the co-channel nterference, second generaton cellular networks lke GSM (Global System for Moble Communcatons) splt the avalable frequency spectrum among the cells to have dstnct frequences n the adjacent cells. Ths allows the possblty of dynamc spectrum allocaton to the cells to match the traffc dynamcs. However, 3G cellular networks are frequency-re-use-1 systems, meanng they use the complete avalable spectrum n each cell, to ncrease the spectrum utlzaton [12]. Therefore n these networks, dynamc capacty enhancement n a cell by spectrum allocaton s not a feasble soluton. 3) Base Staton Densty The densfcaton of Base Transmtter Statons (BTS), such that the nterference remans under a certan lmt can provde sgnfcant gans n network coverage and capacty. However, BTSs cannot be deployed at arbtrary places. Due to legal oblgatons and cost, they can only be deployed at some carefully selected places. Moreover, fnancal and tmng constrans also make ths opton feasble to cater for the long term coverage and capacty upgrades only. 4) Sectorzaton The BTS coverage can be dvded nto multple sectors usng drectonal antennas [12]. Unlke Omn-drectonal antennas, drectonal antennas radate the transmtted sgnals n a partcular drecton and therefore can ncrease the capacty of the network by reducng the nterference n other drectons. In tradtonal networks the number of sectors each BTS has, s decded at the plannng phase. As t requres ste vst and hardware upgrades to change the sectorzaton confguraton, t can only be done over large perods of tme. 5) Antenna Azmuth Antenna azmuth s defned as the angle of man beam of a drectonal antenna w.r.t. the North Pole n the horzontal drecton. It can be used to steer the antenna radaton pattern and to reduce the nterference to the adjacent cells. If the adjacent antennas pont towards each other they produce more nterference compared to f they are drected away from each other. The value of azmuth s normally nfluenced by the relatve postons of the adjacent BS and the targeted coverage areas. Therefore, the possblty of dynamc capacty enhancements by antenna azmuth adaptaton are lmted [13]. 6) Antenna Heght Antenna heght of the BSs also nfluences the receved sgnal strengths n ts coverage area. Hgher the antenna heght s, further the rado sgnals can propagate and therefore larger s the coverage area. However, ts value s fxed at the plannng phase and t s extremely dffcult to modfy t dynamcally. 7) Antenna Tlt Antenna tlt s defned as the elevaton angle of the man lobe of the antenna radaton pattern relatve to the horzontal plane. If the man lobe moves towards the earth t s known as downtlt and f t moves away t s known as uptlt. Hgher antenna downtlts move the man lobe closer to the BTS and vce versa. Therefore, the antenna tlt value has a strong nfluence on the effectve coverage area of the cell. Moreover, wth relatvely close drecton of the man lobe to the BS the receved sgnal strengths n own cell mproves and the nterference to neghborng cells reduces [14]. Ths mproves the sgnal to nterference plus nose (SINR) rato for the moble termnals and the network capacty ncreases. Therefore, antenna tlt can be used to alter both coverage and capacty of the network at the same tme [15]-[17]. 8) Handover for load balancng In a cellular network, load balancng can also be performed by shftng traffc between neghborng cells. Handover (HO) s the process of transferrng a call whch s n progress from one channel to another. It conssts of three man phases: measurement phase, decson phase and executon phase. By adjustng HO parameters settngs, the sze of a cell can be modfed to send users from the current cell to neghborng cells [18, 19]. Thus, the coverage area of the cell wth hgh congeston can be reduced and that of neghborng cells take up traffc from the congested cell edge and as a result of a more even traffc dstrbuton, the call

3 blockng probablty n the congested cell decreases [20]. Several studes n handover for load balancng have been done. In [21], a real tme traffc balancng n cellular network by mult-crtera handoff algorthm usng fuzzy logc for GERAN s presented. An algorthm to decde the balancng of the load between LTE, UMTS and GSM s presented n [22]. The load balancng and handover optmzaton functons may be used to mprove the network performance and a conflct may arse when both functons attempt to adjust the same parameters at the same tme. In [23] coordnaton algorthm of both functons s proposed wth and am to avod the stuaton n whch a parameter s smultaneously ncreased by both functons, achevng extreme values that may negatvely affect network performance. II. METHODOLOGY A fuzzy logc controller was desgned and smulated n Mat lab. The FLC was nvolved n the detecton of hgh load 3G cells that do not have enough cell resources avalable and could beneft from CPICH power adjustment as a rado optmzaton engneer would normally do manually. The FLC was desgned wth 3 nputs whch are the downlnk cell load, receved total wdeband power (RTWP) and the neghborng cells load. The output of the FLC was the CPICH power settng whch would determne whether to ncrease or decrease the coverage footprnt of the cell hence nfluencng the cell downlnk power utlzaton. C. Downlnk Cell The downlnk cell capacty s lmted by ts total avalable transmt cell power, whch s determned by the NodeB RF module capablty and the maxmum output power confgured for the cell. The proporton between voce and data traffc vares all the tme. The capacty left over from Voce traffc s reserved for the best effort data traffc. The overall goal s to provde as much capacty as possble to the users. The downlnk transmt power conssts of the followng, as shown n fg. 1: 1. Common channel (CCH) power 2. Non- hgh speed packet access (HSPA) power-voce 3. HSPA power-data traffc 4. Power margn Fg. 1. Dynamc power resource allocaton Downlnk power resources are allocated as follows: 1. Downlnk power resources are frst reserved for Common Control physcal channels and allocated to the Dedcated Physcal Channel. The remanng power resources are avalable for Data traffc. 2. The Data power resources are frst allocated to the hgh speed uplnk packet access (HSUPA) downlnk control channels whle the remanng power resources are allocated for hgh speed downlnk packet access (HSDPA). 3. The HSDPA power resources are frst allocated to the downlnk control channel hgh-speed shared control channel (HS-SCCH) whle the remanng power resources are allocated for the traffc channel hgh-speed physcal downlnk shared channel (HS-PDSCH). Downlnk power consumpton s related to cell coverage, user equpment (UE) locatons, and the traffc load n the cell. Large cell coverage, UEs beng physcally far away from the cell center, and heavy traffc load all contrbute to large downlnk power consumpton. Therefore, downlnk power overload s more lkely to occur n hotspots or n cells wth large coverage. When the Downlnk transmt power s nsuffcent, the followng occurs: 1. The data throughput decreases. 2. The servce qualty declnes. 3. New user servce requests are lkely to be rejected. D. Uplnk Interference (RTWP) The WCDMA system s lmted by nterference (the less nterference there s, the more capacty the system can offer to the users). Every user equpment (UE) accessng the network generates a sgnal whch, from the pont of vew of the base transcever staton (BTS), ncreases nterference n the system. At the same tme, the capacty of a WCDMA system s proportonal to the level of nterference n the system. The less nterference there s, the more capacty the system can offer. The relatonshp between the rse over thermal (RoT) and the uplnk load factor s as ndcated n fg. 2 below:

4 Fg. 2. Relatonshp between RTWP, nose ncrease, and uplnk load E. Neghbour cells load The Neghbor cells load as shown n fg 3 below; s factored so as to prevent traffc steerng to hgh loaded cells and the plot power adjustment s not adjusted when neghbor cell s already overloaded. As llustrated below the servng cell s hghly loaded same apples to the neghborng cells. Therefore n ths case t wouldn t be advsable to ramp down the CPICH power of the current servng cell as the traffc would be off-loaded to an already overloaded cell therefore negatvely mpactng further on the cell qualty. Fg. 4. Hourly Downlnk Cell and Call Setup Success Rate for Cell A Fg. 5. Hourly Downlnk Cell Utlzaton for Cell A and Cell B Fg. 3. Neghbor cells load Increasng plot power makes the cell bgger whle reducng plot power makes the cell smaller. Therefore, plot power can be used as a tool for traffc load balancng among cells. For ths study we shall consder a network of sx cells as n Fg. 6. Cells before plot power adjustment are as shown. The cell experencng excessve load s represented n red color whle green cells represent cells wth normal load. The ntersecton between the red and green cell on Fg. 6(a) represents an area whch s recevng plot power greater or equal to a gven threshold from more than one cell. Fg. 6(b) shows the cells after plot power adjustment on cell A durng whch the CPICH power s reduced and hence reducng the cell coverage. The yellow cells represent load after redstrbuton. A 3G network cell was dentfed (Cell A) whose Downlnk Cell and Call Setup Success Rate are as shown n fg 4 below. As the cell load ncreases, the qualty of the cell degrades and t has poor call set up success rate. The Overall goal s to balance the load between neghborng cells therefore mprovng the performance of the cell wth congeston ssues. Fg.5 below shows the Hourly Downlnk Cell load for two neghborng cells A and B. It can be seen that Cell A has so much load especally n the busy hour as compared to cell B. Therefore, Cell A load can be releved by handng over some of ts traffc to Cell B. Cell Cell (a) (b) Cell Cell Fg. 6. Network cells wth dfferent cell loadng

5 III. DEVELOPMENT OF FUZZY LOGIC CONTROLLER FOR CPICH POWER OPTIMIZATION The CPICH power of a cell was controlled based on Fuzzy rules based on the downlnk cell utlzaton. Frst a FLC was desgned on Mat lab wth 3 nputs whch are the Downlnk cell load, Receved Total Wdeband Power (RTWP); Interference n the cell and the neghborng cells load. The output of the FLC wll be the CPICH power settng whch wll determne whether to ncrease or decrease the coverage footprnt of the cell hence nfluencng the cell DL power utlzaton. Secondly the effect of varyng the CPICH power on the downlnk cell utlzaton based on fuzzy logc was nvestgated. The proposed FLC based cell capacty enhancement approach was then evaluated through a comparson wth a cell wth constant proporton CPICH power. It was expected that the cell would be utlzed better wth FLC n place as would be shown by the Key Performance Indcator (KPI) montorng tool. There were also expected gans on the data speeds n bts/sec at peak hours of the day. In the proposed algorthm the range of the Downlnk cell load s taken to be 0% to 100%, the RTWP vares from -110 dbm to -70 d m and the neghborng cells load ranges from 0% to 100%. The CPICH power settng was from 30dBm to 36dBm. For system smulaton Mamdan Fuzzy Inference system was used due to fact that the method s well suted to human nput and the nature of wreless networks s nonlnear. Fuzzy nference gathers the nput values of Downlnk Cell, Uplnk Interference and Neghbor cells load fuzzy set as crsp nputs and then evaluates them accordng to the fuzzy rules base. The composed and aggregated output of rules evaluaton was defuzzfed usng the centrod of area method and crsp output s obtaned. The nput lngustc varables were chosen as n Table 1 to Table 3 below: Table 1: Downlnk Cell fuzzy set Fuzzy set or label VLL: Very Low MLL: Medum Low NL: Normal MHL: Medum Hgh VHL: Very Hgh Table 2: Uplnk Interference fuzzy set Set Descrpton The load s very low as compared to the desred value The load s low but close to the desred value The load s n the normal range The load s hgh but close to the desred value The load s very hgh as compared to the desred value MSI: Medum Strong Interference VSI: Very Strong Interference Table 3: Neghbor cells load fuzzy set Fuzzy set or label NLL: Neghbour Low NNL: Neghbour Normal NHL: Neghbour Hgh The cell has medum strong nterference The cell has very strong nterference (very poor qualty) Set Descrpton The Neghbor cells load s very low as compared to the desred value The Neghbor cells load s n the normal range The Neghbor cells load s very hgh as compared to the desred value The output lngustc varables are gven Table 3-4 below: Table 4: CPICH Power fuzzy set Fuzzy set or label NLC: Negatve Large CPICH Power NSC: Negatve Small CPICH Power ZC: Zero CPICH Power PSC: Postve Small CPICH Power PLC: Postve Large CPICH Power Set Descrpton CPICH power to be large n the negatve drecton CPICH power to be small n the negatve drecton CPICH power to be around the normal value CPICH power to be small n the postve drecton CPICH power to be large n the postve drecton Fg. 7 to 10 show the fuzzy nput varable for Downlnk cell load, RTWP, the neghborng cells load and CPICH power settng respectvely. Each of the fuzzy nput varables has subsets whch are mapped to the correspondng membershp functons. The membershp functons defne how each pont n the nput space s mapped to a membershp value between 0 and 1. The Gaussan type membershp functon was chosen for the nputs and outputs because t represents the nonlnear nature of the problem n a better way than trangular or trapezodal membershp functons. Furthermore, the trangular and trapezodal membershp functons contan dscontnutes n ther dervatves, whch can result n abrupt changes n the output of the controller. The membershp functons for the downlnk cell load and uplnk nterference were made denser at the center n order to provde more senstvty. 75 IF-THEN rules provde knowledge to the system and decde whether to ncrease or decrease the output CPICH power of the cell. These rules very much resemble the human thought process, thereby provdng artfcal ntellgence to the system. Fuzzy set or label VWI: Very Weak Interference MWI: Medum Weak Interference ZI: Zero Interference Set Descrpton The cell has very weak nterference The cell has medum weak nterference The cell nterference s wthn the acceptable levels

6 Neghbour cells load s NLL then CPICH power s PLC Fg. 7. Membershp functon for Downlnk cell load IV. RESULTS AND DISCUSSION In the network scenaro, we used 2 base statons wth 3-sector antennas thus comprsng 6 cells. The calculaton for the Downlnk cell load wthout fuzzy at any tme n percentage (%)) wll be gven by the equaton below; (D Utl D Utl (%) = 10(P Utl(dBm) 10) (43) 10 ) (1) Where P Utl (dbm) s the Cell Utlzaton n dbm. The constant CPICH power P CPICH (dbm) for the cell s 33dBm. The power converson of dbm to watts s gven by the formula: P CPICH (W) = 1W. 10 (P CPICH (dbm) 10 ) 1000 (2) The calculaton for the Downlnk cell load wth fuzzy at any tme n percentage (D Fuzzy Utl (%)) wll be gven by the equaton below; Fg. 8. Membershp functon for RTWP (Interference n the cell) D Fuzzy Utl (%) = {(D Utl (%) 10) ) 1000 ) } 1000 ) (Fuzzy CPICH(dBm) (33) 10 CPICH(dBm) 10 10(Fuzzy + { 10 (43) 10 ) } (3) Where D Utl (%) s the Downlnk cell load and Fuzzy CPICH(dBm) s the output of the Fuzzy logc controller Fg. 9. Membershp functon for neghborng cells load Fg. 10. Membershp functon for CPICH power settng F. Effect of varyng the CPICH power on the downlnk cell utlzaton Fg. 11. below shows the normal utlzaton trend of a cell wth a constant CPICH power assgned. It can be seen that the CPICH power s constant at 33dBm, that s, 10% of the total downlnk cell power. The downlnk cell utlzaton vares throughout the day dependng on the traffc. As seen from the fgure n the wee hours of the nght.e. 4am the cell utlzaton s very low at 20%. As the day progresses, the downlnk cell utlzaton ncreases and at 9pm the cell utlzaton s at the hghest at 96%. As stated before ths wll have an mpact on the data throughput as the data only utlzes the remanng power on sharng mode among all the users. It therefore shows that there s a need for better utlzaton of the cell. An example of the of the IF-THEN rules nclude; 1. If Downlnk s VLL and nterference s VWI and Neghbour cells load s NLL then CPICH power s PLC 2. If Downlnk s VLL and nterference s MWI and Neghbour cells load s NLL then CPICH power s PLC 3. If Downlnk s VLL and nterference s ZI and

7 Fg. 11. Normal Downlnk Cell Utlzaton and CPICH power Fg. 12. Fuzzy logc optmzed Downlnk cell utlzaton and CPICH power Fg. 12 above shows the Downlnk cell utlzaton trend of a cell after fuzzy logc has been appled to vary the CPICH power. In ths case the CPICH power s no longer constant at 33dBm but vared from a hgh of 36dBm, that s, 20% of the total cell power to a low of 30dBm, that s, 5% of the cell power. Also the cell load utlzaton s seen to be better as the wee hours of the nght carry more traffc than before as the CPICH power was ncreased, hence ncreasng the cell coverage footprnt and the cell traffc. Better data speeds at the busy hour are acheved as the cell utlzaton s now lower. A and cell B after the CPICH plot power adjustment of cell A was done usng fuzzy logc. The adjustment of the CPICH plot power n cell A especally n the busy hours when the cell has excessve load resulted n the excessve traffc beng steered to cell B whch has less traffc durng that tme. Therefore t can be concluded that the tunng of the CPICH power has a great mpact on the downlnk cell utlzaton largely due to the mpact the tunng of the CPICH power has on the cell coverage area. Fuzzy logc control was successfully used to automatcally control the CPICH power and mproved the downlnk cell load. In addton t was found that the CPICH power tunng has a large mpact on the downlnk cell utlzaton and t can be used to mtgate the load mbalance between neghbourng cells by changng the coverage area of the cells. G. Evaluaton of the CPICH Power optmzaton based on Fuzzy Logc Controller The evaluaton of the performance of the CPICH power control system based on fuzzy logc was done through 2 metrcs. These were the cell s call setup success rate (CSSR) and fnally a comparson of the downlnk cell utlzaton wth and wthout fuzzy control was done. Fg 14 below shows the call setup success rate (CSSR) trend before and after the CPICH plot power control usng fuzzy logc. There was a great mprovement especally n the peak hours when there was a dp as below as 38.5% prevously mprovng to 88.5%. Ths ndcates over 100% mprovement n ths KPI durng the peak hour. Fg. 14. Comparson of the Downlnk Cell load wth and wthout fuzzy logc control Fg. 13. Hourly Downlnk Cell Utlzaton for Cell A and Cell B after CPICH power control usng fuzzy logc Fg. 13 above shows the hourly downlnk cell utlzaton for cell Fg. 15 below shows the comparson of the Downlnk cell utlzaton trend of a cell wth and wthout fuzzy logc control beng appled to vary the CPICH power. It can be seen that the Downlnk cell utlzaton wth fuzzy s better that wthout fuzzy. In the off peak hours of the day, that s, from mdnght to md-day the cell wthout fuzzy was very low utlzed but after the applcaton of fuzzy logc control the traffc ncreased and the cell s more effcently utlzed. In the peak hours of the day, that f, between 7pm and 10pm the cell was over utlzed resultng to poor experence but after fuzzy logc s appled the cell has optmal utlzaton after the decrease n the CPICH power resultng to a lower traffc

8 Fg. 15. Comparson of the Downlnk Cell load wth and wthout fuzzy logc control V. CONCLUSION AND FUTURE WORK We demonstrated a CPICH power control algorthm for 3G cellular networks for cell capacty based on fuzzy logc control. Through Fuzzy Logc Control, the developed algorthm addresses the challenge of ncreasng complexty of manually optmzng the ever growng and dynamcally changng traffc patterns of a 3G network. The developed algorthm can be used to autonomously optmze CPICH power and balance traffc load n the network, for example n the case of a traffc surge due to an accdent or an event happenng at a locaton. Smulaton results show that the CPICH power control based on fuzzy process acheves a sgnfcant mprovement n the Downlnk cell utlzaton whch n turn mproves the cell performance. The current study was lmted to 2 neghborng base statons; further research can be done to nclude a number of cells n a cluster. In addton further work should be done wth regards to nvestgatng the mpact of the CPICH power optmzaton based on Fuzzy logc on the cell data throughputs. REFERENCES [1] C. Johnson, Rado Access Networks For UMTS; Prncples and Practce, John Wley & Sons, [2] H. Holma,. Toskala, WCDMA for UMTS Rado Access for Thrd Generaton Moble Communcaton, John Wley & Sons, 2000 [3] M. N. Islam, R. Abou-Jaoude, C. Hartmann, A. Mtschele-Thel, Self Optmzaton of ntenna Tlt and Plot Power for dedcated channels, Proceedngs of the 8th Internatonal Symposum on Modelng and Optmzaton n Moble, Ad Hoc and Wreless Networks (WOpt), June pp.196,203, [4] L. Chen, D. Yuan, "CPICH power plannng for optmzng HSDPA and R99 SHO performance: Mathematcal modellng and soluton approach", Wreless Days WD '08. 1st IFIP, pp. 1, Nov [5] H. Mfula, T. Isotalo, J. Nurmnen, "Self-optmzaton of power parameters n WCDMA networks", Hgh Performance Computng & Smulaton (HPCS) 2015 Internatonal Conference on, pp , [6] F. Karray and C. De Slva, Soft computng and tools of ntellgent systems desgn. Harlow: ddson-wesley, [7] J. Sánchez-González, J. Pérez-Romero and O. Sallent, "A Rule-Based Soluton Search Methodology for Self-Optmzaton n Cellular Networks", Communcatons Letters IEEE, vol. 18, pp , 2014, ISSN [8] I. Somna, P. V arbrand, and D. Yuan, "Plot power optmzaton and coverage control n WCDMA moble networks", Omega, 35 pp , [9] I. Somna and D. Yuan. Mnmum, "Plot power for servce coverage n WCDMA networks", Wreless Networks, 14, pp , 2008 [10] M. N. Islam, R. Abou-Jaoude, C. Hartmann, A. Mtschele-Thel, Self Optmzaton of ntenna Tlt and Plot Power for dedcated channels, Proceedngs of the 8th Internatonal Symposum on Modelng and Optmzaton n Moble, Ad Hoc and Wreless Networks (WOpt), June , pp.196,203, [11] A. Gerdentsch, S. Jakl, Y. Y. Chong, M. Toeltsch. A Rule-Based Algorthm for Common Plot Channel and Antenna Tlt Optmzaton n UMTS FDD Networks, ETRI Journal, Vol. 26, Issue 5, pp , Oct [12] M. Terre, M. Pschella and E. Vver, Wreless telecommuncaton systems. Hoboken: Wley, [13] Okumuş H.İ., Şahn E., kyaz Ö., " ntenna zmuth Poston Control Wth Fuzzy Logc And Self-Tunng Fuzzy Logc Controllers", 8th Internatonal Conference on Electrcal and Electroncs Engneerng, ELECO 2013, pp [14] D. W. Kfle, B. Wegmann, I. Verng, A. Klen, "Impact of antenna tltng on propagaton shadowng model", 2013 IEEE 77th Vehcular Technology Conference (VTC Sprng), pp. 1-5, June [15] B. Yu, L. Yang, H. Ish, X. Cheng, " Balancng wth Antenna Tlt Control n Enhanced Local Area Archtecture", 2014 IEEE 79th Vehcular Technology Conference (VTC Sprng), pp. 1-6, 2014 [16] K. Toda, T. Yamamoto, T. Ohsek, S. Konsh, " Balancng Technques Based on Antenna Tlt and Handover Tmng Control", Vehcular Technology Conference (VTC Fall), 2013 IEEE 78 th, pp. 1-6, 2013 [17] I. Forkel,. Kemper, R. Pabst, R. Hermans The effect of electrcal and mechancal antenna down-tltng n UMTS networks, 3G Moble Communcaton Technologes, Thrd Internatonal Conference on (Conf. Publ. No. 489), pp 86 90, 2002 [18] P. Muñoz, R. Barco, I. de la Bandera, "Optmzaton of load balancng usng fuzzy Q-Learnng for next generaton wreless networks", Expert Syst. Appl., vol. 40, no. 4, pp , Mar [19] P. Muñoz, R. Barco, I. Bandera, "On the Potental of Handover Parameter Optmzaton for Self-Organzng Networks", Vehcular Technology IEEE Transactons on, vol. 62, pp , [20] W. dnan; R. Rahman; M.Ibrahm, M. Kassm, Performance evaluaton of soft handover parameters n WCDM system, Instrumentaton Control and Automaton (ICA), nd Internatonal Conference on, pp , Feb [21] P. Munoz, R. Barco, I. De la Bandera, M. Torl, and S. Luna-Ramfrez, Optmzaton of a fuzzy logc controller for handover-based load balancng, n Proc. IEEE 73rd IEEE Vehcular Technology Conference (VTC), May [22] H M ElBadawy, "Optmal RAT selecton algorthm through Common Rado Resource Management n heterogeneous wreless networks", Rado Scence Conference (NRSC) th Natonal, pp. 1-9, [23] P. Munoz, R. Barco, S. Fortes, "Conflct resoluton between load balancng and handover optmzaton n LTE networks", IEEE Commun. Lett., vol. 18, no. 10, pp , Oct