A Utility-Approached Radio Resource Allocation Algorithm for Downlink in OFDMA Cellular Systems

Transcription

1 A Utility-Approached Radio Resource Allocation Algorithm for Downlin in OFDMA Cellular Systems Lue T. H. Lee Chung-Ju Chang Yih-Shen Chen and Scott Shen Department of Communication Engineering National Chiao Tung University Da-Shieh Rd. Hsinchu 3 Taiwan { lue.cm92g@nctu.edu.tw. cjchang@cc.nctu.edu.tw atlas@bn3.cm.nctu.edu.tw brahms.cm87g@nctu.edu.tw } Abstract In this paper we propose a utilityapproached radio resource allocation () algorithm for downlin transmission in orthogonal frequency division multiple access (OFDMA) cellular systems. The proposed utility-approached algorithm is to efficiently allocate the radio resource to mobiles according to a utility function which contains a rate-power function a QoS function and a priority function to ensure QoS requirements among heterogenous services and improve the system throughput. Also a simulatedannealing (SA) method is applied to find out the solution within a reasonable time. Simulation result shows that the proposed utility-approached algorithm can allocate the radio resource efficiently and guarantee the QoS requirements of each user under the power constraints of the OFDMA system. I. INTRODUCTION The orthogonal frequency division multiplexing (OFDM) technique is a promising modulation scheme for high data rate wireless communications. The main idea of the OFDM scheme is to split a high rate data stream into a number of lower rate streams and transmit them by a set of orthogonal subcarriers. The OFDM system possesses many outstanding properties such as: () flat fading per subcarrier (2) comparatively short inter-symbol interference (ISI) (3) short equalizers needed and (4) maximum spectral efficiency (Nyquist rate). In OFDM systems when the channel state information (CSI) is available at the transmitter the radio resource allocation () for each subcarrier can be adapted according to the CSI in order to increase the data rate []-[6]. Among many OFDM-based access networs OFDM- FDMA (OFDMA) is considered in this paper for its easy implementation. For OFDMA it had been proved that the data rate of a multiuser OFDM system is maximized when each subcarrier is assigned to only one user with the best channel gain and the transmit power is distributed over the subcarriers by a water-filling policy []. Some wors focused the in OFDMA systems including subcarrier rate and power allocation [2-6]. The goal of these papers was to minimize the total transmission power under satisfying users QoS requirements. Minimum transmission rate is considered as a QoS parameter in [3]-[6] while bit error rate (BER) is additionally considered in [2]. In this paper we further consider more sophisticated QoS requirements to accommodate multiple services such as the required transmission rate required BER delay bound and pacet dropping ratio for real-time (RT) services and the minimum transmission rate and required BER for non-realtime (NRT) services. Accordingly a utility-approached algorithm for a downlin OFDMA cellular system is proposed which allocates transmission power rate and subcarriers to mobile users according to their CSI and QoS requirements. Simulation results show that the proposed utility-approached algorithm has lower mean delay and pacet dropping ratio of RT users than a lin gain-based algorithm which only allocates the radio resource according the lin gain performance among users while not considering the users QoS requirements. The rest of this paper organizes as follows. The operation of the OFDMA system is described in section II. The problem formulation and SA-based scheme are presented in section III and IV respectively. The simulation results and discussions are shown in section V. Finally concluding remars are given in section VI. II. SYSTEM MODEL An OFDM transmitter separates a symbol of serial bits within symbol time Ts into a parallel form; i.e. each bit will be transmitted within the same Ts and fed into the subcarrier different from each other. In the discrete time domain an OFDM modulator can be easily implemented by inverse fast Fourier transform (IFFT) and a parallel-to-serial (P/S) converter. In order to cancel the inter-carrier interference of a subchannel due to multipath a cyclic prefix (CP) is attached which is a copy of the last part of OFDM signal to the front of itself. In contrast to OFDM transmitter an OFDM receiver removes the CP first and then converts serial bits into parallel form for performing the fast Fourier transform by a demodulator. Finally the output bits from FFT bloc will be fed into the de-mapper bloc to restore the original bits and P/S conversion is followed. In this paper users are served in the downlin OFDMA cellular system with N subcarriers. Perfect estimation of the channel state for a mobile is assumed. To increase spectrum efficiency the adaptive modulation is applied to each subcarrier with a suitable M-QAM symbol according to CSI. Fig. shows the downlin OFDMA system with the proposed utility-approached algorithm. At the transmitter of a base station (BS) users data streams are allocated to subcarriers by the utility-approached algorithm according to the QoS requirements and the CSI of all mobiles. We assume that there is an OFDM shared control channel for transmitting radio resource allocation information from BS to all mobiles and for transmitting the QoS requirements and the CSI from all mobiles to BS. The radio resource allocation information includes the results of power allocation subcarrier allocation and the adaptive modulation order for each mobile. The CSI contains the lin gain and SIR of each subchannel from BS to mobile which can be estimated through the OFDM common pilot channel at the receiver by mobiles [7]. In the wireless environment the bandwidth of each subchannel should be smaller than the coherence bandwidth of the channel to overcome the frequency-selective fading. correlation between each subchannel is assumed and /5/$2. (c)25 IEEE

2 each subchannel is a flat-fading channel with the lin gain α ( n) (db) which is given by α ( n) FL + FS ( n) () d s FL log( ) PL( d ) + X (2) σ d where n is subchannel index and is user index F L is largescale fading of the -th user due to free-space degrading and shadowing effect [8] d is the distance between BS and mobile d is the close-in reference distance PL( d ) is the mean path loss at the reference distance d s is the path loss exponent depending on the surroundings and building type and Xσ is the shadowing effect with a normal random variable having a standard deviation σ. FS ( n) is the smallscale fading of the -th user in the n-th subchannel with Rayleigh distribution due to multipath. All the subchannels have the same large-scale fading for one mobile at a random time instant. QoS Requirement of All Mobiles st user traffic n-th user traffic Utility- Approached algorithm Adaptive Modulator for Subchannel Adaptive Modulator for Subchannel N Channel State Information from All Mobiles -th user traffic Bits and Subcarrier Extraction Adaptive Demodulator for Subchannel Adaptive Demodulator for Subchannel N Resource Allocation Information from BS OFDM Modulator Transmitter of BS OFDM Demodulator Channel for Mobile + AWGN Receiver of Mobile Fig. Downlin OFDMA System with utility-approached algorithm III. PROBLEM FORMULATION Four QoS requirements are considered in the proposed utility-approached algorithm for multimedia services. For RT services required transmission rate R required error rate BER delay bound D and pacet dropping ratio P are D taen into account. On the other hand minimum transmission rate R and BER are required for NRT min interactive services while only BER is required for NRT best-effort services. Also adaptive modulation is used according to the quality of each subchannel to improve throughput. The utility-approached algorithm will allocate subcarriers transmission power and rate according to a utility function whose parameters are CSI long-term QoS requirements and priority among mobiles. The utility function is defined as [9] U ( ) ( ) ( ) ( ) nm Rnm Q P (3) where R( nmis ) a power-rate function of the -th mobile on the n-th subchannel with modulation order m Q ( ) is a QoS function of the -th mobile and P( ) is a priority function that gives RT services higher subcarrier allocation probability over NRT services. R( nm ) derived from the BER of M-QAM in AWGN channel [] is given by pnm Rnm ( ) B log 2 ( + β γ n) (4) q where B is the bandwidth of a subchannel and β.5 / ln(5 BER) is the SIR gap p represents the nm power allocated to -th user with n-th subcarrier and modulation order m q is pilot power and γ n is the SIR measured from pilot signals. Q ( ) is defined as Q( ) log( PD ) D ( ) ( ) t D t D exp{ / 2 + D( Lˆ ( L( exp{ } / 2 + L( where () time t () for NRT best - effort service } for RT service for NRT inter - active service (5) D t is the transmission delay of the -th mobile at D t is the weighted mean delay Lˆ () t is a normalized measurement on the difference of R and min R () t and L() t is the average value of Lˆ () t. D() t Lˆ () t and L() t are expressed as log PD Dt () ( ) (6) D Lˆ ( RT RT ˆ Rmin R ( t T) λ L ( ) + ( ) t T λ (7) Rmin L() t Lˆ () t (8) NRT NRT P( ) is defined as ρrt for RT services P ( ) (9) ρnrt for NRT services where ρ RT is larger than ρ NRT so that RT services can get higher utility than NRT services when their R( nm ) Q ( ) are the same. The goal of the utility-approached algorithm is to maximize the summation of all utilities which is expressed as follows N Max U ( n m) δ ( n m) n m M ()

3 N p n m δ( n m) ptotal n m M () st.. pnm δ( n m) pmax for all n m M N Rnm ( ) δ( nm ) R for RT n m M where δ ( nm ) is the n-th subcarrier allocation function and for subcarrier n mobile δ ( nm ) and modulation order m (2) otherwise. IV. SIMULATED ANNEALING ALGORITHM There would be 2 NM combinations to find out the optimal solution of the proposed utility-approached algorithm while exhausted search algorithm is applied. The exhausted search algorithm is not practically feasible due to its long computation time. Therefore a powerful search method to reduce computation time is necessary. Simulated annealing (SA) method is adopted in this paper to solve the optimization problem () under the constraints (). The theoretical analysis of SA algorithm can be found in [] for further reading. In order to adapt SA algorithm to problem (4) is rearranged as ( m ) q pnm (3) β γ n The optimal subcarrier power vector is defined as p [ pˆ pˆ pˆ ] where ˆ 2 N pˆ p p2 p3 p m p M }; pˆ 2 p2 p22 p23 p 2 m p 2 M }; (4) pˆ N p N p N 2 p N 3 p N m p N M }. ˆp can be obtained by iterating from the randomly initial subcarrier power vector p with ( M) N possible combinations. Furthermore given the constraints () p should be carefully selected such that it will not fall into an unreasonable region. Fig.2. shows the complete flowchart of the utilityapproached algorithm using the SA method and the detail discusses are as following. Step : [Initialization] First of all initial temperature T max final temperature T min decrement rate of temperature γ the maximum number of iteration at a temperature level L max are set with appropriate values. Then the initial subcarrier power vector p is selected randomly among users of NRT service without violating the constraints (). The SA algorithm is insensitive to the initial state so choosing a reasonable p is good enough for the fast convergence. L Select one neighbor solution as the next state pj LL+ Calculate accpetance function f Receive the next state? Update state LLmax? Temperature annealing (TζT ) L Received CSI from all mobiles Set SA parameters (Tmax Tmin Lmax) Set pn{pn.4 n is maximum for all where is NRT service} n2...n TTmin? End Initialization Selection and Comparison Annealing Fig.2 Flowchart of SA algorithm Step2: [Selection and Comparison] The object function Uj for next state j is defined as N U U ( n m) δ ( n m) (5) j n m M Our goal is to find an optimal subcarrier power vector ˆp to maximize U j ; i.e. the throughput is maximized and the constraints () are satisfied at the same time. In Step 2 the subcarrier power vector of next state j p j is randomly selected from the neighborhood of the subcarrier power vector of present state i p i. The SA algorithm will determine whether to accept p j or not according to the acceptance function Uij ft( Uij) Min[exp( )]. (6) T p j will be absolutely accepted if U j is larger than U i ; otherwise it will be accepted with a probability. Step3: [Annealing] When the number of iteration at a temperature level exceeds L max the temperature then will be decreased by the decrement rate ζ. The procedure will not be terminated until the temperature decreases down to the final temperature T min. At the initial stage of the SA algorithm higher T max is appropriate to increase the probability of accepting a worse next state to prevent from falling into suboptimal solution. Large decrement rate ζ reduces the convergence time but

4 suboptimal solution may be achieved; on the contrary small ζ performs better at the expense of convergence time. V. SIMULATION RESULTS AND DISCUSSIONS Parameter Symbol Value Thermal noise density N -28 dbm/hz Reference distance d m Cell size d max m Fading factor s 4 Reference path loss PL(d ) 6 db Shadowing factor Xσ mean: variance: 8 Small-scale fading Fs mean: variance: 2 / π Number of subcarrier N 25 Central frequency f 2 GHz OFDM symbol period Ts 8.333μs Subcarrier spacing B 2 Hz Frame time Tf. ms Useful modulation order m QPS 6QAM 64QAM Maximum subcarrier p max.36 watt transmission power Maximum total transmission power p total 35 watt Table System parameters Table I shows all the system parameters of the OFDMA system in this simulation. For the effectiveness of SA method the initial temperature T max should be set at least 85% of accepting a worse next state to avoid trapping into a local optimal solution while the final temperature T min should be set at most.% of accepting a worse next state. Both of the two temperature parameters can be obtained by (6). Furthermore the decrement rate γ set within is suggested and the maximum number of iteration L max can be set according to the present temperature and the number of decision variables. In our simulation the number of RT users is 2 and the number of NRT users is ranging from to 2. The data stream of each user will be segmented into pacets of 5- bit unit and the delay of the -th RT user D () t is measured as the mean delay of each pacet which means the average time of each pacet stays in queue. However a pacet of the -th RT user will be dropped if the time that it stays in queue is longer than the corresponding delay bound D. Also the traffic models for RT interactive NRT and best-effort NRT users are modeled by ON-OFF process Pareto process and batched Poisson process respectively. For performance comparison a lin gain-based algorithm is considered which allocates subcarriers power and rates according to the lin performance of each user. Fig. 3 shows the average throughput versus the number of NRT users. It could be seen that the proposed utilityapproached algorithm has lower average throughput than the lin gain-based algorithm. The lin gain-based algorithm employs the water-filling policy for resource allocation to achieve higher system throughput at the cost of QoS degradation. However QoS fulfillment of multimedia services is more critical than throughput from the point of view of the OFDMA system operators. Fig. 4 shows the mean delay of RT users and fig. 5 shows the mean pacet dropping ratio versus the number of NRT users. From the two figures it could be found that the proposed utility-approached algorithm achieves superior QoS performance (lower mean delay and pacet dropping ratio) over the lin gain-based algorithm. The proposed scheme has lower delay than the lin gain-based algorithm by an amount of 5%. Additionally there is a interesting phenomenon shown in Fig. 3 that the average throughput peas when the number of NRT users is 9 for the utility-approached algorithm. It implies that there is a capacity bound of the OFDMA system under the QoS consideration; therefore a call admission control (CAC) scheme should be accompanied with the design of scheme to prevent from system overloading. Throughput (Mbps) Lingain-Based Fig. 3 Averaged throughput vs. number of NRT users Mean Delay of RT Users (.sec) Lingain-Based Fig. 4 Mean delay of RT users vs. number of NRT users

5 Pacet Dropping Ratio Lingain-Based Fig. 5 Mean pacet dropping ratio vs. number of NRT users VI. CONCLUDING REMARS In this paper a utility-approached algorithm for the downlin OFDMA cellular system is proposed. Different from previous wors we specifically consider the mean pacet delay bound and the pacet dropping ratio of RT users as the QoS requirements. To reduce the computation time of searching an optimal solution of the proposed algorithm the SA method is applied to find the solution effectively. From the simulation results it could be seen that the proposed utility-approached algorithm provides lower pacet delay and lower pacet dropping ratio of RT users than the lin gain-based algorithm does. Our future wor is two-folded: first a sophisticated CAC scheme will be designed to maximize the system throughput and fulfill the QoS requirements; second the proposed scheme will be extended to the multi-cell OFDMA environment with the consideration of cell-level coordination. [5] H. S. im J. S. wa J. M. Choi and J. H. Lee Efficient Subcarrier and Bit Allocation Algorithm for OFDMA System with Adaptive Modulation. IEEE Vehicular Technology Conference (VTC24-Spring) May 24. [6] G. Song and Y. Li Adaptive Subcarrier and Power Allocation in OFDM Based on Maximizing Utility IEEE Vehicular Technology Conference 23-Spring (VTC23-Spring) vol. 2 April [7] S. Cinem M. Ergen A. Puri and A. Bahai Channel Estimation Techniques Based on Pilot Arrangement in OFDM Systems. IEEE Trans. On Broadcasting vol. 48 issue: 3 pp Sept. 22. [8] T. S. Rappaport Wireless Communications: Principles and Practice. Upper Saddle River NJ: Prentice Hall 996. [9] S. Shen C. J. Chang Generalized Scheduling Algorithm with Differentiated QoS Provisioning for Multimedia CDMA Cellular Networs IEEE Vehicular Technology Conference (VTC24-Spring) May 24. [] A. J. Glodsmith S. G. Chua Variable-Rate Variable- Power MQAM for Fading Channels IEEE Trans. On Commun. vol. 45 issue: pp Oct [] F. I. Romeo Simulated Annealing: Theory and Appplications to Layout Problems. Memorandum. UCB/ERL M89/29 March 989. ACNOWLEDGEMENT This wor was supported by National Science Council Taiwan under contract number NSC E-9- and Ministry of Education of Taiwan under Grants 9-E-FA REFERENCE [] J. Jang and. B. Lee Transmit Power Adaptation for Multiuser OFDM Systems IEEE J. Select. Areas Commun. vol. 2 no.2 pp.7-78 Feb. 23. [2] M. Ergen S. Coleri and P. Varaiya QoS Aware Adaptive Resource Allocation Techeniques for Fair Scheduling in OFDMA Based Broadband Wireless Access Systems IEEE Trans. Broadcasting vol. 49 no. 4 pp Dec. 23. [3] C. Y. Wong R. S. Cheng. B. Letaief R. D. Murch Multiuser OFDM with Adaptive Subcarrier Bit and Power Allocation IEEE J. Select. Areas Commun. vol.7 no. pp Oct. 999 [4] D. ivanc G. Li and H. Liu Computationally Efficient Bandwidth Allocation and Power Control for OFDMA IEEE Trans. Wireless Commun. vol. 2 issue: 6 pp v. 23.