A combined MAC and Physical resource allocation

mechanism in IEEE 802.16e networks

Sondes Khemiri and Guy Pujolle
LIP6 University Paris 6
104 Avenue du Pr esident Kennedy
75016 Paris, France
Email: [email protected]
Khaled Boussetta and Nadjib Achir
L2TI Institut Galil ee University Paris 13
99 Avenue J-B Cl ement
93430 Villetaneuse, France
Email: [email protected]
AbstractThis paper addresses the issue of wireless resource
allocation of a Mobile WiMAX cell based on OFDMA access
technique. We compare two strategies: the MAX SNR which is
the most used channel assignment technique and our proposal
which is based on channel condition learning. As we distinguish
four IEEE 802.16e-2005 service classes: UGS, rtPS, nrtPS and
ErtPS, we combine these strategies with a scheduler block in
order to satisfy the QoS constraints of these classes. In addition,
we take into consideration the Adaptive Modulation and Coding
(AMC) scheme, through the selection of the slot which is based
on the modulation scheme. The main concern here is to propose
a simple method to assign WiMAX physical resources called slot
to satisfy QoS constraints of a users scheduled MPDUs, which
can be easily and efciently used in a realistic Mobile WiMAX
network.
I. INTRODUCTION
This work focus on IEEE 802.16e-2005 WiMAX network
with Wireless MAN OFDMA physical layer. This technology,
also referred as a mobile WiMAX, aims to offer a quadruple-
play service to multiple mobile subscribers (MSS), which
can have access anytime and anywhere to various application
types, like le downloading, video streaming, emails and VoIP
with and without silence suppression. In order to guarantee the
quality of service required by these applications, the standard
denes ve service classes. Namely: UGS for VoIP without
silence suppression, rtPS for video streaming, nrtPS for le
downloading, ErtPS for voice with silence suppression and
nally BE for web and mailing applications. For notation
simplicity, we will refer to UGS, rtPS, nrtPS, ErtPS and BE as
a class 1, 2, 3, 4 and 5 respectively. Let C = {1, 2, 3, 4, 5}. In
our work we will only study class 1,2,3 and 4 as BE service
does not requires any QoS guarantees.
II. WIMAX PHYSICAL LAYER OVERVIEW
The mobile WiMAX physical layer is based on Orthogonal
Frequency Multiple Access (OFDMA), which is a multi-
users extension of Orthogonal Frequency-Division Multiplex-
ing (OFDM) technique. The latter principles consists of a
simultaneous transmission of a bit stream over orthogonal
frequencies, also called OFDM sub-carriers. Precisely, the total
bandwidth is divided into a number of orthogonal sub-carriers.
There are three types of sub-carriers: 1) Data sub-carriers are
used to transport in parallel low rate data streams, 2) Pilot sub-
carriers are used for synchronization and channel estimation
and 3) null sub-carriers enable the signal to naturally decay
and prevent leakage of energy. The set of parallel low rate
data streams are resulting from the division of the high rate
data stream using a Fast Fourier Transform (FFT) function.
Depending on the sub-carrier signal quality (e.g. SNR level)
an adapted modulation scheme has to be employed to transmit
the corresponding lower data stream. In this work we need to
know two aspects that specify mobile WiMAX systems: sub-
channelization and adaptive modulation and coding scheme.
A. sub-channelization and permutation scheme
In mobile WiMAX [1], the OFDMA sharing capabilities are
augmented in multi-users context thanks to the exible ability
of the standard to divide the frequency/time resources between
users. Precisely, the basic unit of allocation in the time-
frequency grid is named a slot. Broadly speaking, a slot is an n
x m rectangle, where n is a number of sub-carriers and m is a
number of contiguous symbols. The standard supports multiple
subchannelization schemes (PUSC, AMC, FUSC, TUSC, etc.),
which dene how an OFDMA slot is mapped over subcarriers.
First, the sub-channelization could be: adjacent when sub-
carriers in each sub-channel are grouped in the same frequency
range to enable beam-forming or distributed when they are
pseudo-randomly distributed across the frequency spectrum
for frequency diversity. Although a slot will always contain
48 data carriers, their arrangement is different, depending on
the subchannelization scheme. Precisely, each slot consists
of a single sub-channel (n = 1) by one (m = 1), two
(m = 2), three (m = 3) or six (m = 6) OFDMA symbols.
For instance: For downlink Full Usage Sub-Carriers (FUSC)
scheme each slot is 48 subcarriers by one OFDM symbol,
for downlink Partial Usage of Sub-Carrier (PUSC) scheme
each slot is 24 subcarriers by two OFDM symbols, for uplink
PUSC and TUSC (Tile Usage of Sub-Carrier) scheme each
slot is 16 subcarriers by three OFDM symbols and for Band
Adaptive Modulation and Coding (BAMC) scheme each slot
is 8, 16, or 24 subcarriers by 6, 3, or 2 OFDM symbols,
respectively. These permutations are respectively called (1X6,
2X3, or 3X2).
978-1-4244-2519-8/10/$26.00 2010 IEEE
B. Adaptive modulation and coding scheme (AMC)
In order to adapt the transmission to the time varying
channel conditions that depends on the radio link character-
istics WiMAX presents the advantage of supporting the link
adaptation called Adaptive Modulation and Coding scheme
(AMC). It is an adaptive modication of the combination
of modulation, channel coding types and coding rate also
known as burst prole that takes place in the physical link
depending on a new radio condition. the following table I
shows examples of burst proles in mobile WiMAX there are
52 in IEEE802.16e-2005 [1][2]:
Prole Modulation L Coding scheme Rate
3 16 QAM (RS +CC/CC)
1
2
5 64 QAM (RS +CC/CC)
2
3
6 64 QAM (RS +CC/CC)
3
4
TABLE I
BURST PROFILES: (RS) REED SOLOMON, (CC) CONVOLUTIONAL CODE
In our work we will only consider the downlink. The
downlink SNR is provided by the mobile to the base station
-Using the channel quality feedback indicator- with feedback
on the downlink channel quality, the BS takes into account the
time varying channel conditions by using this (SNR) value and
then decides which burst prole must be used for the SS. When
the distance between the base station and the subscriber station
increases the signal to the noise ratio decreases due to the path
loss. This is why modulation must be used depending on the
station position starting from the lower efciency modulation
(near the BS) to the higher efciency modulation (far away
the BS).
The WiMAX PHY layer has the responsibility of slot
allocation and framing over the radio interface. The minimum
time-frequency resource that can be allocated by a WiMAX
system to a given user is called a slot. In this paper we focus
on this slot allocation problem in mobile WiMAX OFDMA
system which consists of how to assign PHY resource to a
user in order to satisfy a QoS request.
The rest of this paper is organized as follows: the state of
the art is presented in section III. The problem formalization is
detailed in section IV. OFDMA channel assignment strategies
are presented in section V. Their performances evaluation are
discussed in section VI. Finally, section VII summarize the
paper.
III. STATE OF THE ART
In [3], authors propose an algorithm that employs a combi-
nation of dynamic sub-channel selection, modulation selection
and power allocation for packet transmission in OFDMA
System. But this proposal did not take into account QoS
constraints of packets. In [4], Orthogonal Frequency Division
Multiple Access (OFDMA) provides possibilities to exploit
diversity of a frequency selective fading channel by appro-
priate resource scheduling strategies. This paper presents the
performance comparison of different strategies with fair allo-
cation of transmission capacity in a typical multiuser scenario.
The analyzed performance metrics comprise both throughput
and delay, which are required for a comprehensive analysis
regarding Quality of Service (QoS). The evaluation has been
conducted by means of stochastic event-driven simulations of
a WiMAX based OFDMA system. Therefore, the developed
simulation tool provides a detailed implementation of both
medium access control (MAC) and physical layer (PHY). The
presented results clarify the signicant impact of different
scheduling strategies on the tradeoff between throughput and
delay. In [5] and [6] Channel segregation performance has
been examined by applying it to FDMA systems. This paper
discusses its application to the TDWFDMA or multi-carrier
TDMA system. Spectrum efciency of the TDMA/FDMA
cellular system deteriorates due to the problem of inaccessible
channel: a call can be blocked in a cell even when there
are idle channels because of the restriction on simultaneous
use of different carrier frequencies in the cell. This paper
shows that channel segregation can resolve this problem with
a small modication of its algorithm. The performance of
the TDWFDMA system with channel segregation on the call
blocking probability versus trafc density is analyzed with
computer simulation experiments. The effect of losing the
TDMA frame synchronization between cells on the perfor-
mance is also discussed. in [7] Orthogonal Frequency Division
Multiple Access (OFDMA) base stations allow multiple users
to transmit simultaneously on different subcarriers during
the same symbol period. This paper considers base station
allocation of subcarriers and power to each user to maximize
the sum of user data rates, subject to constraints on total power,
bit error rate, and proportionality among user data rates.
These works did not consider the double problem of MP-
DUs selection for transmission and the channel assignment
technique, which it is a purpose of this work.
IV. SYSTEM DESCRIPTION AND PROBLEM STATEMENT
We consider a network relying on IEEE 802.16e 2005
technology with Wireless MAN OFDMA physical layer. The
system provides a quadruple-play service to multiple mo-
bile subscribers (MSS), which can have access anytime and
anywhere to various application types, like le downloading,
video streaming, emails and VoIP. In order to guarantee the
quality of service required by these applications, the service
provider has to distinguish four service classes. Namely: UGS
for VoIP, rtPS for video streaming, nrtPS for le downloading
and ErtPS for voice without silence suppression.
We only study the system in the downlink, in order to satisfy
the QoS constraints of users service ows we assume that there
is in the base station a classier that assigns users ows to the
service class described before (UGS, rtPS, ErtPS or nrtPS), we
also suppose that there is a call admission control mechanism
to make sure that the accepted calls did not degrade the QoS of
the ongoing calls, and there enough available system capacity
for the accepted call and if not the call is rejected.Since we
consider a mobile radio environment this system capacity vary
with channel condition so a scheduling mechanism must be
used in order to select which MPDUs must be transmitted in
addition to a physical resource assignment strategy in order to
select the best slot (physical resource) that satises the QoS
constraints of the selected MPDUs. So a good knowledge of
a system physical layer characteristics and the radio mobile
environment is needed in order to obtain this aim, this is the
purpose of the rest of this section.
In order to adapt the transmission rate according to the
channel quality we consider that the system supports an
Adaptive Modulation and Coding (AMC) scheme. Precisely,
we assume that the cell is divided into 3 coverage zones where
a same modulation and coding scheme is used for all calls
ongoing in a specic area. Without loss of generality, we
assume that Z = {1, 2, 3} and that the areas border shape
are concentric perfect circles.
In our model we consider a typical downlink WiMAX
OFDMA system. we consider the following parameters that
denes a typical IEEE 802.16e-2005 WiMAX Prole II:
Parameters Meaning
K Number of users
S Number of subcarriers
g
k,s
Envelope of the channel gain for user k in subcarrier s
p
k,s
Transmit power allocated for user k for subcarrier s
No The noise power spectral density

2
AWGN power spectrum density
2
=
BN
0
S
P Total power transmission
B Total transmission bandwidth
Permutation BAMC (1X6) adjacent permutation scheme
TABLE II
WIMAX OFDMA SYSTEM PARAMETERS
At the base station transmitter it is better to understand how
this access mechanism works. In fact, bits for each of the
different K users are allocated to the S subcarriers, and each
subcarrier s (1 s S) of user k (1 k K) is assigned
a power p
k,s
. Each of the users bits are then modulated into
S M-QAM symbols (M = 16, 64), which are subsequently
combined using the IFFT into an OFDMA symbol. This is
then transmitted through a slowly time-varying, frequency-
selective Rayleigh channel with a bandwidth B. In this work
we consider that we allocate a xed power p
k,s
=
P
S
for
each subcarrier since we did not focus on a power allocation
problem. We assume that each user experiences an indepen-
dent fading and the channel gain of user k in subcarrier s is
denoted as g
k,s
so we can easily deduce the k
th
users received
signal-to-noise ratio (SNR) on subcarrier s as follows :
SNR
k,s
= p
k,s
g
2
k,s

2
(1)
where
2
=
BN0
S
. The slowly time-varying assumption is
crucial since it is also assumed that each user is able to
estimate the channel perfectly and these estimates are made
known to the transmitter via a dedicated feedback channel.
These channel estimates are then used as input to the resource
allocation algorithms. We suppose that the channel condition
did not change during the frame duration.
As we consider a mobile WiMAX system supporting Adap-
tive modulation and coding we can deduce from [7] and [8]
the number r
k,s
of bit that a given subcarrier s can transmit
if we know channel condition for a given user k, so we have:
r
k,s
= log
2
_
1 +
SNR
k,s

_
(2)
where = ln
_
5BER
1.6
_
, the BER is a bit error constraints to
be met described in [7]. We suppose that the system supports
a frame format with a length L as we have a mobile WiMAX
OFDMA system with a nominal bandwidth B the total frame
length corresponds to L =
1
B
T where
1
B
is the OFDM
symbol length and T is the total number of symbols that
compose the frame. As we described previously we suppose
to have one permutation scheme precisely a BAMC (1X6)
scheme where a slot corresponds to 8 data subcarriers by 6
OFDM symbols. Lets have an allocation matrix of a k
th
user
denoted by A
k
, this matrix is expressed as following:
A
k
=
_
a
k
t,s

(t,s){0,T}{0,S}
(3)
where
a
k
t,s
= 1
{1
(t,s)
=k}
(4)
Using 1,2,3 and 4 The throughput per user per frame is
obtained as follows:
th
k
=
B
S

S

s=0
T

t=0
(a
k
t,s
log
2
_
1 +
p
k,s
.g
2
k,s
.
2
_
) (5)
We can deduce that in order to satisfy a user k request
corresponding to a rate equal to th
k
we must allocate a
total number of subcarriers per a single frame equal to:
n
k
=

S
s=0

T
t=0
a
k
t,s
, note that this number will be a
multiple of 48, since a slot corresponds to 8 adjacent data
subcarriers X 6 OFDM symbols. The total capacity C per
frame corresponds to:
C =
K

k=0
_
B
S

S

s=0
T

t=0
(a
k
t,s
log
2
_
1 +
p
k,s
.g
2
k,s
.
2
_
)
_
(6)
Our aim is to nd an efcient resource assignment strategy
that takes into account two aspects: 1)the varying channel con-
dition and 2) the QoS constraints of users MPDUs scheduled
to be transmitted into the physical frame.
V. PROPOSAL
In this work our purpose is to respond to two questions:
1) Which MPDUs to serve? and
2) Which slot to assign to satisfy the bandwidth request of
the selected MPDUs?
To ask to the rst question we propose the following Hybrid
scheduling block: Here the idea is to use two Round Robin
(RR) schedulers in a rst stair to provide fair distribution of
bandwidth especially between ErTPS and UGS classes since
they have the same packet sizes. In the second stair we propose
Fig. 1. DL scheduler block
to use a Priority queuing scheduler in order to give a high
priority for VoIP applications and a lower priority for video
streaming and web browsing applications.
To respond to the second question we compare two resource
assignment methods: The MAX SNR method-which is the
most used method [3], [1]-to our proposal which is based on
channel condition learning.
1) Max SNR gain: This strategy nds the rst available
slots with the maximum SNR gain that satisfy the bandwidth
user request. So it assigns slots with the highest SNR.
Fig. 2. The MAX SNR scheme
2) Channel learning mechanism: This scheme is the most
used one because it is simple to implement but the choice of a
resource unit could be very selsh if there are other users that
they need the same slot than this user, this scheme is based
on rst in rst out mechanism. Our proposal is based on a
channel segregation approach, in fact in order to satisfy the
number x of bit required for the user k packet scheduled to
be served we apply the following steps to select slot :
a) Step 0: The sort: As we consider that we have three
modulation and coding schemes described in table I, the rst
step is to sort slot into 3 sets, namely I
1
, I
2
and I
3
which
respectively represent the set of slots that their SNR
k,s
is in
]22.7, +[ dB, ]16.4, 22.7] dB and [6.4, 16.4] dB and which
respectively corresponds to the modulation [64-QAM,
3
4
], [64-
QAM,
2
3
] and [16-QAM,
1
2
]. We denote by P
k,s
the priority
selection function which characterize a degree of a priority of a
given slot s for a given user k described as follows:P
k,s
=
N
s
N
t
.
Where N
s
is the number of success and N
t
is the number of
the allocation try. The value of this parameter is initially set
to 1.
b) Step 1: The selection: After sorting, if there are
more than one slot candidate in I
1
, select the candidate
c that have the maximum priority, c corresponds to c =
arg[max
sI1
(P
k,s
)], and update P
k,s
. If I
1
is empty, choose
slot from set I
2
. If there are more than one slot candi-
date in I
2
, select the candidate c corresponding to c =
arg[max
sI2
(P
k,s
)], and update P
k,s
. If I
2
is empty, choose
slot from set I
3
. If there are more than one slot candi-
date in I
3
, select the candidate c corresponding to c =
arg[max
sI3
(P
k,s
)], and update P
k,s
. Repeat until the x bits
requested by a user k are satised. Note that the number of
bits that a slot could transmit is easily deduced by equation 2.
c) step 2: The priority function update: The priority
increases when a slot allocation is a success and decreases
if not. A slot selection is called a success when a selected slot
which belongs to a set I
i
, i {1, 2, 3} is kept in the same set
in the next frame or due to the varying channel condition it
becomes belonging to a set I
j
, j {1, 2, 3} and j < i. In other
words a success is when a channel condition of the selected
slot becomes better (i.e the corresponding SNR increases). It
is called a try when the channel condition in the next frame
becomes bad and the selected slot which belongs to a set I
i
,
i {1, 2, 3} becomes in j {1, 2, 3} and j > i .
In case of success we increase a number of success and
the number of try as follows: N
s
= N
s
+ 1 and N
t
= N
t
+
1.Otherwise we increase only the number of tries :N
t
= N
t
+1.
The aim is to promote the selection of the best channel while
keeping some stability in order to not change too much the
modulation scheme.
VI. PRELIMINARY RESULTS
A. Simulation parameters
To evaluate our proposal we carried out discrete event simu-
lation experiments using a MATLAB toolbox called eventSIM.
The purpose of this study is to compare two resource allocation
schemes described in section V namely our proposal and the
most used allocation resource scheme called MAX SNR. The
topology of the simulated network consists of a BS with
12 mobile SS and 24 mobile SS. These SS are randomly
distributed around the BS, and they turn around a BS. The
mobile SS velocity vary from 0.1 to 20 m/s and the trajectory
is a perfect circle with radius varying from 1m to 2 km.
The duration time of our simulation is 20s.We choose system
parameters corresponding to the mobile WiMAX prole, with
10 MHz bandwidth and an FFT size of 1024. The mobile
WiMAX frame with 5ms duration provides 69*4 units of phys-
ical resource or OFDMA slots. The base station provides the
following applications to MSS: In our preliminary evaluation
Classes Mean rate MSSs in Sim. 1 MSSs in Sim. 2
Telephony: UGS 64 kbps 1 2
Video: rtPS [0.5, 1] Mbps 2 4
FTP: nrtPS [1, 2] Mbps 5 10
VOIP: ErtPS 64 kbps 4 8
TABLE III
SIMULATION PARAMETERS
we focused on two evaluation parameters: the average data
rate of each MSS and the frame occupation ratio.
B. Analysis
Figures 4 a, b, 3 c and d show respectively the varying of
user average rate function of frame index obtained when we
apply MAX SNR method and our proposal for the 2 simulation
scenarios. In These gures, we retrieve the same applications
characteristics described in III, so both of these method satisfy
the QoS constraints in terms of average rate.
0
100000
200000
300000
400000
500000
600000
2 4 6 8 10 12
d
a
ta
ra
te
b
p
s
Index user
g(a) g(b)
Fig. 3. Users average data rate for MAX SNR and proposed scheme for 12
obtained by MAX SNR (a) and the proposed scheme (b)
0
50000
100000
150000
200000
250000
300000
350000
0 5 10 15 20 25
d
a
ta
ra
te
b
p
s
Index user
0
50000
100000
150000
200000
250000
0 5 10 15 20 25
d
a
ta
ra
te
b
p
s
Index user
g(c) g(d)
Fig. 4. Users average data rate for MAX SNR and proposed scheme for 24
destinations obtained by MAX SNR (c) and the proposed scheme (d)
But the improvement that provides our proposal is rep-
resented in the frame occupation ratio variation function of
the frame index as we see in gure 5 and 6 . Figures a, b,
c and d show respectively that the average occupation ratio
obtained by the MAX SNR strategy and the proposal scheme is
respectively around 0.90 and 0.95 for simulation 1 and around
0.80 and 0.85 for simulation 2.
VII. CONCLUSION
In this paper we propose a combined scheduler block with
assignment slot technique based on channel learning. We
compare the proposed solution with a MAX SNR assignment
and we obtain with our proposal an improvement of the frame
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 20 40 60 80 100
F
ra
m
e
o
c
c
u
p
a
tio
n
ra
tio

Index user
g(a) g(b)
Fig. 5. Users average data rate for MAX SNR and proposed scheme for 12
obtained by MAX SNR (a) and the proposed scheme (b)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 20 40 60 80 100
F
ra
m
e
o
c
c
u
p
a
tio
n
ra
tio

Index user
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 20 40 60 80 100
F
ra
m
e
o
c
c
u
p
a
tio
n
ra
tio

Index user
g(c) g(d)
Fig. 6. The frame occupation ratio for 24 destinations obtained by MAX
SNR (c) and the proposed scheme (d)
occupation ratio. As future works, we will improve the sched-
uler block by introducing other scheduling algorithms like
Decit Round Robin that provides fairness between variable
packet size ows, we will compare our proposed assignment
technique with other resource allocation techniques and we
will investigate other performance parameters like delays,
packet loss and jitter.
REFERENCES
[1] P. G. P. M. Jeffrey G. Andrews, Fundamentals of WiMAX Understanding
Broadband Wireless Networking. Prentice Hall Communications Engi-
neering and Emerging Technologies Series, 2007.
[2] I. S. 802.16e 2005, IEEE Standard 802.16e-2005, Amendment to IEEE
Standard for Local and Metropolitan Area Networks - Part 16: Air Inter-
face for Fixed Broadband Wireless Access Systems- Physical and Medium
Access Control Layers for Combined Fixed and Mobile Operation in
Licensed Bands. 2005.
[3] M. Einhaus and O. Klein, Performance evaluation of a basic ofdma
scheduling algorithm for packet data transmissions, p. 7, Jun 2006.
[4] M. Einhaus, O. Klein, and B. Walke, Comparison of ofdma resource
scheduling strategies with fair allocation of capacity, p. 5, Jan 2008.
[5] Y. A. H. Akaiwa, Channel segregation-a self-organized dynamic channel
allocationmethod: application to tdma/fdma microcellular system, tech.
rep., Selected Areas in Communications, IEEE Journal, Volume 11, Issue
6, Aug 1993 Page(s):949 - 954.
[6] I.katzela and M.Naghshineh, Channel assignement schemes for cellu-
lar mobile telecommunication systems: a comprehensive survey, IEEE
personal communications, June 1996.
[7] I. Z. S. E. B. A. J. Wong, A low complexity algorithm for proportional
resource allocation in ofdma systems, Signal Processing Systems, 2004.
SIPS 2004. IEEE Workshop on, pp. 16, 13-15 Oct. 2004.
[8] S. Chung and A. Goldsmith, Degrees of freedom in adaptive modulation:
A unied view, 2000. Submitted to IEEE Trans. Commun., January 2000.