Enhanced Uplink Scheduling for Continuous Connectivity in High Speed Packet Access Systems

Transcription

1 Int. J. Communcatons, Network and System Scences, 212, 5, Publshed Onlne August 212 ( Enhanced Uplnk Schedulng for Contnuous Connectvty n Hgh Speed Packet Access Systems Saed M. Abd El-Atty 1, Konstantnos Lzos 2 1 Department of Computer Scence and Informaton, College of Arts & Scence, Salman Bn Abdulazz Unversty, Wad Addawasr, KSA 2 Head of IT Department, Embassy of Greece Nobels Gate 45, Oslo, Norway Emal: sabdelatty@gmal.com, klzos@mfa.gr Receved September 21, 211; revsed June 15, 212; accepted July 1, 212 ABSTRACT The effcency of hgh speed packet access (HSPA) systems s manly based on channel qualty nformaton (CQI) reports transmtted by user equpment (UE) to Node B. In ths paper, we propose an mproved CQI predcton scheme based on a fnte state Markov chan (FSMC) model for wreless channel, n order to reduce CQI sgnalng overhead n the HSPA system. Then, we ntroduce an enhanced uplnk packet schedulng (EUPS) scheme to provde qualty of servces (QoS) guaranteed for contnuous packet connectvty n the enhanced uplnk (EUL). EUPS serves the actve UEs not only accordng to buffer status but also accordng to reported state of the wreless channel of each UE. The performance of the proposed scheme n terms of average packet delay, average packet drop and average cell throughput s compared to the classcal scheduler of 3GPP standards. The smulaton results show the effectveness of the proposed scheme. Keywords: EUL; Schedulng; CQI; FSMC; HSPA 1. Introducton Nowadays, HSPA s the best communcaton system to provde dfferent servces wth QoS guaranteed to UEs at any tme, any place, and supportng full moblty n a cost-effectve and effcent manner. HSPA s also the best avalable rado-based technology when consderng hgh download and upload speeds. The uplnk s becomng ncreasngly mportant, wth many UEs nterested n uploadng ther clps onto onlne streamng provders and smlar vdeo-sharng stes [1]. The enhanced uplnk (EUL) of WCDMA or commercally called hgh speed uplnk packet access (HSUPA) s deployed n moble marketng at the end of 29. The man benefcal of the HSUPA s not only provdng hgh uplnk spectral effcency compared to R99 UMTS systems but also delverng hgh data rate at the end users [2]. The schedulng procedure n HSPA systems s based on a set of rules that contans sgnallng control nformaton n the downlnk and uplnk between Node B and UEs. The hgh speed downlnk packet access (HSDPA) schedulng s manly based CQI reports that are transmtted from UEs to Node B, n order to acheve hgh system effcency. On the other hand, the HSUPA utlzes fast schedulng, hybrd automatc repeat request (HARQ), and shorter transmsson tme nterval TTI (2 ms) n order to enhance system throughput and mnmze transmsson delay [3]. However, accordng to 3GPP standards, the HSUPA scheduler at Node B s manly based on the schedulng nformaton (SI) that contans logcal nformaton, related to the power transmsson and buffered data of UEs. As a result, the schedulng procedure for multmeda applcatons packet data does not qute guarantee the delay requrements of dfferent servces. Also n HSPA systems, CQI s perodcally reported from UEs to Node B. However, the frequent CQI reports ncrease uplnk nterference, reducng the sgnal recepton qualty at the uplnk, causng degradaton of the system spectral effcency. Hence, the uplnk scheduler experences some challenges to serve contnuous packet connectvty and guarantee the QoS requrements of the multmeda traffc flows. Therefore, our goal n ths work s to mprove the performance effcency of HSPA systems, hence we propose an mproved CQI reportng scheme to mnmze the sgnallng overhead and then ncrease the uplnk recepton qualty. Subsequently, based on the proposed CQI reportng scheme, we adopt an EUPS scheme for packet data transmsson on the EUL or HSUPA. The packet schedulng procedure n EUPS does not only take nto account the delay senstvty but also consder the fluctuatng channel condtons. Therefore, the deployment of EUPS n HSPA systems would Copyrght 212 ScRes.

2 S. M. ABD EL-ATTY, K. LIZOS 447 provde an opportunty for UEs n order to guarantee the QoS requrements.e., lessen average packet delay as well as ncrease cell throughput. The remander of the paper s organzed as follows. In Secton 2 related work and problem statement are nvestgated. The proposed CQI reportng scheme and the predcton mechansm are presented n Sectons 3 and 4 respectvely. In sequel, the detals of the EUPS are nvestgated n Secton 5. The smulaton results and performance evaluaton are presented n Secton 6. Fnally, we conclude ths work at Secton Related Work and Problem Statement Schedulng packet data transmsson on the enhanced dedcated channel (E-DCH) n HSPA systems can be performed by tme or code dvson scheduler [4]. In enhanced uplnk (EUL) or namely HSUPA, code schedulng grants a far rado resource wth low data rate to multple UEs contnuously whle tme schedulng allocates rado resource to dfferent UEs accordng to channel state nformaton [5]. However, the nose rse level may be ncreased at Node B above the desred level n both types. A combned scheduler between code and tme dvson schedulng s proposed n [4]. In addton, hybrd schedulng (HS) for resource allocaton n HSUPA s presented n [6]. HS conssts of traffc classfer, QoS montor, queue selector and adaptve weght factor. Although, HS provdes a QoS guaranteed servce for mxed multmeda accordng to traffc characterstcs and packet drop rate, the scheduler desgn and mplementaton s complex. On the other hand, the tme scheduler n HS does not take nto account the channel condtons of UEs. In 3GPP standards [3], the classcal schedulng (CS- 3gpp) operaton at the E-DCH s based on the schedulng nformaton (SI) and the happy bt (HB) durng the transmsson of the uplnk logcal channels when the UEs request the requred resources from the Node B [2]. Accordngly, HSUPA schedulng procedure depends on downlnk and uplnk sgnallng data. In the downlnk, a resource ndcaton (schedulng grant) s requred to notfy the UE on the lmted amount of uplnk resources t s allowed to use. Schedulng grants (SG) carry dfferent control sgnals such as maxmum allowed E-DCH Dedcated Physcal Data Channel (E-DPDCH)/Dedcated Physcal Data Control Channel (DPCCH) power rato of the actve UEs and E-DCH TFC selecton algorthm. The SG s performed once per TTI by usng absolute grant and relatve grant. In every E-DCH transmsson, UE transmts (SI) n the uplnk logcal channels contanng the followng nformaton: 1) SNPL: denotes to the path losses of the servng cell and neghbour cells, the reserved sze s 5 bts. 2) TEBS: denotes to the total buffer occupancy status of UE, the kept sze s 5 bts. 3) UPH: denotes to the power avalable to the UE, the kept sze s 5 bts. 4) HLID: denotes to the hghest logcal channel dentfcaton that s used to dentfy prorty logcal channel, the kept sze s 5 bts. 5) HLBS: denotes to the hghest prorty logcal buffer status, used for ndcatng the usage status of the hghest logcal prorty channel buffer and the percentage of the total buffer beng used, the kept sze s 5 bts. In addton to the above SI, the UE n each E-DPCCH transmsson nstance should send happy bt (HB) n order to ndcate f UE asks more resources or not. The UE wll ndcate that s unhappy when the followng states are met: UE s transmttng as much data as allowed by the current (SG); and UE has enough power avalable to transmt at hgher data rate; Total buffer states would requre more than Happy_ Bt-Delay_Condton ms to be transmtted wth the Current_Servng_Grant the rato of actve processes to the total number of processes. In sequel, the Node B scheduler can control the power transmsson and data rate to each UE. Therefore, n 3GPP standards the classcal scheduler CS-3gpp at Node B s manly based on the uplnk and downlnk sgnallng, carryng SI and HB nformaton. However, when the numbers of UEs ncrease n the servng cell, the amount of the allocated resources per UE s decreased due to the ncrease n the uplnk nterference causng degradaton of the cell throughput. On the other hand, the SI enables only the Node B scheduler to prortze the traffc flows of an UE accordng to the nformaton carred on logcal channel and the buffered data, whle t does not take nto account the delay experenced by each UE. Therefore, the classcal scheduler n 3GPP standards does not support the delay senstvty of multmeda traffc [2,3]. The man am of ths work s to mprove the uplnk recepton qualty and to reduce the sgnallng overhead by ntroducng a predcton CQI (P-CQI) reportng scheme that reveals the measured sgnal and nterference nose rato (SINR) of the actve UE. Subsequently, based on the 3GPP standards, we present an enhanced uplnk packet schedulng (EUPS) n order to serve the traffc flows of UE, takng nto account not only buffered status but also the qualty of ts channel. In addton, we consder the UE plays varous applcatons wth dfferent QoS requrements and then the EUPS prortes those applcatons based on the experenced delayed data and ther channel qualty. Therefore, the servce prorty crteron n EUPS s able to serve dfferent UE traffc flows correspondng to the delay senstvty and the reported Copyrght 212 ScRes.

3 448 S. M. ABD EL-ATTY, K. LIZOS predcted channel state. In concluson, EUPS guarantees contnuous packet connectvty wth dfferent QoS requrements of the selected UE. 3. New ng Scheme In Release 5 [7], 3GPP proposed a perodc CQI feedback scheme as llustrated n Fgure 1(a) that has a report cycle of r c. The possble values of r c are [, 2, 1, 2, 4, 8, 16] msec. The Enhanced CQI reportng (E-CQI) scheme descrbed at Release 6 extends Release 5 specfcatons by ntroducng addtonal CQI reports durng perods of downlnk actvty. As shown at Fgure 1(b) the addtonal CQI reports are transmtted wth every packet acknowledgment (ACK) (and/or Non-ACK). The am of the Enhanced ng s to use longer report cycles compared to the perodc CQI scheme and ncrease the number of the CQI reports when t s needed (.e. when the downlnk actvty ncreases). Obvously, the shorter the report cycle s the more effcent performance of HSPA, as t provdes better adaptve modulaton codng (AMC) adaptaton to the varatons of the wreless channel. However, at the same tme, the frequent CQI sgnallng ncreases uplnk nterference and thus decreases the average UE throughput as well as the achevable energy-per-bt to nose E b /I rato at the enhanced uplnk E-DCH. Therefore, we propose an mproved CQI reportng scheme whch ams to reduce the requred CQI sgnalng even when the downlnk data actvty s relatvely hgh, by employng a predcton CQI (P-CQI) scheme. Accordng to ths scheme, Node B predcts all the ntermedary CQI reports between two subsequent CQI reports by utlzng a FSMC model of the wreless channel. Thus, as shown n Fgure 1(c), n our proposed mechansm a number of CQI reports can be predcted nstead of transmtted. Adoptng the smplfed nterference analyss presented n [8], we can measure the beneft of usng P-CQI scheme. Consderng M u = 3 the number of users n the cell, we calculate an approxmate estmaton of the uplnk sgnal recepton qualty gan at the Node B. Fgure 2 shows how the gan of the E b /I rato ncreases as the rato of predcted CQI reports to the total number of CQI reports, ncreases. The Perodc-CQI scheme wth a reportng cycle of 2 ms s used as a reference base. As we can see n Fgure 2, the recepton qualty of HSUPA mproves as the number of CQI predctons ncreases. Although a hgh rato of CQI predctons may be mpractcal n a real system, as predcton accuracy decreases NodeB cycle UE a) Perodc CQI reportng scheme cycle NodeB Data ACK Addtonal CQI Report UE b) Enhanced CQI reportng scheme cycle NodeB CQI predcton at Node B CQI predcton at Node B CQI predcton at Node B CQI predcton at Node B UE c) Predcton based CQI reportng scheme Fgure 1. CQI reportng schemes. Copyrght 212 ScRes.

4 S. M. ABD EL-ATTY, K. LIZOS 449 Fgure 2. E b /I gan vs ncreasng rato of CQI predctons n HSPA. when the number of predcted samples s ncreased, we can conclude from Fgure 2 that even by predctng, and thus avodng, the transmsson of just one every two CQI reports we can acheve a sgnfcant gan of approxmately 1.5 db. 4. Predcton Mechansm Based-CQI Scheme The CQI reports ndcate the requested TFC and thus they reflect ther current channel condtons. Therefore, they can be nterpreted to SINR measurements. The obtaned SINR values are then used by the P-CQI scheme to predct, through a FSMC model, the next state of the wreless channel and therefore the next CQI. Consequently, predctng the next CQI at the Node B helps n reducng the number of the requred CQI reports. Modelng Wreless Channel Based-FSMC We consder a multpath fadng envronment, n whch the receved SINR (γ ) of an actve UE s proportonal to the square of the sgnal envelop. The probablty densty functon (pdf) of γ can be expressed as n [9]. p 3 e 1, where = { } s the mean value of SINR, whch can estmated by averagng the SINR of all M u actve UEs n the E-DCH cell as follows: j jmu M u Accordng to the FSMC model depcted n Fgure 3, can be parttoned nto K non-overlappng ntervals by thresholds k, k {, 1, 2,, K}, the wreless channel of actve UE s consdered to be n state S k f the measured les n the nterval { k, k+1 }. Assumng that the channel fades slowly wth respect to CQI feedback report cycle (r c ) and Doppler shft f d n the carrer frequency f c, (1) (2) Fgure 3. P-CQI scheme n HSPA system. s f d = vf c /c, then probablty of transmsson n state k, can be obtaned as follows [9]: K 1 k k1 πk p d e e (3) K Takng nto account the prevous equatons as well as the fundamental theory of Markov chans and snce the summaton of all transton probabltes equals 1, t s possble to determne the steady state transton probabltes from state k to k + 1 or k 1 or remanng n state k. Hence, the derved transton probablty of transmssons at the FSMC model can be approxmated by the followng equatons: k 1 1 2π k 1 pkk, 1 fdts e πk (4) kk, 1 d s πk k 1 2π k p f T e (5) p 1 p p (6) kk, kk, 1 kk, 1 Therefore, the predcted state for each UE can be transmtted va the hgh-speed dedcated physcal control channel (HS-DPCCH) n the uplnk every TTI (2 ms). Snce k s the current state of the channel, computed by a SINR report, the next transton state, namely m, accordng to the hghest state transton probablty of a Markov chan model s as follows: p max p, p, p (7) km, kk, kk, 1 kk, 1 More future states may be predcted n the same manner usng the same calculated transton probabltes. The Copyrght 212 ScRes.

5 45 S. M. ABD EL-ATTY, K. LIZOS state transton probabltes are updated perodcally based on the real SINR levels receved from the UEs at each tme the real CQI measurements are collected. By ths method the UE s able to select the approprate applcaton for transmsson based on the reported state. For nstance, as shown n Fgure 3, when Node B reports the predcted state (S 1 ) to UE, the UE selects the TFC of the VoIP applcaton and so on. 5. Enhanced Uplnk Packet Schedulng (EUPS) Scheme The am of proposed scheme s to reduce the CQI sgnalng overhead and thus mprove the sgnal recepton qualty at the enhanced uplnk (EUL). Therefore, the prncple of EUPS procedure on E-DCH s governed by the proposed P-CQI reportng mechansm and at the same tme the operaton of EUPS s stll certfed on 3GPP standards. We consder that the UE s able to play dfferent applcatons of multmeda traffc whle UE can be notfed to send SINR measurement reports perodcally to Node B durng the roamng n the servng cell. Accordng to the measured SINR, the EUPS predcts the next state channel by usng the P-CQI scheme as shown n Fgure 3. The UE sends the requred data rate of each applcaton to Node B through grant request (GR). Dependng on the receved GR and the predcted states of UE channels, the EUPS examnes the avalable resources, computes the Transport Format Combnaton Set (TFCS) and sgnals the TFCS to the UE va grant allocaton (GA). TFCS s the set of Transport Format Combnatons (TFCs) that can be used by the UE for data transmsson. On the other hand, EUPS employs a prorty crteron (PC) at Equaton (8) based on the delay senstvty and the channel qualty for data transmsson of each flow n the E-DCH. PC s defned as the rato of Head of Lne packet delay over the packet delay threshold of a flow. In addton, we consder the successful probablty of flow transmsson as a dscrmnated prorty between the delayed flows. Therefore, the servce of the hghest prorty logcal channel ID (HPLC) of flow transmsson durng each TTI (2 ms) can be expressed as follows: P, 1, 2, 3 PC HPLC HOL s (8) D th, where HOL denotes the Head of Lne packet delay of flow and D th, s the delay threshold of packets for flow 1, 2, 3 that referred to VoIP, Vdeo, and Web browsng respectvely. P s, s the successful probablty of bt transmsson durng the next sub-frame and s determned through predcton mechansm of the wreless channel. The schedulng procedure of EUPS s performed every TTI correspondng to the servce prorty. The EUPS steps as follows: 1) Set a lst of UEs prortzaton accordng to PC, solve tes based on P s. 2) Select the hghest prorty UE from the lst and assgn a Schedulng Grant (SG) and, 3) Exclude any UE, not beng able to get SG from the current TTI, due to power lmtatons. 4) For all the remanng of UEs n the prortzaton lst, share the resdual capacty accordng to the followng steps: Set UEs n decreasng prorty lst accordng to TEBS. Solve tes based on P s. Assgn a SG accordng to the hghest power transmsson. End the assgnment of SG when the resdual capacty s not suffcent. Fnally, data transmsson takes place on the uplnk E-DCH Dedcated Physcal Data Channel (uplnk E- DPDCH). Subject to the amount of data n the buffer and on the requred transmsson power, the UE selects the most approprate TFC between those ncluded n the TFCS. In addton, EUPS makes use of computng P s, of each flow, to avod erroneous packet transmssons and thus ncreases the bandwdth utlzaton. Thus computng P s, s preferable nstead of employng HARQ and power headroom requrng hgher power requrements, especally when channel condtons have worsened. Therefore, EUPS guarantees contnuous packet connectvty to UE correspondng to the predcted state reported va HS- DPCCH. 6. Smulaton and Performance Evaluaton The performance evaluaton of the proposed scheme based on the P-CQI predcton mechansm s drven by developng a system level smulator for enhanced uplnk n 3GPP. We compare the performance of EUPS wth a conventonal scheduler based on 3GPP standards (CS- 3gpp) [3] and the hybrd schedulng (HS) algorthm [6]. We evaluate the performance of EUPS n terms of average packet delay, average packet drop rate and cell throughput. The smulaton scenaro was saturated wth realstc parameters to reflect forecasted growth of UE future traffc requrements. We consder a cell wth a radus of 1 km. Node B s located at the centre of the cell. Each UE s unformly dstrbuted n the cell and can play mxed traffcs. A mxed multmeda traffc ncludes VoIP, Vdeo and Web browsng [1] wth average data rates 2 kpbs, 6 kpbs, and 9 kbps respectvely. The proportons of the multmeda traffcs for VoIP, Vdeo and Web browsng are 7%, 2% and 1% respectvely. As we mentoned prevously, the operaton of EUPS s based on the proposed P-CQI scheme correspondng to Copyrght 212 ScRes.

6 S. M. ABD EL-ATTY, K. LIZOS 451 the nstantaneous value of SINR of UEs, then the nterference generated at neghbourng cells can be reduced, and cell throughput s ncreased. Furthermore, for the evaluaton the average packet drop rate, we consder the average packet error probablty (PER) when the applcaton s transmttng, depends solely on the bt errors and the data rate. It can be expressed as follows: R s, PER 1 P (9) where R s the transmsson rate of flow and P s, s computed as follows: 1 k 1 Ps, 1 pe m p d π (1) k k where p em () denotes the error probablty for a specfc modulaton scheme n AMC map. By computng the PER at each TTI, EUPS can determne the average packet drop rate (P d, ) of traffc flow as follows: P d, No.of transmtted packets 1PER total No. of transmtted pacekts (11) Smulatons have ntally been run for an average number of UEs per cell equal to 5 and for a pedestran profle at 3 km/h [11]. The sesson arrval rate s modelled by a Posson process. The traffc load ncreases by ncreasng the number of UEs n the servng cell. The ntal locaton of each UE s randomly dstrbuted n the cell; the drecton of UE movement s unformly dstrbuted. The macrocell propagaton model s adopted for calculatng the path loss at dstance d (km) from Node B. Therefore, the attenuaton L p of the transmtted sgnal for a Node B antenna heght of 15 meter and 2.1 GHz carrer frequency s defned as L p (d ) = log 1 (d ) [db]. The modellng of the wreless channel of UE s performed through a four-state FSMC whch provdes the requred accuracy wthout addng excessve complexty. We employed equal probablty method (EPM) [9] for estmatng the steady state probabltes and afterwards the transton probabltes. The CQI sgnallng delay s not taken nto account n the process of evaluatng packet delay and t s stll open ssue for studyng n HSPA systems. At the uplnk, the UE s data are transmtted through E-DCH Dedcated Physcal Data Channel (uplnk E- DPDCH) wth the selected TFC whle the HS-DPCCH s employed for fast power control and sgnalng data. The detaled smulaton parameters are summarzed n Table 1. We now present a snapshot of our smulaton results, whch provde farly nterestng nsght and performance effcency compared to the prevous works [3,6] Average Packet Delay Comparson Fgures 4 and 5 show the average packet delay of VoIP Table 1. Basc smulaton parameters n HSPA systems. Parameter Carrer Freq. Ch. Bandwdth TTI Cell Radus UE dstrbuton Channel Models Traffc Models Path Loss Model Thermal Nose UE Tx power Power allocaton Tx power Other-to-Own cell nterference rato Value 2.1 GHz 5 MHz 2 ms Hexagonal, 1 km Unform Raylegh VoIP, Vdeo, Web browsng L p (d ) = log 1 (d ) [db]. 174 dbm/hz 2 dbm (.1 Watt) Control 2% E-DPDCH 8% from total power Iteratons per smulaton run 7 Smulaton tme.5 36 sec Fgure 4. Average packet delay for VoIP traffc. Fgure 5. Average packet delay for Web browsng traffc. Copyrght 212 ScRes.

7 452 S. M. ABD EL-ATTY, K. LIZOS and Web browsng traffc flow respectvely wth ncreasng the number of UEs per cell. As expected, by ncreasng the number of actve UEs n the cell, the average packet delay s ncreased for both VoIP and Web browsng traffc servces. However, by employng EUPS a sgnfcant delay reducton n both cases s acheved. Ths occurs on the bass of the prorty crteron (PC) where t can sort the hghest prorty logcal channels not only accordng to the delay senstvty but also accordng to the reported predcted channel state of each UE. In addton, f two UEs have the same delay, EUPS serves the UE wth the hgher successful transmsson probablty (P s ). Therefore, the number of UEs awatng servce s decreased. Ths leads to a reducton of the delayed packet n the UEs buffers. For comparson reasons, assumng an average number of UEs = 25 n the cell as shown at Fgure 4 of VoIP traffc, the average packet delay s reduced by approxmately 11% compared to HS and CS-3gpp schemes. Fgure 6. Average packet drop rate for VoIP Average Packet Drop Rate Comparson Fgures 6 and 7 show the average packet drop rate, for VoIP and Web browsng traffc servces respectvely wth ncreasng the number of actve UEs per cell. As llustrated n the Fgures 6 and 7 the EUPS scheme s capable of reducng the packet drop rate of the scheduled traffc flow for VoIP or Web browsng compared wth CS-3GPP and HS dscplnes. Also besdes the reducton of delayed packets, EUPS explots the computng of P s n order to mnmze the retransmsson of erroneous packet and hence the average packet drop rate s sgnfcantly decreased. Subsequently, the bandwdth utlzaton s ncreased and the cell throughput s ncreased compared to CS-3GPP and HS dscplnes as shown n Fgure 8. To summarze the performance of the proposed scheme, EUPS s able to sort the traffc flows not only based on the delay senstvty but also takes nto account the reported channel states of UEs. Therefore, the packet delay and packet drop rate are sgnfcantly decreased. As well as when UE plays dfferent multmeda traffc, EUPS can provde QoS guaranteed for contnuous packet connectvty accordng to reported channel state. 7. Concluson In ths paper, we have proposed an enhanced uplnk packet schedulng (EUPS) scheme for multmeda applcatons packet data transmsson n HSPA systems. EUPS operaton based on an mproved predcton CQI reportng scheme. The smulaton results show that the performance effcency of HSPA systems employed EUPS s superor compared to the conventonal scheduler n 3GPP standards. Fgure 7. Average packet drop rate for web browsng traffc. Fgure 8. Comparson of cell throughput. 8. Acknowledgements A part of ths work s carred out n Communcaton LAB at Faculty of Electronc Engneerng, Menoufa Unversty, Egypt n the last summer. REFERENCES [1] HSPA Brngs the World, [2] H. Holma and A. Toskala, HSDPA/HSUPA for UMTS Copyrght 212 ScRes.

8 S. M. ABD EL-ATTY, K. LIZOS 453 Hgh Speed Rado Access for Moble Communcatons, John Wley & Sons Ltd., Hoboken, 26. do:1.12/ [3] 3GPP TS V7.2., Enhanced Uplnk; Overall Descrpton; Stage 2 (Release 7), 27. [4] C. Rosa, J. O. Carnero, T. B. Sørensen, J. Wgard and P. E. Mogensen, Combned Tme and Code Dvson Schedulng for Enhanced Uplnk Packet Access n WCDMA, IEEE Proceedngs of the 6th Vhecular Technology Conference, Los Angeles, September 24, pp [5] D. C. Dmtrova, J. L. van den Berg, G. J. Hejenk and R. Ltjens, Flow Level Performance Comparson of Packet Schedulng Schemes for UMTS EUL, Proceedngs of 6th Internatonal Conference on Wred/Wreless Internet Communcatons, Tampere, 28-3 May 28, pp [6] Y. J. Kang, J. Km, D. K. Sung and S. Lee, Hybrd Schedulng Algorthm for Guaranteeng QoS of Real- Tme Traffc n Hgh Speed Uplnk Packet Access (HSUPA), IEEE 18th Internatonal Symposum on Personal, Indoor and Moble Rado Communcatons, Athens, 3-7 September 27, pp [7] 3GPP TR V5.., Hgh Speed Downlnk Packet Access: Physcal Layer Aspects (Release 5), 22. [8] N. Fuku, Study of Channel Qualty Feedback n UMTS HSDPA, 14th IEEE Proceedngs on Personal, Indoor and Moble Rado Communcatons, 7-1 September 23, pp [9] Q. Q. Zhang and S. A. Kassam, Fnte-State Markov Model for Raylegh Fadng Channels, IEEE Transactons on Communcaton, Vol. 47, No. 11, 1999, pp do:1.119/ [1] 3GPP TR V6.., Feasblty Study for Enhanced Uplnk for UTRA FDD, 24. [11] ETSI, Unversal Moble Telecommuncatons System (UMTS); Selecton Procedures for the Choce of Rado Transmsson Technologes of the UMTS, TR , V3.2., Copyrght 212 ScRes.