A Novel Broadband Wireless OFDMA Scheme for Downlink in
`Cellular Communications
`Junqiang Li, Hojin Kim, Yongsuk Lee and Yongsoo Kim
`i-Networking lab
`Samsung Advanced Institute of Technology, PO BOX 111
`Suwon,, Korea, 440-600
`{jqli, wireless, proakis, ysk}@sait.samsung.co.kr
`
`Abstract-A new air interface scheme based on adaptive
`resource allocations method is proposed for downlink in
`cellular OFDMA systems. The proposed method is featured as
`a novel technique for high spectral efficiency (i.e. frequency
`reuse factor approaching one) and high power efficiency. In
`order to reduce the co-channel interference (CCI) and improve
`the capacity in a cell, a cell is partitioned into three 1200
`sectors and three adjacent sectors from different cells are
`composed as a “virtual cell” which is centrally controlled.
`Assuming knowledge of the instantaneous channel information
`for all sub-channels, bit rates and Quality of Service (QoS)
`requirements of all active users in Base-Stations (BSs), an
`adaptive resource allocations based wireless multi-user OFDM
`access is employed independently in each “virtual cell” by
`using the techniques such as multi-user diversity, interference
`measurement and bad channel avoidance over the whole
`system bandwidth. It
`is demonstrated that the system
`performance is greatly improved compared to the traditional
`systems with
`fixed
`channel
`allocation
`in
`cellular
`communications.
`
`I. INTRODUCTION
`Orthogonal Frequency Division Multiplexing (OFDM)
`is an effective multiplexing technique that can significantly
`mitigate inter-symbol-interference (ISI) induced by wireless
`multi-path fading channels while supporting high data rate
`transmission over wireless radio channels [1]. As this
`technology advances, OFDM is promising to be the key
`modulation
`technique
`in
`the next generation mobile
`communication systems. It is well known that a cellular
`system uses frequency reuse concept to enhance the
`efficiency of
`spectral utilization, which
`introduces
`co-channel interference mainly from adjacent co-channel
`cells that is one of the major sources of performance
`degradation. In contrast to the traditional OFDMA scheme
`based on adaptive allocations of sub-carriers, bits and
`power in the single cell case [5,6,7,11], these access
`schemes based on the water-filling algorithm no longer
`maintain their optimality in the multi-cell environment
`because the water-filling algorithm assigns greater power to
`sub-carriers with high channel gain and lower interference
`level so that the assigned greater power will cause larger
`CCI to users in other co-channel cells. Due to the effect of
`CCI in the cellular OFDMA systems, it is not easy to find
`the optimal solution of sub-carriers, bits and power
`allocations. On other hand, in order to track the dynamic
`nature of the radio channel, a fast and low complexity
`resource allocation algorithm should be required for the
`________________________
`This work is supported by I-networking Lab of Samsung
`Advanced Institute of Technology, Korea
`
`new air interface in the practical systems
`
`In recent years, the interference averaging based
`broadband wireless access schemes have been proposed as
`in [2,3]. In [3], interference averaging is achieved due to
`interference diversity effect by random frequency hopping
`between cells as in BDMA (Band Division Multiple
`Access). Similar interference averaging method using
`multi-carrier CDMA is proposed for downlink access in [2],
`where it was demonstrated that one and a half times bit
`throughputs could be obtained over
`the
`traditional
`OFDM-TDMA scheme. According to [8,9], it was shown
`that interference averaging techniques could perform better
`than
`fixed
`channel
`assignment
`techniques
`(i.e.
`OFDM-TDMA
`and
`fixed OFDM-FDMA), whereas
`interference avoidance techniques outperform interference
`averaging techniques by a factor of 2-3 in spectrum
`efficiency. Thus, the combination of OFDM, sectored
`antennas, adaptive modulation,
`interference avoidance
`based on dynamic sub-channel allocations with
`low
`complexity is promising to significantly mitigate the effect
`of deep fading and CCI while improving the system
`capacity [10].
`In this paper, we consider a new air interface for
`wireless OFDMA systems with time-division duplex (TDD)
`mode in cellular communications systems to improve the
`spectrum and power efficiency, where we assume that the
`transmitting BS knows
`the
`instantaneous
`channel
`characteristic of the mobile users (e.g., such estimation can
`be obtained for the downlink channels using received
`uplink transmissions in the TDD scheme). Since the CCI is
`one of the major sources of performance degradation, our
`object is to mitigate the effect of CCI while achieving high
`spectrum efficiency and power efficiency as well as
`satisfying the required QoS along with multi-rates of all
`active users. In the proposed new air interface, a new
`concept of “virtual cell” which is defined as a cell
`composed of three centrally controlled adjacent sectors
`from different cells is given and the cellular structure based
`on the “virtual cell” is effective to improve the spectrum
`efficiency and the resistance to the CCI compared to the
`conventional structure. As mentioned in [11], it is unlikely
`that a sub-carrier will be in a deep fade or with high CCI
`levels over all the links between BSs and MSs in “virtual
`cell”. This is because of the multi-user diversity where the
`statistics of fading and interference over all these links are
`mutually independent. Thus, with the information of
`channel estimation and rapid interference measurements,
`adaptive resource (i.e. sub-carriers, bits and power)
`allocations can be realized in the “virtual cell”. By doing
`
`0-7803-7700-1/03/$17.00 (C) 2003 IEEE
`
`1907
`
`(cid:57)(cid:58)(cid:42)(cid:82)(cid:36)(cid:3)EX1020
`U.S. Patent No. 10,965,512
`
`

`

`so, high spectrum efficiency by means of hign frequency
`reuse approaching 1 and high power efficiency can be
`achieved by interference avoidance based on dynamic
`sub-carrier allocations, adaptive modulations and power
`control.
`This paper is organized as follows. In Section II, the
`proposed new air interface scheme is described for the
`downlink OFDMA systems including a new cellular
`structure and adaptive OFDMA. Numerical results are then
`presented in Section III. Finally, our conclusions are drawn
`in Section IV.
`
`
`
`
`dPL(
`
`)(
`
`dB
`
`)
`
`
`
`dPL )(
`
`+
`
`X
`
`σ
`
`=
`
`
`
`dPL(
`
`)
`
`0
`
`+
`
`10
`
`n
`
`log
`(cid:168)(cid:168)
`(cid:184)(cid:184)
`0
`Multi-path Rayleigh fading is assumed in a small-scale
`radio environment. With a further assumption that the
`mobile speed is not too high so that the Doppler frequency
`shift and ICI (Inter-Channel Interference) can be ignored,
`the received SINR (Signal-to-Interference-and-Noise Ratio)
`for a particular n-th sub-channel at the k-th mobile in
`“virtual cell” can be expressed as
`PH
`=
`
`
`nk ,
`+
`I
`N
`o
`
`nk,
` is the channel gain resulting from large path
`where
`nkH ,
`knP
`loss (shadow-fading) and small scale Rayleigh fading.
`is the transmitted power to the k-th mobile station at the
`0N denotes the AWGN and
`nkI .
`n-th sub-channel,
`is
`interference on the n-th sub-channel at the k-th mobile
`station which is the summation value of the total CCI from
`all adjacent “virtual cell”. We assume the sum in the
`denominator of (2) can be treated as a Gaussian process.
`2.2 Multi-cell structure
`
`Since the bandwidth is limited and expensive, the
`concept of
`frequency
`reuse
`is adopted
`in cellular
`communication systems to improve the spectrum efficiency.
`In order to further improve the spectrum efficiency and
`increase the resistance to the CCI coming from the
`downlink of the co-channel cells. In Fig. 1, a new multi-cell
`structure is proposed, where three adjacent sectors from
`three adjacent cells BS 1,2 and 3 are composed as a “virtual
`cell”. Within a “virtual cell”, the total bandwidth F is
`adaptively assigned among the active users by the centrally
`
`SINR
`
`
`
`nk,
`
`nk ,
`
`~f
`~f
`~f
`BS8
`BS10
`1,2
`~f
` , (2)
`~f
`~f
`~f
`BS9
`
`1,6
`
`~f
`~f
`
`2,7
`
`~f
`
`3,1
`
`BS1
`
`~f
`1,4
`~f
`
`2,3
`
`~f
`
`3,12
`
`BS12
`
`~f
`
`3,7
`
`BS7
`
`~f
`~f
`
`1,1
`
`2,2
`
`~f
`
`3,3
`
`BS3
`
`~f
`~f
`
`1,7
`
`2,8
`
`~f
`
`3,2
`
`BS2
`
`~f
`1,3
`~f
`
`2,10
`
`~f
`
`3,11
`
`BS11
`
`~f
`~f
`
`1,12
`
`2,11
`
`~f
`
`1,11
`
`3,8
`
`3,10
`
`2,9
`
`1,8
`
`3,9
`
`1,10
`
`~f
`
`1,9
`
`Figure 1: Multi-cell scheme with dynamically frequency band
`allocation in “virtual cells”
`
`2.3 Adaptive OFDMA within “Virtual Cell”
`
`According to multi-user diversity, a sub-carrier with
`deep fading or strong CCI cannot be used by a user,
`however, it may not be in a deep fading or strongly
`interfered over all links between BSs and MSs in “virtual
`cell”, where the statistics of fading and interference over all
`these links are mutually independent. Thus, within the
`“virtual cell”, the frequency channels, bits and power can be
`adaptively allocated according to the noise and interference
`measurement, channel gains for all sub-carriers, the bit rates
`and the required QoS of all the active users. Since it is
`difficult to complete the optimization problem over the
`whole cellular network, we propose a multi-stage
`sub-optimal method with low complexity to realize the
`sub-carrier allocation, bit allocation and power control,
`separately, in one “virtual cell” independently.
`
`1908
`
`controlled BSs in the “virtual cell”, which is based on the
`information of interference measurement and channel
`estimation fed back to BSs. To realize the measurement of
`the CCI of the active users in the “virtual cell”, the
`downlink signals transmitted by all the neighboring “virtual
`cell” are measured by the active users while the desired
`“virtual cell” (i.e. desired three sectors) turns off its
`downlink
`transmitted
`signals. The downlink CCI
`measurement is obtained by monitoring the interference
`within an assigned time slot. In an adjacent “virtual cell”,
`the total frequency bandwidth can be reused and the
`sub-carriers are dynamically allocated within the active
`users. Such processing seems independent among “virtual
`cells”. In practical situation, the traffic is random in each
`sector and the CCI mainly comes from BS4 through BS12
`randomly, the strong co-channel interference can then be
`avoided by using interference avoidance scheme. In the
`macro-cell environment, both sectoring and adaptive
`beamforming techniques can be used in each “virtual cell”
`as in Fig.2 (a) and (b), it is evident that much high
`frequency reuse can be achieved by using adaptive
`beamforming because of refined spatial separation. For
`simplicity, we assume that the antenna gain is equal to 1,
`and sectoring technique is adopted in the following
`sections.
`
`3,5
`
`2,6
`
`3,6
`
`~f
`
`2,5
`
`BS5
`
`~f
`
`2,4
`
`~f
`~f
`
`1,5
`
`2,1
`
`~f
`
`3,4
`
`BS4
`
`~f
`
`2,12
`
`~f
`~f
` . (1)
`~f
`BS6
`
`+
`
`X
`
`σ
`
`(cid:185)(cid:183)
`
`dd
`(cid:169)(cid:167)
`
`II. PROPOSED NEW AIR INTERFACE SCHEME
`In this section, we develop a new air interface for
`
`broadband wireless OFDMA systems. Multi-cell structure
`along with adaptive OFDMA based on dynamic resource
`allocation scheme is investigated. Firstly, we describe the
`channel model in a cellular environment.
`2.1 Channel Model
`In a macro cellular environment, the following path
`loss (PL) notation was used with assumption for the carrier
`frequency around 5GHz. We assume that the average
`nd −
`received power decreases with distance d as
`,
`4=n
`, and the large-scale shadow fading is log-normal
`10=σ dB,
`distribution with a standard deviation of
`which is given by
`=
`
`

`

`K
`
`N
`
`k
`
`=
`1
`
`=
`1
`
`nkp ,
`
`
`
`n
`c
`
`min
`
`Subject to:
`
`k
`
`~f
`
`2,3
`
`~f
`
`1,3
`
`~f
`
`2,1
`
`~f
`
`3,1
`
`~f
`
`3,3
`
`~f
`
`1,1
`
`~f
`
`2,2
`
`~f
`
`1,2
`
`~f
`
`3,2
`
`~f
`
`2,3
`
`~f
`
`1,3
`
`~f
`
`2,1
`
`~f
`
`3,1
`
`~f
`
`3,3
`
`~f
`
`1,1
`
`~f
`
`2,2
`
`~f
`
`1,2
`
`~f
`
`3,2
`
`
`
` A
`
`3
`
`≈
`
`Q
`4
`
`⋅
`
`, (4)
`
`. (3)
`}K
`
`
`{
`k ,...,2,1∈∀
`
`
`
`(cid:166)(cid:166)
`= N
`
`
`
`nk
`,
`
`(cid:166)=
`
`n
`
`R
`1
`P ≤,
`SER
`and
`
`ne
`k
`where each sub-carrier can only be used by one user at one
`nkp ,
`is the transmitted power at the n-th
`time slot.
` (a) (b)
`nkc .
`sub-carrier of the k-th user,
` is the number of the bits
`Figure 2: (a): Sectoring based “virtual cell”; (b) Adaptive
`neP ,
`of the MQAM, and the symbol error rate
` over the
`bemforming based “virtual cell”
`n-th sub-carrier occupied by the k-th user which satisfies the
`QoS requirements. We consider the uncoded transmission
`and assume that an adaptive MQAM constellation is used
`on the sub-channels over the whole frequency bandwidth.
`In particular, based on the Gaussion approximation of the
`sum of
`the AWGN and
`the
`interference,
`in high
`signal-to-noise ratio with large signal constellation size, the
`symbol error probability (or bit error probability, assuming
`the use of gray encoding) of an M-ary QAM transmission
`can be approximated by
`
`(cid:184)(cid:184) (cid:185)(cid:183)
`
`MS
`
`(cid:168)(cid:168) (cid:169)(cid:167)
`
`P
`ne
`,
`
`(cid:185)(cid:183)
`
`4,
`
`(cid:169)(cid:167)
`
`−
`1
`
`Q
`
`MS
`
`=Γ
`
` ,nk
`
`INR
`nk
`,
`−
`1
`nk
`,
`eP ) on the sub-channel
`where the symbol error probability (
`nkM .
`is related to the SINR and modulation level.
` is the
`modulation level at the n-th sub-channel occupied by the
`( )⋅Q denotes the normal probability integral.
`k-th user and
`nk,Γ , satisfying
`We define a service requirement parameter,
`⋅
`INR
`3
`=
`. (5)
`
`neP
`
`,nk
`(cid:168)(cid:168)
`(cid:184)(cid:184)
`−
`1
`
`,nk
`Since the average power of the QAM symbol mainly
`mind
`depends on
`, defined as the minimium distance
`between two points on the signal constellation [12].
`Approximately, we have
`(
`)
`{ }6,5,4,3,2
`nk −
`c
`2 ,
`1
`, ∈nkc
`≈
`d
`cP
`2
`nk
`nk
`nk
`,
`,
`,
`6
`Here, we set the modulation level as between 4QAM to
`64QAM as long as the energy is poured into sub-channel
`with good channel characteristics (i.e., channel gains and
`interference) [6]. We get the minimum distance of the
`constellation at the n-th sub-channel of the k-th mobile as
`(
`)
`Γ
`+
`N
`I
`2
` (7)
`nk
`H
`
`nk ,
`In order to realize the real time allocations of sub-carriers,
`bits and power with low complexity in practical systems,
`we divide the joint optimal allocations into three steps with
`the decision of the number of sub-channels of each users,
`the sub-channel selection of each user, and the bits and
`power allocation of each user based on water filling.
`Step1: Bandwidth allocations (# of sub-carriers) & initial
`bits allocation
`In a wireless environment, due to the effect of
`“near-far” problem, the users with distance far from the BS
`
` where
`
`
`
` (6)
`
`o
`
`
`
`nk ,
`
`2,
`
`d
`
`
`
`nk,
`
`=
`
` downlink OFDM system with K users and N sub-carriers
`is shown in Fig. 3. The users can transmit multimedia
`data such as speech and video with each satisfying data rate
`kR . The bandwidth of each sub-carrier is
`requirements of
`much smaller than the channel coherent bandwidth so that
`the fading on each sub-carrier is flat. The BS is able to
`know the instantaneous channel characteristics of all BSs
`-to- MSs links as long as the channel varies relatively
`slowly. The basic information of channel and interference
`are exchanged among BSs and centrally controlled by BS
`control (BSC). According to channel information, total N
`sub-channels are adaptively allocated by the K active users
`in the “virtual cell”. Each sub-carrier can only be occupied
`by one user and the information on resource allocations is
`sent to mobiles via a dedicated channel so that the mobiles
`are able to extract its own data from the received OFDM
`symbols. Within the “virtual cell”, Macro transmit diversity
`(MTD) can be easily achieved by transmitting the same data
`through the same sub-channels by three sectored antennas
`of three sectors to improve the performance of “weak” user
`who is located at the center of “virtual cell”.
` Our objective is to minimize the total transmitted
`power in a “virtual cell” while satisfying the service
`requirement of each user in each sector. Especially, the
`problem can be formulated as
`
`BS Cont r ol
`
`Sector1
`
`User 1
`(cid:23)
`Weak user
`
`Adapt i ve
`OFDMA & MTD
`
`Sector2
`
`Sector3
`
`MTD
`
`(cid:23)
`
`MTD
`
`(cid:23)
`User K
`
`Adapt i ve
`OFDMA & MTD
`
`Adapt i ve
`OFDMA & MTD
`
`(cid:23)
`
`(cid:23)
`
`(cid:23)
`
`I FFT
`
`I FFT
`
`I FFT
`
`Add
`cycl i c
`pr ef i x
`& P/ S
`Add
`cycl i c
`pr ef i x
`& P/ S
`
`Add
`cycl i c
`pr ef i x
`& P/ S
`
`(cid:23)
`
`(cid:23)
`
`(cid:23)
`
`Channel gai n
`i nt er f er ence
`i nf or mat i on
`&
`modul at i on
`par amet er
`
`BS1
`
`BS2
`
`BS3
`
`MS
`
`User k
`
`Subcar r i er
`sel ect or
`demodul at i on
`
`Channel
`est i mat i on
`i nt er f er ence
`measur ment
`
`(cid:23)
`
`FFT
`
`(cid:23)
`
`(cid:23)
`
`Remove
`cycl i c
`pr ef i x
`& S/ p
`
`
`Figure 3: Adaptive resource allocation OFDMA with Marco
`Transmit Diversity in a “virtual cell”
`
`1909
`
`

`

`nkG . in the ascending order.
`After reordering the parameter
`, l , to each
`ˆ
`We assign the index of sub-carriers of
`nkG ,
`user, where the l-th sub-carrier can be used by only one user
`at a time slot, especially, the sub-carrier allocation process
`is as follows:
` =Φ k
`,1 (cid:22)=
`null
`Initialization:
`where
`
`Φ=Φ
`1=
`k
`m
`(
`
`k
`
`(cid:27)K
`
`K
`
`<Φ
`)
`
`N
`
`, do
`
`
`
`k
`
`,
`
`K
`
`, and
`
`=Φ
`
`null
`
`,
`
`(
`
`<Φ )
`k
`
`m
`
`k
`
`
`
`m
` and
`Φ∈l
`
`
`is expected to achieve a lower overall SINR than other
`users, these users tend to require much higher power to
`transmit the same bits over a sub-channel. If each user is
`given enough sub-channels to satisfy their minimum rate
`requirements, giving the extra sub-channels to users with
`lower average SINR will help
`to reduce
`the
`total
`transmission power [7].
`kα be the shadow-fading gain of the k-th user,
`Let
`kR , and the k-th
`the bit rate requirement of the k-th user
`km sub-carrier. According to [6], since
`user be allocated
`each user selects the best sub-channels for bits allocation, it
`is argued that flat transmit PSD (power spectral density)
`could hardly reduce the data throughput of a multiuser
`OFDM system, so the initial bit allocations after the
`decision of the number of sub-carrier of the k-th user is
`given as
`
`
`
` best ssubcarrier
`
`
`
`
`
` (8)
`
`m
`
`k
`
`(cid:188)(cid:187)
`
`×(cid:187)
`
`mR
`(cid:171)(cid:172)(cid:171)
`
`kk
`
`−
`
`k
`
`1
`
`
`
`
`
`
`
` the
`
`R
`
`(cid:188)(cid:187)
`
`+(cid:187)
`
`kk
`
`(cid:176)(cid:176)
`
`c
`
`
`
` ,nk
`
`=
`
`×
`
`(cid:184)(cid:184)(cid:185)(cid:183)
`
`m
`
`k
`
`(cid:188)(cid:186)
`
`⋅(cid:187)
`
`mR
`(cid:171)(cid:172)(cid:170)
`
`kk
`
`m
`
`k
`
`−
`
`R
`
`k
`
`+
`
`(cid:168)(cid:168)(cid:169)(cid:167)
`
`×
`
`P
`
`1
`
`+
`
`(cid:184)(cid:184)(cid:185)(cid:183)
`
`m
`
`k
`
`(cid:188)(cid:187)
`
`mR
`(cid:171)(cid:172)(cid:171)
`
`kk
`
`k
`
`(cid:168)(cid:168)(cid:169)(cid:167)
`
`P
`k
`
`total
`
`=
`
`mR
`(cid:171)(cid:172)(cid:171)
`(cid:176) (cid:174)(cid:173)
`
`
`
`subchannelOther
`
`(cid:187)(cid:188)(cid:187)
`
`mR
`(cid:171)(cid:172)(cid:171)
`
`kk
`
`(cid:176)(cid:176)(cid:176) (cid:175)
`
`Thus, the average total transmit power of the k-th user can
`be presented as
`−
`⋅(cid:187)
`R
`
`While (cid:166)
`=
`k
`1
`For k=1,...,K
`Φ∉l
` if
`l Φ∈
`
` and
`k
` End if
` End for
`End while
`( )⋅m denotes the number of elements in coset.
`where
`Step3: Refined bit allocation
` After the sub-carrier allocation, we initialize the bit
`allocations as in the Eqn. (8). we then refine the bit
`allocations by water-filling algorithm. Mathematically, we
` (9)
`have the optimization problem for the k-th user as
` (12)
`lkP ,
`min
`
`(cid:166)Φ
`
`(cid:184)(cid:184)(cid:185)(cid:183)
`(cid:187)(cid:188)(cid:187)
`
`mR
`(cid:171)(cid:172)(cid:171)
`(cid:168)(cid:168)(cid:169)(cid:167)
`
`kk
`
`K
`
`, such
`
`α
`k
`k
`
`km ,
`
`P
`,
`
`,1 (cid:22)=
`
`(cid:184)(cid:184)(cid:185)(cid:183)
`
`(cid:188)(cid:187)
`
`+(cid:187)
`
`mR
`(cid:171)(cid:172)(cid:171)
`(cid:168)(cid:168)(cid:169)(cid:167)
`
`kk
`
`,
`
`α
`k
`Our objective is to find a set of
`that
`
`(cid:166)=K
`
`min
`
`
`
`(cid:166)Φ
`
`k
`
`
`
` ,lk
`
`>
`
`,
`
`(cid:191)(cid:190)(cid:189)
`
`,...,
`
`N
`
`(cid:187)(cid:187)(cid:186)
`
`k
`
`max
`
`(cid:171)(cid:171)(cid:170)
`(cid:175)(cid:174)(cid:173)
`
`∈
`
`m
`
`k
`
`kP
`total
`,
`k
`1
`subject to
`K
`
`,
`k =(cid:166)
`R
`m
`N
`R
`=1
`k
`maxR
`is the maximum number of modulated bits
`where
`max =
`R
`6
`over sub-carriers (e.g.
`)
`.
` The optimal
`distribution of sub-carriers among users can be found as
`follows,
`Initialization:
`
`∈ k
`l
`subject to
`P ≤,
`SER
` and
`.
`=
`R
`c
`le
`k
`k
`lk
`,
` (10)
`∈
`l
`Since the initialized bit allocation approaches the optimal
`bit allocation, we can use the greedy method proposed in [5]
`to refine bit allocation. By finding two sub-carriers, indexed
`by l and l′ , such that
`−
`−
`−+
` (13)
` (cP
`
` (cP
`
` (cP
` (cP
`
`
`)
`)1
`)1
`)
`′
`′
`′
`′
`
` ,lk
` ,lk
`
` ,lk
`
`
` ,lk
`
` ,lk
`
` ,lk
`
` ,lk
`
`
`if such as ( )ll ′,
` pair exits, moving one bit from
`sub-channel l to sub-carrier l′ will lower the overall
`transmit power of the k-th user. The optimal bit allocations
`for the k-th user can be obtained by satisfying
`Δ
`+ Δ≥
`−
`P
`P
`, (14)
`k
`k
`(
`)
`(
`
`{
` )}lk
`<
`While (cid:166)
`Δ
`−+
`k N
`+
`P
`cP
`cP
`
`min
`1
`k
`
`lk ,
`
`lk ,
`,
`Φ∈∀
`l
`=
`1
`)
`)}1
`{
`
`(
`(
`
`=(cid:8)
`Δ
`−
`−
`P
`cP
`cP
`P
`max
`k
`lk
`lk
`lk
`lk
`,
`,
`,
`,
`k
`Φ∈∀
`l
`,1 (cid:22)=
`K
`,
`Since the minimum distance in the signal constellation of
`,
`the transmitted symbol is given in Eqn. (7), both BS and
`(cid:8)
`=
`l
`P
`arg
`min
`nkd , directly in order to
`k
`MSs can calculate the value of
`(cid:37)≤
`Kk
`1
` = ll mm
`1+
`
`reduce the side information transmitted to MSs by BSs.
`
`End while
`Step2: Sub-carriers allocation
`III. NUMERICAL RESULTS
`We define the channel characteristic parameter
`In this section, we provide some simulation results,
`nkG . at the n-th sub-carrier of k-th user as
`which demonstrate the potential of our proposed adaptive
`+
` (11)
`resource allocation based OFDMA systems in the downlink
`N
`I
`=
`0
`G
`for cellular communications. In the following simulations,
`H
`channel model “Vehicular A” in Table.1 is adopted with a
`
`=
`
`=
`
`k
`
`k
`
`
`
`lk ,
`
`−
`
`
`
` ,
`
`k
`
`,1 (cid:22)=
`
`,
`
`K
`
`,
`
`where
`
`)
`1 −+
`
`P
`k
`
`,
`
`total
`
`(
`m
`
`)k
`
`and
`
`,
`
`(cid:187)(cid:187)(cid:186)
`
`R
`k
`maxR
`
`(cid:171)(cid:171)(cid:170)
`
`=
`
`m
`
`k
`
`K
`
`k
`
`m
`
`P
`k
`
`,
`
`k
`
` do
`(
`m
`
`k
`
`total
`
`
`
`
`
`
`
`nk,
`
`
`
`nk ,
`
`
`
`nk,
`
`1910
`
`

`

`Delay Profile (nsec)
`(0,250,500,750,1000,1700,
`1900,2400,2600,2700)
`
`is
`maximum Doppler frequency of 550 Hz, which
`=df
`4.0
`720
` μs.
`corresponding to the coherent time
`The available bandwidth F is assumed to be 18 MHz at
`5GHz radio frequency and the total number of sub-carriers
`N =1024. This corresponds to a sub-channel separation of
`16 kHz and an effective OFDM frame duration of 56.8889
`μs, which means that the channel is static for 12 OFDM
`frame duration. In each OFDM frame, a cyclic prefix of
`5.68889 μs duration is added which is larger than the
`maximum time delay of multi-path Rayleigh fading channel
`
`
`Type
`
`Vehicula
`r A
`Table1: Channel model for simulations
`
`Without loss of generality, we consider a “virtual cell”
`with (6, 3, 2) users in each sector and the bit rates of all
`users are (4Mbps, 4Mbps, 4Mbps, 4Mbps, 8Mbps, 8Mbps;
`4Mbps, 4Mbps, 8Mbps; 4Mbps, 8Mbps), respectively. We
`0 =
`d
`R
`5.0
`consider the reference distance
`, and average
`propagation distance of CCI from the adjacent co channel
`5.2=
`d
`R
`. The standard deviation of
`“virtual cell”,
`10=σ dB over
`large-scale shadow-fading distribution is
`all sub-carriers. We adopt a set of shadowing gain of the
`active users in “virtual cell” with values (0.70, 1.13, 1.04,
`1.33, 0.84, 0.72; 0.91, 0.89, 0.54; 0.77, 1.22). The value of
`nk,Γ , of all users over all sub-carriers is set to be
`gamma,
`the same for convenience. We consider Fig. 4, where we
`present the performance comparison between our proposed
`adaptive resource allocations based OFDMA scheme and a
`traditional fixed OFDM-FDMA scheme. It is demonstrated
`that the proposed scheme can greatly improve the system
`performance
`compared
`to
`the
`traditional
`fixed
`OFDM-FDMA scheme with the same total transmit power
`and the same bit throughout.
`
`
`Power (dB)
`(0,-3,-6,-9,-1
`2,-13,-15,-25
`,-21,-25)
`
`IV.
`
`CONCLUSIONS
`
`
`
`A new air interface scheme based on adaptive
`resource allocations method was proposed for downlink in
`the multi-cell OFDMA systems. Based on the statistic
`multiplexing and bad quality channel avoidance, our
`proposed method can achieve high spectral efficiency (i.e.,
`frequency reuse factor approaching 1) and high power
`efficiency. Assuming
`perfect
`knowledge
`of
`the
`instantaneous channel information at all sub-channels, bit
`rates and QoS requirements of all active users in the
`“virtual cell”, adaptive resource allocations based wireless
`multi-user OFDM access is employed independently in
`each “virtual cell” benefiting from multi-user diversity.
`From the numerical results, it is demonstrated that the
`system performance is greatly improved compared to the
`traditional
`systems with
`fixed
`channel
`allocation.
`
`1911
`
`
`
`
`Figure 4: Performance comparison between proposed method and
`fixed OFDM-FDMA scheme
`
`
`
`References
`1.
`J. A. C. Bingham, “Multicarrier modulation for data
`transmission: an idea whose time has come,” IEEE Trans. on
`Commun., Vol. 28, No. 5, pp. 5-14, May 1990
`2. H. Atarashi, S. Abeta, and M. Sawahashi, “Broadband packet
`wireless access appropriate for high-speed and high-capacity
`throughput,”IEEE VTC’2001, pp. 566 - 570, Rhodes, Greece,
`May 2001
`3. M. Suzuki, R.Bohnke, and K. Sakoda, “Band division
`multiple access (BDMA) system: A novel approach for next
`generation mobile
`telecommuniation system, based on
`OFDM and SFH-TDMA,” IEEE VTC’1998.
`4. H-J Su and E. Geraniotis, ”A distributed power allocation
`algorithm with adaptive modulation for multi-cell OFDM
`systems,” IEEE SSTA, pp. 474 – 478, South Africa Sept.
`1998
`5. S. K. Lai, R. S. Cheng, K. B. Letaief and C. Y. Tsui,
`“Adaptive tracking of optimal bit and power allocation for
`OFDM systems in time-varying channels,” IEEE WCNC’99,
`pp. 776 – 780, New Orleans, LA, USA Sept. 1999
`6. W. Rhee and J. M. Cioffi, “Increase in capacity of Multiuser
`OFDM system using dynamic subchannel allocation,” IEEE
`VTC’2000, pp. 1085 - 1089 Tokyo, Japan, May 2000
`7. D. Kivanc and H. Liu, “Subcarrier allocation and power
`control
`for OFDMA,”
`the Thirty-Fourth Asilomar
`Conference on Signals, Systems and Computers, pp. 147 -
`151 Nov. 2000
`in personal
`8. G.
`J. Pottie, “System design choices
`communications” IEEE pers. Commun., Vol. 2,no. 5, pp.
`50-67, Oct. 1995.
`Justin C. -I Chuang and N. R. sollenberger, “Beyond 3G:
`wideband wireless data access based on OFDM and dynamic
`packet assignment,” IEEE communications Magazine, PP.
`78-87, July 2000.
`10. L. J. Cimini, Justin C. –I Chuang, and N. R. Sollenberger,
`“Advanced cellular
`internet
`service
`(ACIS),”
`IEEE
`Communications Magazine, pp. 150-159, Oct.1998.
`11. C.Y. Wong, R. S. Cheng, K. B. Letaief and R. D. Murch,
`“Multuser OFDM with adaptive subcarrier, bit, and power
`allocation,” IEEE JSAC, Vol. 17, No. 10, Oct 1999.
`12. Inhyoung Kim, Hae Leem Lee, Beomsup Kim and Lee Y.H,”
`On the use of linear programming for dynamic subchannel
`and bit allocation in multiuser OFDM,” IEEE GLOBECOM
`'01, pp. 3648 –3652, November 2001
`
`9.
`
`