Abstract-This paper proposes an enhanced rate adaptive resource allocation scheme in ... high data rate applications and services, next generation broadband ...
Enhanced Rate Adaptive Resource Allocation Scheme in Downlink OFDMA System Junhong Hui
Yongxing Zhou Samsung Advanced Institute of Technology Beijing Branch, China yongxing.zhou(samsung.com
Samsung Advanced Institute of Technology Suwon, Korea junhong.hui(samung.com Abstract-This paper proposes an enhanced rate adaptive resource allocation scheme in an OFDMA (Orthogonal Frequency Division Multiple Access) downlink transmission system. It proves to be a great improvement on capacity over the latest available rate adaptive OFDMA methods, which seek to maximize the total system capacity subject to the total power and BER constraints. We introduce a priority based sequential scheduling criteria instead of random numerical ordering during the first time of subcarrier allocation, which provides higher capacity by exploiting the multiuser diversity gains. Thus it can be seen as an efficient means to improve the system spectrum efficiency through adaptive resource allocation of OFDMA.
Keywords-adaptive resource allocation; OFDM; OFDMA; multiple access; multiuser diversity. I.
INTRODUCTION
With the spectacularly increasing demand of customer for high data rate applications and services, next generation broadband multimedia communication is coming earlier than ever expected. Supporting large data rates with sufficient robustness to radio channel impairment requires careful choosing of modulation technique. OFDM has already been widely accepted by industry leaders as a suitable solution for the high speed and broadband wireless system due to its robustness against frequency selective fading and narrowband interference like inter symbol interference (ISI) and inter carrier interference (ICI). For example, not only the IEEE 802.11 a/g wireless LAN and DSL system in production today, but also the digital TV standards, and UWB in development now all exclusively choose OFDM as the air interface. 3GPP activities including 3.9G and 4G cellular standards are all likely to be OFDM-based. As an extension, OFDM could be used more than just a modulation scheme, but as part of the multiple access technique as well, namely OFDMA (Orthogonal Frequency Division Multiple Access), which is also referred to as multiuser OFDM. OFDMA as one of the multiplexing techniques for the future broadband wireless communication system recently has attracted rising research interests. In OFDMA, the subcarriers are efficiently granulated into many small sub-channels and each user is assigned a fraction of the available subchannels. OFDMA signal splits a single high rate data stream transmitted over a carrier frequency into a number of low rate parallel data streams transmitted simultaneously over a number of orthogonal subcarrier channels. In OFDMA, these orthogonal
subcarrier spacing is a multiple of 1/T instead of the normal 2/T where T is the symbol period. Reducing the subcarrier spacing by half makes OFDMA a highly spectrum efficient multiplexing scheme. To achieve high transmission rate depends on the ability of the OFDMA system to provide efficient and flexible resource allocation scheme appropriately adapted to the changing channel conditions. Since in OFDMA, for a specific subcarrier, if a user experiences deep fading, the others may not be in deep fading, we can choose the user in a good channel condition to transmit data on that subcarrier and this eventually results in multiuser diversity effects. Recent studies although not so many on resource allocation demonstrate that OFDMA technique together with DCA (dynamic channel allocation) can significantly improve the system capacity [1][2] and thus is envisioned to be a potential alternative of the wireline broadband access techniques such as copper line, xDSL and cable modem. Margin adaptive and rate adaptive are two classes of adaptive resource allocation schemes being investigated currently. The margin adaptive allocation [1][2] is to minimize the overall transmit power given a set of the fixed user data rates and BER while the rate adaptive allocation [3]-[7] is to maximize the total data rate of all users subject to power and BER constrains. In [3], the author proposed a transmit power adaptation scheme which only transmit symbols through the subcarriers with the best gains among multiple users. However some users might have no best subcarrier thus being starved. In [4], the author considered a scheme by maximizing the minimum user's data rate, which guarantees the minimum quota for each user and partially allows fairness among users. Next, in [5] the fairness among users is taken into account by maximizing the total system capacity under the constraint of user rate proportionality. But the author in [5] used iterative method for root finding of non-linear equation which is complex and time consuming. Therefore, based on the proportional resource allocation method of [5], in [6] a low complexity non-iterative rate adaptive subcarrier allocation scheme that linearizes the power allocation problem and yields higher total capacity than the previous schemes while achieves approximate rate proportionality was proposed. But the author in [6] just used random ordering during the initial subcarrier assignment, favoring the users who are able to choose their best subcarrier earlier than others and vice versa. Furthermore, most available schemes only discuss the case when the number of subcarriers is much larger than the number of users.
0-7803-9392-9/06/$20.00 (c) 2006 IEEE 2464
In this paper, we introduce a priority based sequential scheduling criterion instead of random numerical ordering during the initial subcarrier allocation and then utilize the rate adaptive resource allocation scheme similar to [6] satisfying approximate rate proportional fairness function. From the simulation results, the OFDMA downlink system capacity could be greatly improved by our proposed rate adaptive resource allocation scheme based on priority sequential scheduling criterion [8]. Besides, this scheme has no constraint that the number of subcarriers should be much larger than the number of users, which benefit it to have wider application scenarios. This paper is organized as follows: In section II we describe the OFDMA system model and capacity optimization objective function. Then in Section III the proposed rate adaptive subcarrier allocation method is demonstrated and corresponding simulation results are also shown. Finally in Section IV conclusions are drawn.
Controller at base station Channel information feedback from user 1
,
>
from user 2
Resource Allocation
Algorithm
II.
SYSTEM MODEL We consider a typical point to multipoint downlink OFDMA system and the system block diagram is depicted in Fig. 1. The controller at base station gets the channel state information (CSI) feedback from all the K users and then based on these CSI inputs, adopts the proposed adaptive "resource allocation algorithm" to control the subcarrier and power/bit allocation block of OFDMA system base station to assign the N subcarriers to K different users respectively, modulates the user data by dynamic bit loading and distributes power among subcarriers consequently. It is assumed that each subcarrier is assigned to only one user to simplify the design. Subsequently, these modulated signals on all N orthogonal subcarriers are simultaneously inputted to the IFFT module to form an OFDMA symbol, in which a guard interval is added to prevent the ICI. We also assume that the OFDMA symbol undergoes a slowly time-varying frequency selective Rayleigh fading channel, which is constant during one OFDMA symbol but varies from symbol to symbol. Each subband of OFDMA is assumed to be narrow enough that each user signal on it experiences independent fading. By this assumption, we can simplify the channel estimation problem into such a situation that perfect CSI is known to controller at base station via a dedicated feedback channel so as to adjust the resource allocation scheme in a timely manner once the channel changes. At the receiver, the specific resource allocation scheme is also informed through a control channel so that the user can use the subchannel selector to decode the data on its own subcarrier. Assuming that M-level QAM with Gray bit mapping is used, the BER for the k -th (1 < k < K ) user's n -th (1>K).
(5
K.
This is approximated by Nk =LkNJ . HereLok NJ stands for the value tk N rounded to the upper integral. From this, we can obtain the number of those unallocated subcarries F7 = N -E Nk which is left for residual k Nk subcarrier allocation. Other initializations are as follows:
...
Ckn = O,Vke {1,. K}and Vne {1, N}; Rk = O,V ke {1, K}, p = Pmax IN;
q7 {1,2...,K};F = {1,2...,N};
171
> 0, 2) While let the user k with the best priority to choose first its subcarrier n = argmaxneF Ckn =l;Nk = Nk -1; F= F \{n}; Rk = Rk +BIN log2(l +pHk,n);17 = i\{k} 3) While JIFII > rF
IHk,n
r
=
{1, 2,.. K, k
n
=
arg maxn,,
if
=
arg minkc
Rk I Ok;
Hk,n |;
Nk > 0, Ckn =l;Nk =Nk -1; F =F\{n};
Rk Rk
+BINlog2(1+pHk,n;
else 17 =17
\{k}.
4) Residual subcarrier allocation to maximize capacity (This is optional if N>>K). 17 = {1, 2,--- K}1;
forn =1 toF*:
k = argmaxkE 7Hk,n ,Ck,n = 1, Rk = Rk + B I N log2 ( + pHk,n), 77 = 77 \ {k}. In the Step 1, Rk is a variable to keep track of the capacity for each user, F and z7 are the updated sets of all the currently available subcarriers and users respectively during subchannel allocation process. Then in next step, we will utilize maximized bit loading to each user and choose uniform power distribution among all subchannels to simplify designing since we already know from [3] [4] that the resource allocation algorithm with
2466
equal power distribution over all subcarriers can perform almost as well as the optimal power and subchannel allocation. Our algorithm is developed from [6], by introducing a priority scheduling criterion instead of using the random numerical ordering of [6] during the initial subcarrier allocation. From this improvement, we can exploit much more of the multiuser diversity and spectral diversity to increase the capacity greatly, which is proved in our simulation testing. We choose our system under frequency selective fading channel consists of six independent Rayleigh multipaths with an exponentially decaying profile. A maximized delay is 5us and maximum Doppler frequency shift is 30 Hz. The channel information is sampled every 0.5 ms to update the subchannel. The total bandwidth is assumed to be 1 MHz. We choose the average subchannel SNR as 38 dB and M-QAM modulation. The required BER < 10-3 gives an SNR gap of 3.3. The number of users for the system varies from 32 to 64, and the number of subcarrier varies from 32 to 256. The set of rate proportionality constants (expressed as integers) 5k = Ok /min Ok follows the probability mass function (6), in which p stands for probability. I with p =0.5
Pk=2 14
ilvwith p =0.3 with p
=
(6)
0.2
We shall refer to the algorithm in [5] and [6] respectively as ROOT-FINDING and LINER, and compare our scheme termed as "Hui" with them. Shown in Fig.2, Fig.4 and Fig.6 are the simulation results of total system capacity versus user numbers in increments of 2 of our proposed scheme, LINER and ROOTFINDING method. The maximum number of users N and the number of subcarriers K in Fig.2 equal 16 and 64 respectively and we refer to this as case 1. In Fig.4 K=32, N=64 is case 2 and in Fig.6 K=32, N=256 is the case 3. It is demonstrated from these figures that under the same conditions our scheme "Hui" shows a consistently much higher total capacity by introducing a priority criterion compared with the LINER and ROOT-FINDING schemes. The schemes of LINER and ROOT-FINDING all use optimal power allocation while in our scheme only simple equal power distribution is adopted. Still higher capacity is obtained by us. It is observed that in Fig.2 the LINER scheme has an advantage of almost 0.05-0.1 bit/s/Hz than the ROOT-FINDING method for all the numbers of users, while our scheme Hui has even approximate 0.4 bit/s/Hz better gains than the LINER method, a prominent improvement of capacity. Because in case 1 all these schemes use the same simulation parameters, such great improvement in spectral efficiency is brought solely by the priority scheduling introduced during the initial subcarrier allocation which exploits more efficiently the multiuser diversity. The same conclusions can also be obtained from case 2 and case 3. We also calculate the AWGN capacity as a benchmark, given by C =log2 (1 + SNR)=5.2854 bit/s/Hz We can see that our scheme sometimes even surpasses AWGN channel capacity. Thus it is proved from another aspect that our scheme is a spectrum efficient choice.
Simulation results of the normalized rate proportion about each user from case 1 to case 3 are shown in Fig. 3, Fig.5 and Fig.7 respectively. The normalized rate proportion is given
by Rk
K
i=l
Rk
In these figures, Gamma means the strict rate
proportion as specified by pi: jIVi, j E {1, , K}, i . j. Shown from these bar plots, ROOT-FINDING, LINEAR and Hui scheme each obtains a tradeoff between capacity and the relaxed proportionality among users. But what we concern most is the total system capacity and rough rate proportions as just a soft guarantee of fairness among users in practical system will be acceptable. In case 2 of Fig.4, N=2K, which doesn't satisfy condition N>>K, our scheme Hui has a quite higher capacity, even at the same level of AWGN channel capacity. It is known through simulation that it is the priority scheduling and residual subcarriers allocation to users who has best gains that improve the whole system capacity when N is not much larger than K. We notice that in [4][5][6], those mentioned schemes all assume that the number of subcarriers should be much larger than the number of users, but our scheme hasn't such constraint Thus it has a broader application environment. Still in the case 3 shown in Fig.6, when N=8K (N>>K), our scheme Hui achieves a higher capacity than the latest available resource allocation scheme LINEAR. IV. CONCLUSIONS This paper proposes an enhanced rate adaptive resource allocation scheme for the downlink OFDMA system. It introduces the initial priority scheduling in the rate proportion resource allocation method and thus obtains much higher system capacity by exploiting more multiuser diversity and
spectral diversity. Based on the results in the above section, conclusion can be drawn that our enhanced rate adaptive resource allocation scheme is a flexible and efficient air interface suitable for the future high speed wireless system due to its high spectral efficiency. Without solving non-linear equation, it also fits well with the real time applications. Our next research directions include extending such scheme to uplink OFDMA system and using adaptive power allocation among users, which will be forthcoming. REFERENCES
[1] C. Y. Wong, R. S. Cheng, K. B. Lataief, and R. D. Murch, "Multiuser OFDM System with Adaptive Subcarrier, Bit, and Power Allocation," IEEE J. Select. Areas Commun., vol. 17, pp. 1747-1758, Oct 1999. [2] Ergen, M,; Coleri, S.; Varaiya, P.; "Qos Aware Adaptive Resource Allocation Techniques for Fair Scheduling in OFDMA Based Broadband Wireless Access Systems," IEEE Transactions on Broadcasting, vol. 49, Issue 4, pp.362 -370, Dec. 2003. [3] Jiho Jang and Kwang Bok. Lee, "Transmit Power Adaptation for Multiuser OFDM Systems," IEEE J. Select. Areas Commun., vol. 21, pp. 171-178, February 2003. [4] W.Rhee and J. M. Cioffi, "Increase in Capacity of Multiuser OFDM System Using Dynamic Subchannel Allocation," in Proc. IEEE VTC, Tokyo, Japan, May 2000, pp. 1085-1089.
2467
[5] Z. Shen, J. G. Andrews, and B. L. Evans, "Optimal Power Allocation in Multiuser OFDM Systems," in Proc. IEEE Global Communications Conference, San Francisco, CA, Dec.2003, pp. 337-341. [6] Wong, I.C., Zukang Shen, Evans, B.L., Andrews, J.G. " A Low Complexity Algorithm for Proportional Resource Allocation in OFDMA Sstems," Signal Processing Systems,SIPS 2004. IEEE Workshop on 2004 pp.1 - 6. [7] A.J. Goldsmith and S.-G Chua, "Variable-rate variable-power MQAM for fading channels," IEEE Trans. Comm., vol.45, pp1218-1230, Oct. 1997. [8] Teng, Y., Nagaosa, T., Mori, K., Kobayashi, H., "Proposal of Adaptive Subchannel and Bit AllocationMethod for OFDM Access Wireless LAN Systems," VTC, 2003-Spring. 22-25 April, vol.2, pp.910 - 914.
Fig.5. Case 2: K=32, N=64, Normalized rate ratios per user with the required proportions Gamma shown as the leftmost bar.
Fig.2. Case 1: K=16, N=64, total capacity vs. user number in OFDMA system when SNR=38dB and SNR gap 3.3.
Fig.6. Case 3: K=32, N=256, total capacity vs. user number in OFDMA system when SNR=38dB and SNR gap 3.3.
Fig.3. Case 1: K=16, N=64, Normalized rate ratios per user with the required proportions Gamma shown as the leftmost bar.
Fig.7 Case 3: K=32, N=256, Normalized rate ratios per user with the required proportions Gamma shown as the leftmost bar.
Fig.4. Case 2: K=32, N=64, total capacity vs. user number in OFDMA system when SNR=38dB and SNR gap 3.3.
2468