Sei sulla pagina 1di 9

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 27, NO.

2, FEBRUARY 2009

217

Adaptive Subcarrier Allocation Schemes for Wireless OFDMA Systems in WiMAX Networks
Alessandro Biagioni, Romano Fantacci, Fellow, IEEE, Dania Marabissi, Member, IEEE, and Daniele Tarchi, Member, IEEE

AbstractWiMAX is one of the most important technologies for providing a Broadband Wireless Access (BWA) in a metropolitan area. The use of OFDM transmissions has been proposed to reduce the effect of multipath fading in wireless communications. Moreover, multiple access is achieved by resorting to the OFDMA scheme. Adaptive subcarrier allocation techniques have been selected to exploit the multiuser diversity, leading to an improvement of performance by assigning subchannels to the users accordingly with their channel conditions. A method to allocate subcarriers is to assign almost an equal bandwidth to all users (fair allocation). However, it is well known that this method limits the bandwidth efciency of the system. In order to lower this drawback, in this paper, two different adaptive subcarrier allocation algorithms are proposed and analyzed. Their aim is to share the network bandwidth among users on the basis of specic channel conditions without loosing bandwidth efciency and fairness. Performance comparisons with the static and the fair allocation approaches are presented in terms of bit error rate and throughput to highlight the better behavior of the proposed schemes in particular when users have different distances from the BS. Index TermsAdaptive systems, Subcarrier allocation, Orthogonal Frequency Division Multiple Access, WiMAX.

I. I NTRODUCTION

HE INCREASING interest in multimedia applications and high data rate services has lead to a grow in the development of wireless communication systems. Wireless systems have the capacity to address broad geographic areas without the costly infrastructure required to deploy cabled links and satisfy the users connection requirements both in nomadic and mobile application scenarios. In particular, the IEEE 802.16 family of standards [1], [2], supported by the WiMAX commercial consortium, denes the Physical and Medium Access Control (MAC) layers specications for a Broadband Wireless Access (BWA) communication protocol. Among several alternatives, the IEEE 802.16 standard proposes the use of Orthogonal Frequency Division Multiplexing (OFDM) for mitigating frequency-dependent distortion across the channel band and simplifying the equalization in a multipath fading environment [3]. The basic principle of OFDM is parallelization: by dividing the available bandwidth into several smaller bands that are called subcarriers, the transmitted signal over each subcarrier may experience at

Manuscript received 15 January 2008; revised 8 August 2008. This work has been partially supported by Italian National projects INSYEME under grant number RBIP063BPH and WORLD under grant number 2007R989S. The authors are with the Depertment of Electronic and Telecommunications, University of Firenze Via di S. Marta, 3 I-50139 Firenze, Italy. Digital Object Identier 10.1109/JSAC.2009.090212.

fading. Moreover, Orthogonal Frequency Division Multiple Access (OFDMA) is used to provide a exible multiuser access scheme: disjunctive sets of subcarriers and OFDM symbols are allocated to different users. To have more exibility and higher efciency, adaptive OFDM schemes are adopted to maximize the system capacity and maintain the desired system performance [4], [5]: the inherent multi-carrier nature of OFDM allows the use of link adaptation techniques according to the narrow-band channel gains. Transmission techniques which do not adapt the transmission parameters to the fading channel are designed to maintain acceptable performance under the worst-case channel conditions with a consequent insufcient utilization of the available resources. Conversely, if the channel fade level is known at the transmitter, Shannon capacity can be approached by matching transmission parameters to time-varying channel: the signal transmitted to and by a particular station can be modied to take into account the channel gain variation. Usually, wireless systems use power control as the preferred method for link adaptation: the power of the transmitted signal is adjusted in order to maintain the quality of the received signal. Therefore, the transmit power is kept low when a user is close to the BS; conversely, it increases when the user distance from the BS increases. Most of the algorithms proposed in the literature are based on the water lling principle [6], that provides the optimal power allocation through a spectrally shaped radio channel for either maximizing the channel capacity under the constrained transmit power or minimizing the transmit power under providing xed bit rate and error probability (or equivalently minimizing the bit error rate (BER) under the constrained transmit power and bit rate). Although the use of power control can improve the system performance, usually, the total channel capacity is not used efciently at any transmission time if the modulation scheme is xed. To overcome this drawback, adaptive modulation and coding (AMC) schemes have been introduced. In a system where the AMC is adopted, the modulation and coding orders are changed to match the actual channel gain: users close to the BS usually see high channel gains so that they are typically assigned higher-order modulations and higher code-rates but the modulation-order and/or the code-rate usually decrease when the user distance from the BS increases (low channel gains) [7]. Furthermore, in multiuser OFDMA systems, larger throughput can be achieved by assigning the resources (i.e., the sub-

0733-8716/08/$25.00 c 2008 IEEE


Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

218

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 27, NO. 2, FEBRUARY 2009

carriers) in an adaptively manner. Subcarrier allocation strategies dynamically assign subcarriers to users having the best instantaneous channel conditions (i.e., frequency response) by exploiting the multiuser diversity: the instantaneous channel conditions for different users are almost mutually independent; i.e., the most attenuated subcarriers for a given user may result not to be in a deep fade for other users. Subcarrier allocation strategies can follow different criteria; examples are those having a fair data rate distribution among users or maximizing the overall network throughput, depending on the network requirements. Recently, a particular attention has been devoted to the cross-layer optimization of OFDM systems [8][10]. In particular, the authors of [10] proposed an algorithm to provide almost equal data rates for all users. With this strategy more resources are allocated to users with bad channel conditions or far away from the base station. As a consequence, the capacity of users with good channel conditions are not fully exploited. In [11], [12] the problem of a fairness scheduling in resource allocation have been introduced and discussed for the case of multicarrier systems. In addition to this, in [13][16] adaptive subcarriers allocation techniques have been addressed, jointly with other resources allocation strategies, e.g., power allocation and bit assignment schemes. In particular, the adaptive subcarriers and bits assignment scheme presented in [15] has the aim of minimizing the total transmitted power over the entire network while satisfying the data rate requirement of each link. In [16] a low complexity sub-optimal power and subcarrier allocation for OFDMA systems is considered, proposing a heuristic noniterative method as an extension of the ordered subcarrier selection algorithm for a single user case to OFDMA systems. This paper proposes two new allocation strategies based on the consideration that absolute fairness may lead to a low bandwidth efciency and the throughput maximization is sometimes unfair for those users with bad channel conditions. The aim of the proposed strategies is to increase the bandwidth efciency of the system trying to nd a trade-off between efciency and fairness. Differently from previous papers [13] [16] the resource allocation unit is not the subcarrier but a time/frequency unit (named slot), in accordance with WiMAX systems. A generalized version of the dynamic resource allocation scheme [10] to the case of a fair slots assignment among users is also presented in the paper. [10] is often used as a benchmark to validate the performance of subcarrier allocation schemes [17][20]: we have considered the extention of [10] to the slot assignment case as a benchmark to validate the good behavior of the two novel schemes proposed here. Aims of such allocation strategies are to take into account the distribution of channel capacity among the users and improve performances with respect to a non adaptive allocation scheme. The slot allocation is performed based on channel state information (CSI), which allows to fully exploit the characteristics of the multicarrier transmission. The performance has been evaluated by using a xed modulation scheme and an adaptive modulation scheme [21] in order to evaluate the advantages of the joint use of adaptive modulation and subcarrier allocation strategies.

The paper is organized as follows. In Section II, the system model is presented. Section III introduces the different subcarrier allocation strategies. Section IV shows the simulation results and, nally, in Section V some conclusions are drawn. II. S YSTEM M ODEL In this paper, an OFDMA system for mobile wireless MANs, based on IEEE 802.16e standard [2], is considered. In OFDM communication systems, the available spectrum is divided into a large number of subcarriers. Efcient modulation schemes are adopted to transmit data information on these subcarriers by resorting to the use of the inverse discrete Fourier transform (IDFT) in transmission and the discrete Fourier transform (DFT) at the receiving end. In particular, we focus here, on an OFDM-based wireless communication system where K users can communicate with the base station, by using a set of subcarriers, under the form of slot, or subchannel: in particular, accordingly to [2], we suppose that a slot is formed by a group of adjacent subcarriers subdivided in 48 data subcarriers and 6 pilot subcarriers, for a total of 54 subcarriers. This slot structure is the same for both downlink and uplink, and is based on the concept of bin, which is a set of 9 subcarriers (8 data subcarriers and 1 pilot subcarrier): an AMC slot is, therefore, dened as set of bins satisfying the relation N M = 6, where N is the number of consecutive bins and M the number of consecutive OFDM symbols. In this work, the slots are composed by 18 subcarriers contiguous in frequency (N = 2) over 3 consecutive OFDM symbols (M = 3), accordingly to the IEEE 802.16e standard for the AMC permutation scheme. Within a frame, different subchannels, i.e., different sets of slots (groups of contiguous subcarriers in the time-frequency domain) are assigned to users. We have considered a frame duration TF equal to 8 ms and a 1024 FFT for a 10 MHz channel according to the SOFDMA (Scalable OFDMA) concept [22]: the subcarrier spacing is a xed parameter in all the systems with different bandwidth. The assumption is to adopt a TDD structure: this is the most suitable solution for data trafc such as new IP based multirate and multi-QoS services. The TDD frame begins with the Downlink subframe and a guard time of 5 s precedes the Uplink part; then another guard time of 5 s separates a frame from the following one. We have chosen a balanced division of the frame, dividing it into 40 OFDM symbols for the Downlink subframe and 39 OFDM symbols for the Uplink subframe, namely equally parceling the 79 symbols frame dened in the IEEE 802.16e standard as possible size for the OFDMA frame. The proposed adaptive subchannel allocation strategies are based on the estimation of the different channel responses belonging to different users. By assuming the channel as reciprocal in frequency, after estimating the channel response during the uplink subframe of the current TDD frame, and considering it still valid during the downlink subframe of the same TDD frame, the estimation of the complex channel coefcients H(n) (n-th subcarrier) is performed at the BS within the Uplink subframe. It is important to point out that, even if we assume a perfect knowledge of the channel impulse response at the receiving side, we take into account the delay

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

BIAGIONI et al.: ADAPTIVE SUBCARRIER ALLOCATION SCHEMES FOR WIRELESS OFDMA SYSTEMS IN WIMAX NETWORKS

219

The channel capacity of a subcarrier in the OFDMA multiplex can be expressed as: Cp = Bp log2 1 + Es Ts N0 Bp = Bp log2 1 + Es N0 , (1)

Fig. 1.

An example of Slot Allocation considerting three concurrent users

where Bp is the subcarrier bandwidth (corresponding to the total system bandwidth divided by the number of carriers), Es is the mean symbol energy, Ts is the symbol time and N0 is the power spectral density of the AWGN channel. Note that, it has been considered here that in an OFDM system the carrier spacing (Bp ) is equal to the inverse of the symbol time (Ts ). Eq. (1) denes the maximum capacity of a subcarrier under the assumption of a continuous use of it. In our case, we have to take into account that the capacity of a subcarrier within an OFDMA frame is limited by the duration of a single OFDM symbol time: hence, the previous value must be divided by the number of OFDM symbols per frame (Ns ) in order to obtain the capacity belonging to a single tone in a frame. Cp,OFDMA = 1 1 Es Cp = Bp log2 1 + Ns Ns N0 . (2)

introduced by the channel estimation algorithm: the adaptation is performed considering the channel state at the previous frame time and it can introduce a performance loss in time varying channels.

By introducing the effect of the multipath-fading, the capacity of an N M rectangular slot (N contiguous carriers in the frequency domain per M OFDM symbols) becomes: CSLOT (SLOT ) = N M Es 2 Bp log2 1 + , Ns N0 SLOT (3)

A. Adaptive Modulation Scheme The AMC system [21] is modeled here as a Moores state machine, where each state is represented by a modulation scheme: QPSK, 16-QAM and 64-QAM. The adaptation algorithm is basically characterized by three thresholds, representing the transitions instances between different modulation schemes: when a threshold is reached, the modulation order changes and the state machine keeps a different state until another threshold is reached. Under the assumption of comparing the actual estimated mean channel attenuation factor over a slot to the thresholds, it is possible to select the most suitable modulation scheme to be used. The thresholds denition depend on the goal of the adaptation algorithm, i.e., maximizing the throughput or minimizing the error probability.

where SLOT denotes the mean value of the multipath channel coefcient (with respect to the slot) dened as:
M N

SLOT =
i=1 j=1

i,j N M

(4)

where i,j is the multipath channel coefcient of the i-th subcarrier in the j-th OFDM symbol assumed here as Rayleigh distributed random variable. Let us consider a system with K users and an OFDMA frame formed by A slots along the frequency dimension and B slots along the time dimension, resulting in a total of A B slots. Let us dene:

III. A DAPTIVE SUBCHANNEL ALLOCATION In an OFDMA system, resources are considered both in the frequency and time dimensions: different users can be assigned subcarriers on a slot basis belonging to different frequencies or to different OFDM symbol times within the same frame. Hence, the available resources within the OFDMA frame form a grid of slots that are assigned to users according to suitable criteria. Fig. 1 shows an example of allocation within a frame by considering three users (each one represented by a different gray avor). It is possible to highlight that each slot is independent to each other and allocated to one user and one user only. The adaptive subcarrier allocation algorithms proposed in this paper are based on the estimation of the channel capacity belonging to slots. By means of CSI, it is possible to foreseen the channel capacity for a user if that slot is assigned to it.

k: user index, with k = 1, . . . , K; i: slot index along frequency dimension, with i = 0, . . . , A 1; j: slot index along time dimension, with j = 0, . . . , B1; k i,j : averaged multipath coefcient within the slots SLOT with indices (i, j), for the k-th user.

In this paper, an ideal channel estimation in assumed, i.e., the exact value of SNRk is known at the receiving site. The i,j amount of the channel assigned to a given user is related in our case to the overall number of slots assigned to it. In particular, it results to be:
k CSLOTi,j =

N M Bp log2 1 + SNRk i,j Ns for k = 1, 2, . . . , K. (5)

In order to simplify our notation, in (5) and what follows the k dependence of CSLOTi,j on k i,j is omitted. From above, SLOT it is easy to note that the maximum capacity value assigned

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

220

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 27, NO. 2, FEBRUARY 2009

to each user can be derived as:


A1 B1

Cmax,k =
i=0 j=0

k CSLOTi,j for k = 1, 2, . . . , K.

(6)

In practical applications, all users have to share the available capacity in slots usually according to an optimal dynamic allocation algorithm. In order to investigate optimal slot assignment criteria, as in [10], we formulate the slot assignment problem as a non-linear integer programming by introducing for each frame and each user the slot allocation matrix Xk , with A rows, B columns, and elements xk dened as: i,j xk = i,j 1 SLOTi,j assigned to k-th user 0 SLOTi,j not assigned to k-th user (7)

response. When we have a large pathloss or high user mobility this approach becomes unfair for users under poor channel conditions [10]. Under particular conditions, it may happen that some users are unable to access the channel during a frame, hence, suffering for unacceptable long transmission delay that could degrade the performance of network protocols such as the popular TCP/IP (e.g., by increasing the packet dropping probability). Scope of this paper is to lower this drawback by proposing slot allocation algorithms that permits to improve fairness in the slot assignment. As a consequence, the OSDA will be not further discussed in the remaining part of this paper. In particular, we discuss in what follows three solutions, namely: Fair allocation, Proportional allocation and Equal Capacity Increment allocation. A. Fair allocation The rst allocation algorithm considered in this paper represents a generalized version of that proposed in [10]. The main modication introduced here is that of considering the slot allocation instead of single carrier allocation. The problem to be solved in an optimal manner is to allocate slots to users in order to maximize the minimum of all users assigned capacity Ck (i.e., users throughput). This can be obtained by searching for each user the slots allocation matrix Xk that guarantees the minimum of Ck for k = 1, 2, . . . , K, i.e., the highest among all the possible allocations values resulting by different slot allocation solutions. From above it follows that the fair allocation problem can be formulated as follows:
k argmax min CSLOTi,j Xk k

Moreover, as stated before, we have to take into account that slots need to be allocated to one user and to one user only. Hence, the following constraint has to be fullled: xk = 1 i, j i,j
k

(8)

where is the set of all users. According to [10], the channel capacity assigned to the k-th user by a certain allocation algorithm can be dened as:
A1 B1

Ck =
i=0 j=0

k xk CSLOTi,j for k = 1, 2, . . . , K, i,j

(9)

From above, it follows that the goal of any allocation algorithm is to determine the matrix Xk according to suitable optimal criteria. We can start our analysis by extending to the dynamic slot allocation case under consideration the approach proposed in [8], [9] for the case of an optimal subcarrier assignment (frequency dimension only) named here optimal slot dynamic allocation (OSDA). Under the assumption of a xed power allocation and having as a target to maximize the network throughput, the optimal subcarrier allocation problem [8], [9] can be reformulated here as follows:
A1 B1

(15)

subject to xk = 1 i, j i,j
k A1 B1 xk i,j i=0 j=0

(16)

1 k .

(17)

max
Xk k i=0 j=0

k xk CSLOTi,j i,j

(10)

subject to: xk = 1 i, j i,j


k A1 B1

(11)

xk 1 k . i,j
i=0 j=0

(12)

By means of the analytical method proposed in [9] we have that the set of slots assigned to the user k is globally optimal if and only if:
k l CSLOTi,j CSLOTi,j

i, j ,

k, l

(13)

Hence, the optimal slot assignment results to be:


k l CSLOTi,j = argmax CSLOTi,j l

(14)

The OSDA permits to achieve the maximum network throughput by assign a slot to that user having the best channel

As stated in [10], this algorithm requires high computational complexity due to the recursive nature of the solution. In order to lower this drawback we have resorted to search for a suboptimal solution by the following iterative algorithm. Let us dene: s(i, j): the slot of frequency index i and time index j; S: the set of free slots; Sk : the set of slots assigned to the k-th user; Rk : the capacity assigned to the k-th user. The fair allocation algorithm results to be: 1) Initialization: a) Rk 0 for k = 1, . . . , K; b) Sk for k = 1, . . . , K; c) S {s(i, j) : i = 0, . . . , A 1, j = 0, . . . , B 1}; 2) for k = 1, . . . , K: k k a) nd a slot s(i, j) S so that SN Ri,j SN Rn,m for each slot index (n, m) for which at least a free slot in S exists;

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

BIAGIONI et al.: ADAPTIVE SUBCARRIER ALLOCATION SCHEMES FOR WIRELESS OFDMA SYSTEMS IN WIMAX NETWORKS

221

b) Sk Sk {s(i, j)}; c) S S {s(i, j)}; d) Rk Rk + CSLOTk ; i,j 3) while S = : a) nd the user k so that Rk Ru for each user u; k b) nd the slot s(i, j) S so that SN Ri,j k SN Rn,m for each slot index (n, m) for which at least a free slot in S exists; c) Sk Sk {s(i, j)}; d) S S {s(i, j)}; e) Rk Rk + CSLOTk . i,j The algorithm, after an initialization phase, assigns to a given user a slot within the frequencies in which that user has the best channel conditions. Subsequently, all the remaining slots are assigned through an iterative process: for each iteration, the user with the lowest amount of capacity is selected and a slot is allocated to him. The slot to assign is selected among the ones that provide the best SNR to that user. It is important to stress that the slot selection criterion outlined before allows to exploit multiuser diversity in an efcient manner, hence, achieving very good performance. As in [10] we can state that the complexity of the proposed algorithm is very low in comparison with the complexity of the solution of the max-min problem (15)-(17). Moreover, here again the suboptimal algorithm has achieved results very close to the optimal solution [10] so that only suboptimal algorithm results will be considered in carrying out performance comparison in Figs. 2-9. In particular, as stated in Section I, the obtained results will be considered as benchmark in performing performance comparisons with the novel slot assignment approaches proposed here. B. Proportional allocation The main lack of the previous algorithm, is that users with the best channel conditions obtain a lower number of resources with respect to users with worst channel conditions, hence, the channel capacity is not fully exploited. To avoid this drawback, a different allocation strategy is proposed here by assuming that users with the best channel conditions obtain a larger amount of capacity. Being the maximum capacity allocable to each user given by (6), the slots allocation for the algorithm under consideration searches for capacity values for each user proportional to: C1 : C2 : ... : CK = Cmax,1 : Cmax,2 : ... : Cmax,K . (18)

As before, the optimal solution for the max-min problem requires an intensive computation, making a suboptimal solution more attractive. Towards this end, we have resorted to the following algorithm: 1) Initialization: accorda) compute Cmax,k for k = 1, . . . , K ingly to (6); b) Rk 0 for k = 1, . . . , K; c) Sk for k = 1, . . . , K; d) S {s(i, j) : i = 0, ..., A 1, j = 0, ..., B 1}; 2) for k = 1, . . . , K: k k a) nd a slot s(i, j) S so that SN Ri,j SN Rn,m for each slot index (n, m) for which at least a free slot in S exists; b) Sk Sk {s(i, j)}; c) S S {s(i, j)}; d) Rk Rk + CSLOTi,k ; 3) while S = : Rk Ru a) nd the user k so that Cmax,k Cmax,u for each user u; k k b) nd a slot s(i, j) S so that SN Ri,j SN Rn,m for each slot index (n, m) for which at least a free slot in S exists; c) Sk Sk {s(i, j)}; d) S S {s(i, j)}; e) Rk Rk + CSLOTk . i,j The difference between this algorithm and the previous one is the user selection: in this case, the selected user is the one who has the minimum ratio between the actual capacity value and the maximum obtainable capacity value. For the Fair allocation case, the suboptimal algorithm achieves results close to the optimal solution with a lower complexity, hence, in the next gures, only suboptimal algorithm results will be reported. C. Equal Capacity Increment allocation The goal of the strategy proposed in this subsection is to guarantee to each user an equal capacity increment with respect to a non-adaptive strategy. This technique attempts to distribute the additional amount of capacity in an equal manner among all users. In particular, the value of capacity resulting for each of them through a non-adaptive technique is rst derived; next, accordingly to a suitable algorithm a slot allocation is performed in order to have an equal capacity increment with respect to that value. We consider now a non-adaptive strategy in which slots belonging to different users are distinguished only by the position in the time domain (that is a TDMA system); in this case, the different channel conditions an user has on different frequencies are averaged and, therefore, the amount of capacity assigned to the k-th user results to be: Ck = 1 1 Cmax,k = K K
A1 B1

Hence, we can formulate the optimal proportional slot allocation problem as: argmax min
Xk k

Ck Cmax,k

(19)

subject to: xk = 1 i, j i,j


k A1 B1 xk i,j i=0 j=0

(20)

CSLOTk . i,j
i=0 j=0

(22)

1 k

(21)

This value is used as an estimation of the capacity of a nonadaptive algorithm. Let us dene Gk as the difference between

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

222

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 27, NO. 2, FEBRUARY 2009

the capacity assigned to the k-th user and Ck : Gk = Rk Ck . (23) The value Gk represents the capacity increment for the k-th user with respect to a non-adaptive strategy: the allocation scheme under consideration has the target of maximizing the minimum Gk for all users (i.e., k = 1, 2, . . . , K). It follows that the optimal slot assignment problem according to the Equal Capacity Increment criterion can be formulated as: argmax min Gk
Xk k

TABLE I ITU-R VEHICULAR CHANNEL MODEL A, WITH 6 PATHS [23]

Tap 1 2 3 4 5 6

= argmax min(Rk Ck ) .
Xk k

(24)

Channel A Relative delay Average Power (ns) (dB) 0 0.0 310 -1.0 710 -9.0 1090 -10.0 1730 -15.0 2510 -20.0

Doppler Spectrum Classic Classic Classic Classic Classic Classic

subject to xk = 1 i, j i,j
k A1 B1 xk i,j i=0 j=0

(25)

1 k

(26)

f) Gk Rk Ck for k = 0, ..., K 1. The main difference between this algorithm and the fair allocation consists in the user selection: the user for which the value of Gk is minimum is selected, so as to make approximately equal the capacity increment Ck for all the users. As for the previous cases, the computational complexity of the suboptimal algorithm is very low with respect to the optimal alternative with results very close to the optimal solution and, hence, omitted in the next gures. IV. N UMERICAL RESULTS In this Section numerical results obtained through computer simulations will be presented. In particular, the attention will be focused on performance comparisons between the proposed adaptive allocation algorithms and a non-adaptive allocation strategy. The values of the system parameters assumed in carrying out our analysis have been: radio frequency carrier fc = 3.5 GHz; bandwidth of 10 MHz;

As for the previous cases, the optimal solution has an high computational complexity, so that we have resorted here again to a sub-optimal solution according to the following algorithm: 1) Initialization: accordingly a) compute Ck for k = 1, . . . , K to (22); b) Rk 0 for k = 1, . . . , K; c) Gk Ck for k = 1, . . . , K; d) Sk for k = 1, . . . , K; e) S {s(i, j) : i = 0, . . . , A 1, j = 0, . . . , B 1}; 2) while S = : a) nd the user k so that Gk Gu for each user u; k k b) nd a slot s(i, j) S so that SN Ri,j SN Rn,m for each slot index (n, m) for which at least a free slot in S exists; c) Sk Sk {s(i, j)}; d) S S {s(i, j)}; e) Rk Rk + CSLOTk ; i,j

maximum Doppler deviation fd,max 408 Hz (mobility terminals up to 125 km/h); maximum delay spread max = 2.51 s; guard time 1/8 of an OFDMA symbol duration; ITU-R vehicular channel model A, with 6 paths [23] as reported in Tab. I.

The modulation schemes considered in our analysis are QPSK, 16QAM and 64QAM accordingly to the WiMAX specications. Both static and adaptive modulation schemes have been considered in deriving numerical results as explained in Section II-A. In order to take into account the actual modulation and coding used in the IEEE802.16e standard a constraint must be introduced: we assumed here an upper limit to (5), due to the nite set of used modulation schemes, given by: NMn , (27) Rmax = Ns T s where n is the number of bits per symbols corresponding to the used modulation scheme. Let us start our analysis by neglecting the effects of the pathloss on the performance of the considered allocation scheme. This case corresponds to a particular user distribution where each one is at the same distance from the base station. Under this assumption, the differences in channel conditions among different users are caused only by the multipath fading effect. In practical applications this can be achieved with a good approximation by resorting to efcient power control techniques, that estimate the different user attenuations and compensate them by adopting with different powers in transmission. Numerical results are provided here as a function of the mean signal-to-noise ratio at the receiving end under the assumption of a number of active users equal to 20, and considering the use of a maximum throughput adaptive modulation technique in transmission according to Section II-A and [21]. Fig. 2 and 3 show the performance of the proposed schemes in comparison with the static allocation in terms of bit error rate and throughput, respectively. These gures, as expected, point out the better behavior of the proposed methods. These gures also show that when pathloss effects are removed off the proposed methods achieve almost the same performance. Let us consider a more realistic case where the pathloss effects have been considered. The pathloss model is based

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

BIAGIONI et al.: ADAPTIVE SUBCARRIER ALLOCATION SCHEMES FOR WIRELESS OFDMA SYSTEMS IN WIMAX NETWORKS

223

Fig. 2. System Bit Error Rate comparisons for different slot allocation strategies (pathloss effects have been neglected).

Fig. 4. User throughput comparisons for different slot allocation strategies and pathloss channel case.

Fig. 3. System throughput comparisons for different slot allocation strategies (pathloss effects have been neglected).

Fig. 5. User Bit Error Rate comparisons for different slot allocation strategies, and pathloss channel case.

on [24], [25] with pathloss value, in dB, given by: P L = L + 10 log10 D D0 , D > D0 , (28)

where D is the user distance from BS, D0 is the reference distance (set to 100 meters), is the pathloss exponent (set to 4.375 accordingly to an urban environment) and L equal to 20 log10 ( 4D0 ) is the pathloss value at the reference distance (with respect to the wavelength ) that, in this case, is equal to 83.32 dB. Due to the presence of pathloss, the signal-to-noise ratio of the users (having different distance from the base station) is different, due to pathloss and multipath fading. A cell radius of 10 km has been considered and the system must provide an average SNR of 7 dB at the edge of the cell. The average SNR for each user is computed accordingly to the pathloss formula and the SNR value on the cell edge. The number of active users is set to 20 with their positions randomly distributed within the cell area; the performance of a single user have been evaluated with respect to the distance from the base station: the simulations have been repeated with different positions of the reference user (1 km to 10 km from the base station, with a step of 1 km).

Fig 4 shows performance results in terms of throughput of the reference user; maximum throughput adaptive modulation technique discussed in Section II-A has been used. The comparison of the adaptive strategies with a xed (nonadaptive) slot allocation highlights better performance for the proportional and equal increment allocation with respect to the xed one: the user obtains more channel capacity (and thus more throughput) at any distance from the base station. In particular, the proposed proportional and equal increment allocation strategies have the best performance for distances up to 7 km and, anyway, are better than a xed allocation even for higher distances. The fair allocation shows an almost constant throughput for any distance, resulting disadvantageous for the users near the BS (because it has a lower throughput than a xed allocation) and protable for the users far from the BS. In Fig. 5 the performance in terms of bit error rate in case of a pathloss environment has been evaluated for different values of the average SNR at the receiving side. This gure highlights the better performance for the three adaptive schemes with respect to the xed allocation scheme. The fair allocation scheme seems to have the best performance for high SNR values.

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

224

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 27, NO. 2, FEBRUARY 2009

1.8

1.6

1.6

1.4

1.4

1.2

1.2 Throughput (Mbit/s) Throughput (Mbit/s) Fixed allocation Fair allocation Proportional allocation Equal increment allocation 2000 3000 4000 5000 6000 7000 8000 9000 10000 Distance (meters) 1

0.8

0.8

0.6

0.6 0.4

0.4

0.2

0.2

0 1000

0 1000

Fixed allocation Fair allocation Proportional allocation Equal increment allocation 2000 3000 4000 5000 6000 7000 8000 9000 10000 Distance (meters)

Fig. 6. User throughput comparisons for different slot allocation strategies with 64-QAM modulation, and pathloss channel case.
1 Fixed allocation Fair allocation Proportional allocation Equal increment allocation

Fig. 8. User throughput comparisons for different slot allocation strategies, for the case of pathloss channel, and users close to the BS.
2.5

0.1

Bit Error Rate

0.01

Throughput (Mbit/s)

1.5

0.001

1e-04

0.5 Fixed allocation Fair allocation Proportional allocation Equal increment allocation 2000 3000 4000 5000 6000 Distance (meters) 7000 8000 9000 10000

1e-05 10 15 20 25 SNR (dB) 30 35 40

0 1000

Fig. 7. User Bit Error Rate comparisons for different slot allocation strategies with 64-QAM modulation, and pathloss channel case.

Fig. 9. User throughput comparisons for different slot allocation strategies, for the case of pathloss channel, and users far from the BS.

In Fig. 6 and 7 the results for a xed 64-QAM modulation scheme are shown in terms of throughput and bit error rate per user. By comparing these results with that give in Figs. 4 and 5 it is possible to note that he use of adaptive strategies allows to improve performance with respect to a non-adaptive allocation scheme. Finally, two limit cases have been considered by evaluating the performance in terms of throughput by considering that all the users except one are no more at a random distance from the BS. In particular, in deriving the results shown in Figs. 8 and 9, we have assumed that all the user except one are, respectively, at 1 km and 10 km from the BS. These further results prove that the proportional and equal capacity increment are better for users near the BS while the fair allocation scheme is better for users far from the BS. This is mainly due to the fact that the fair allocation tries to allocate the subcarrier in a fair way while in the other cases the subcarrier are allocated in order to maximize the overall cell capacity that is done by damaging the most far users. V. C ONCLUSION This paper has considered a exible OFDMA wireless system where the available resources are accessed in an

adaptive way by different users. In particular, two adaptive slot allocation algorithms have been proposed and compared with alternatives previously proposed in the literature on this subject limited to the case of an optimal subcarriers allocation. These algorithms are based on the estimation of the channel capacity belonging to the slots to be assigned. The rst proposed method, named proportional allocation scheme, assigns different amounts of capacity to users, proportionally to the channel conditions seen by each user; the second strategy, named equal capacity increment approach, provides an equal increase of capacity to all the users with respect to a nonadaptive slot allocation technique. The proposed algorithms permit to achieve a better trade-off between fairness and bandwidth efciency respect to the fair allocation and show a performance increase with respect to a statical allocation independently of channel conditions and at any distance from the base station. R EFERENCES
[1] IEEE Standard for local and metropolitan area networks - Part 16: Air Interface for Fixed Broadband Wireless Access Systems, IEEE Std. 802.16-2004, Oct. 2004.

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

BIAGIONI et al.: ADAPTIVE SUBCARRIER ALLOCATION SCHEMES FOR WIRELESS OFDMA SYSTEMS IN WIMAX NETWORKS

225

[2] Amendment to IEEE Standard for Local and Metropolitan Area Networks - Part 16: Air Interface for Fixed Broadband Wireless Access Systems - Physical and Medium Access Control Layers for Combined Fixed and Mobile Operation in Licensed Bands, IEEE Std. 802.16e2005, Dec. 2005. [3] R. van Nee and R. Prasad, OFDM for Wireless Multimedia Communications. Artech House Publisher, 2000. [4] T. Keller and L. Hanzo, Adaptive multicarrier modulation: A convenient framework for time-frequency processing in wireless communications, Proc. IEEE, vol. 88, no. 5, May 2000. [5] M. Bohge, J. Gross, A. Wolisz, and M. Mayer, Dynamic resource allocation in OFDM systems: An overview of cross-layer optimization principles and techniques, IEEE Network, vol. 21, no. 1, pp. 5359, Jan./Feb. 2007. [6] R. G. Gallager, Information theory and reliable communication. New York, USA: Wiley, 1968, ch. 12. [7] T. Keller and L. Hanzo, Adaptive modulation techniques for duplex OFDM transmission, IEEE Trans. Veh. Technol., vol. 49, no. 5, pp. 18931906, Sep. 2000. [8] G. Song and Y. Li, Cross-layer optimization for OFDM wireless networks-part i: theoretical framework, IEEE Trans. Commun., vol. 4, no. 2, pp. 614624, 2005. [9] , Cross-Layer Optimization for OFDM Wireless NetworksPart II: algorithm development, IEEE Trans. Commun., vol. 4, no. 2, pp. 625 634, 2005. [10] W. Rhee and J. M. Ciof, Increase in capacity of multiuser OFDM system using dynamic subchannel allocation, in IEEE VTC 2000Spring, vol. 2, Tokyo, Japan, May 2000, pp. 10851089. [11] H. Kim and Y. Han, A proportional fair scheduling for multicarrier transmission systems, IEEE Commun. Lett., vol. 9, no. 3, pp. 210212, 2005. [12] M. Ergen, S. Coleri, and P. Varaiya, QoS aware adaptive resource allocation techniques for fair scheduling in OFDMA based broadband wireless access systems, IEEE Trans. Broadcast., vol. 49, no. 4, pp. 362370, 2003. [13] C. Y. Wong, R. S. Cheng, K. B. Letaief, and R. D. Murch, Multiuser OFDM with adaptive subcarrier, bit, and power allocation, IEEE J. Select. Areas Commun., vol. 17, no. 10, pp. 17471758, Oct. 1999. [14] K. Kim, Y. Han, and S.-L. Kim, Joint subcarrier and power allocation in uplink OFDMA systems, IEEE Commun. Lett., vol. 9, no. 6, pp. 526528, 2005. [15] G. Kulkarni, S. Adlakha, and M. Srivastava, Subcarrier allocation and bit loading algorithms for OFDMA-based wireless networks, IEEE Trans. Mobile Comput., vol. 4, no. 6, pp. 652662, 2005. [16] N. Y. Ermolova and B. Makarevitch, Low complexity adaptive power and subcarrier allocation for OFDMA, IEEE Trans. Wireless Commun., vol. 6, no. 2, pp. 433437, 2007. [17] I. Kim, H. L. Lee, B. Kim, and Y. H. Lee, On the use of linear programming for dynamic subchannel and bit allocation in multiuser OFDM, in IEEE Globecom 01, San Antonio, TX, USA, Nov. 2001. [18] Z. Shen, J. G. Andrews, and B. L. Evans, Adaptive resource allocation in multiuser OFDM systems with proportional rate constraints, IEEE Trans. Wireless Commun., vol. 4, no. 6, pp. 27262737, 2005. [19] C. Mohanram and S. Bhashyam, A sub-optimal joint subcarrier and power allocation algorithm for multiuser OFDM, IEEE Commun. Lett., vol. 9, no. 8, pp. 685687, 2005. [20] P. Parag, S. Bhashyam, and R. Aravind, A subcarrier allocation algorithm for OFDMA using buffer and channel state information, in Proc. IEEE VTC-2005 Fall, vol. 1, Sep. 2005, pp. 622625. [21] D. Marabissi, D. Tarchi, F. Genovese, and R. Fantacci, Adaptive modulation in wireless OFDMA systems with nite state modeling, in Proc. IEEE Globecom 07, Washington, DC, USA, Dec. 2007. [22] H. Yaghoobi, Scalable OFDMA physical layer in IEEE 802.16 WirelessMAN, Intel Technology Journal, vol. 8, no. 3, Aug. 2004. [23] Guidelines for evaluation of radio transmission technologies for IMT2000, ITU-R Recommendation M.1225, Feb. 1997. [24] M. Hata, Empirical Formula for Propagation Loss in Land Mobile Radio Services, IEEE Trans. Vehicular Technology, vol. 29, pp. 317 325, 1980. [25] T. Y. Okumura, E. Ohmori, and K. Fukua, Field strength and its variability in UHF and VHF land-mobile radio service, Rev. Elec. Commun. Lab., vol. 16, no. 9, 1968.

Alessandro Biagioni was born in Pistoia, Italy, in 1982. He received the degree in Telecommunications Engineering from the University of Florence in 2007 with a thesis on wireless broadband OFDMA systems. He is currently working in a company operating in the design of telecommunications networks and security and control systems, especially for application in the eld of transportation.

Romano Fantacci (F05) born in Pistoia, Italy, graduated from the Engineering School of the Universit di Firenze, Florence, Italy, with a degree in electronics in 1982. He received his Ph.D. degree in telecommunications in 1987. After joining the Dipartimento di Elettronica e Telecomunicazioni as an assistant professor, he was appointed associate professor in 1991 and full professor in 1999. His current research interests are digital communications, computer communications, queuing theory, satellite communication systems, wireless broadband communication networks, ad-hoc and sensor networks. He has been involved in several European Space Agency (ESA) and INTELSAT advanced research projects. He is the author of numerous articles published in prestigious communication science journals. He guest edited special issues in IEEE Journals and magazines and served as symposium chair of several IEEE conferences, including VTC, ICC and Globecom. Professor Fantacci received the IEE IERE Benefactor premium in 1990 and IEEE COMSOC Award Distinguished Contributions to Satellite Communications in 2002. He is currently serving as Associate Editor for Telecommunication Systems, International J. Commun. Systems, IEEE Trans. Commun. and Area Editor for IEEE Transactions on Wireless Communications.

Dania Marabissi (M00) born in Chianciano, Italy. She received the degree in Telecommunications Engineering and the PhD degree in Informatics and Telecommunications Engineering from the University of Florence in 2000 and 2004, respectively. She joined the Electronic and Telecommunications Department at University of Florence in 2000 where now works as assistant professor. She currently conducts research on physical layer design for broadband wireless systems. In particular her interests include WiMAX and OFDM systems, resource allocation strategies, channel estimation and synchronization. She has been involved in several national and European research projects and is author of technical papers published in international journals and conferences. She is currently serving as Associate Editor for IEEE Transaction on Vehicular Technology.

Daniele Tarchi (S98, M06) was born in Florence, Italy, in 1975. He received the MSc degree in Telecommunications Engineering and the PhD degree in Informatics and Telecommunications Engineering from the University of Florence, Italy, in 2000 and 2004, respectively. He is now an assistant professor at the University of Florence, Italy. His research interests are in both Data Link and Physical Layers, with particular interests to Resource Allocation algorithms in wireless networks, Link Adaptation and Adaptive Modulation and Coding Techniques, MAC Protocols for Broadband wireless access. He has been involved in several national projects (Insyeme, Rescue, Pattern and Women) as well as European projects (Nexway, Newcom, Satnex, COST289). He is currently serving as Associate Editor for IEEE Transactions on Wireless Communications and has been reviewer of several technical papers submitted to journals and magazines.

Authorized licensed use limited to: TOHOKU UNIVERSITY. Downloaded on August 12, 2009 at 01:18 from IEEE Xplore. Restrictions apply.

Potrebbero piacerti anche