QoS Criteria in IEEE 802.

16 Collision Resolution
Protocol
Abdelillah Karouit and Abdelkrim Haqiq
Department of Mathematics and Computer
Hassan 1
st
University
Faculty of Sciences and Techniques, Settat, Morocco
e-NGN research group, Africa and Middle East
Email: {arkaouit ; ahaqiq}@gmail.com
Luis Orozco Barbosa
Albacete Research Institute of Informatics
University of Castilla La Mancha (UCLM)
02071 Albacete, Spain
Email: luis.orozco@uclm.es
AbstractIn this paper, we study the collision resolution pro-
tocol of the IEEE 802.16 standard for metropolitan broadband
wireless access systems. The standard species the use of a
random access scheme based on the slotted binary exponential
backoff (BEB) algorithm during the bandwidth request phase. We
analyze the distributed choice of the retransmission probabilities
used by the BEB protocol. We follow a cooperative team problem
approach towards the analysis of the throughput and the delay
during the bandwidth reservation process. A Markov chain
analysis is used to obtain the optimal retransmission probabilities,
the optimal value of the initial backoff window size, Wmin, the
throughput and the expected delays experienced by a subscriber
station (SS) for placing its bandwidth requests. the results shows
that the delay of backlogged messages becomes very large in
heavy trafc when our objective is to minimize the expected
delay of all transmitted messages. This motivate us to pose a
problem of minimizing the a expected delay (or maximizing
the throughput) subject to constraints on the expected delay of
backlogged messages as a multicriterion problem.
Index TermsIEEE 802.16, Binary exponential backoff, Team
problem, Markov chain, QoS Criteria.
I. INTRODUCTION
The IEEE 802.16 WiMaX standard [1] is supposed to
play an important role in rapid and ubiquitous adoption
of broadband wireless access systems worldwide. In the
centralized point-to-multipoint architecture of WiMaX,
subscriber stations (SSs) share the uplink to a base station
(BS) on demand basis. This means that if a SS requires
bandwidth, it informs the BS by means of transferring a
request message. The BS scheduler accepts the requests from
different SSs and grants them the transmission opportunities
in time slots by using some scheduling algorithms, which
should take into account the requirement of each SS and the
available channel resources. These grants are based on the
negotiated quality of service (QoS) agreements. Two main
methods to provide transmission opportunities are suggested
: centralized polling and contention-based random access.
According to the IEEE 802.16, the latter may be implemented
in two different ways. The rst and most used solution is
to give all stations an opportunity to access all available
contention slots ; another typical solution is to group some
stations together and assign disjoint subsets of all contention
slots to each group. Previous results have been obtained to
evaluate its performance as well as to propose to vendors
some efcient ways to use different available tools and
methods. One main line of research is devoted to improving
the operation of mesh-topology network based on the IEEE
802.16 standard [4]-[5], where this system is analyzed mainly
by means of simulations. An analytical model for CSMA
based Point-to-Multipoint (PMP) broadband wireless network
was considered in [6]. However, according to the WiMaX
standard [1] a mandatory random access method in IEEE
802.16 is based on a truncated BEB without any carrier sense
features.
A large number of research papers have focused on the
performance evaluation of the various IEEE 802.16 features
[7]-[8]-[9] . In particular, the bandwidth requests transmission
by a subscriber to reserve a portion of the channel resources is
frequently addressed. A detailed description of the reservation
techniques is known from the fundamental work in [10]. The
standard allows a random multiple access (RMA) scheme at
the reservation stage and implements the truncated binary
exponential backoff (BEB) algorithm for the purposes of the
collision resolution. The asymptotic behavior of the BEB
algorithm has been widely addressed in the literature. In
[11] it was shown that the BEB algorithm is unstable in
the innitely-many subscribes stations case. By contrast,
[12]shows that the BEB is stable for any nite number
of subscribes stations, even if it is extremely large, and
sufciently low input rate. These seemingly controversial
results demonstrate the two alternative approaches to the
analysis of an RMA algorithm [13]. The former is the innite
population model, which studies the ultimate performance
characteristics of an RMA algorithm. The latter is the nite
population model that addresses the limits of the practical
algorithm operation. An exhaustive description of both models
may be found in [14] and [15].
BEB algorithm within the framework of IEEE 802.16
was rst investigated in [16] under a saturation condition
assumption by using an analytical approach similar to [18].
However, this assumption is too strong since networks do
not typically operate in saturation conditions. Most Internet
applications exhibit bursty trafc characteristics. A reasonable
model for the fair analysis in more general case of an
arbitrary request arrival rate should consider subscribers in
tow states (active/inactive). Similar idea was introduced in
the analysis of wireless centralized networks, for instance,
wireless ATM in [19]. Following this approach, we do not
consider the packet transmission phase, but limit our scope
to the throughput and delay analysis of bandwidth request
during the reservation process.
In this paper, we undertake a team problem approach.
We obtain the optimal retransmission probabilities in the team
problem to optimize the two parameters of performances
(The throughput and delay of bandwidth request) through an
analytical model. The overviews of the IEEE 802.16 protocol
and its request mechanisms in Section II. We then briey
describe the BEB algorithm in Section III. In Section IV
we formulate and solve the team problem. We then provide
some numerical results in Section V. In section VI we pose
and solve the multicriterion problem of minimizing the
expected delay subject to constraints on the expected delay of
backlogged messages only. Finally we draw our conclusions
in Section VII.
II. IEEE 802.16 MAC PROTOCOL
A. FRAME STRUCTURE
Let us consider the network with a PMP architecture, which
consists of one BS managing multiple SSs. Transmissions
between the BS and SSs are realized in xed-sized frames by
means of time division multiple access (TDMA) / time division
duplexing (TDD) mode of operation. The frame structure
consists of a downlink sub-frame for transmission from the
BS to SSs and an uplink sub-frame for transmissions in
the reverse direction. The Tx/Rx transition gap (TTG) and
Rx/Tx transition gap (RTG) shall be inserted between the
sub-frames to allow terminals to turn around from reception
to transmission and vice versa. In the downlink sub-frame
the Downlink MAP (DL-MAP) and Uplink MAP (UL-MAP)
messages are transmitted by the BS, which comprise the
bandwidth allocations for data transmission in both down-
link and uplink directions, respectively. Another important
management message, which is related to UL-MAP, is called
an Uplink Channel Descriptor (UCD). It can be periodically
transmitted in the downlink sub-frame. The values of the
minimum backoff window, W
min
, and the maximum backoff
window, W
max
, are dened in this message, which are used
by the collision resolution algorithm. The uplink sub-frame
contains transmission opportunities scheduled for the purpose
of bandwidth requests, in which Bandwidth Request (BW-
REQ) messages can be transmitted by the SSs to indicate to
the BS their uplink bandwidth allocation requirements. The BS
manages the number of transmission opportunities through the
UL-MAP message.
B. REQUEST MECHANISM IN IEEE 802.16
Each transmission opportunity may be assigned by the BS
either to exactly one subscriber station or to a group of sta-
tions. In the rst case, a station is provided a so-called unicast
opportunity for issuing its BW-REQ. In other words, the BS
polls a SS to allow it to transmit the request in a contention-
free manner. In the latter case random access algorithm is used
by the group of SSs to contend for the common transmission
opportunities and resolve possible collisions. As mentioned
earlier, the mandatory method of contention resolution, whose
inclusion is mandatory by the standard, is based on a truncated
binary exponential backoff, with the initial backoff window
and the maximum backoff window controlled by the BS. This
algorithm is described in detail in the next section.
The information whether BW-REQ message is successfully
transmitted or distorted (because of collision or noise) is not
explicitly transmitted by the BS in the downlink. The standard
does not specify how the SSs get to know how outcome of
their transmission. It might be based on the correspondence
between the amount of the resources assigned to the given
SS and the amount of the resources it has asked for in the
transmitted BW-REQ message.
III. BINARY EXPONENTIAL BACKOFF
In distributed multiple access, a simple yet effective
random backoff algorithm is widely used to avoid collisions.
In particular, the truncated binary exponential backoff
algorithm adjusts the contention window size dynamically
according to the collision intensity. This algorithm steps may
be summarized as follows:
Rule 1.1. If a new bandwidth request arrives to a SS
in the frame t 1 and this SS has no other pending requests,
it transmits the request in the frame t (transmission attempt).
The slot for the request transmission is sampled uniformly
from the number of contention slots dedicated to the group
to which the SS belongs to. Notice, that in case of broadcast
polling the SS may choose between all the contention slots
K of the frame t, whereas in case of multicast polling the
choice is narrowed to L slots of the respective multicast group.
Rule 1.2. If a request is ready for retransmission at the
beginning of the frame t at its i th retransmission attempt
(i > 0), a SS chooses a number (backoff counter) in the
range {0, 1, ..., 2
min(m,i)
W 1} uniformly, where W and
m are the parameters of the BEB algorithm, named initial
contention window and maximum backoff stage respectively
and i is the number of collisions this request suffered from
so far. The SS then defers the request retransmission for
the chosen number of slots, accounting only for the slots
dedicated to its group.
Rule 2.1. If, after receiving the feedback from the BS, the
SS determines that its last request collided, it increments the
collision counter i for this request. If this counter coincides
with the maximum allowable number of retransmission
attempts, then the request together with the corresponding
data packet is discarded and the collision counter is reset to
i = 0.
Rule 2.2. If, after receiving the feedback from the BS,
the SS determines that the (re)transmission of the last request
was successful, it resets the collision counter to i = 0.
IV. PROBLEM FORMULATION
We use a Markovian model based on [21]-[2]-[8]. We
assume that there are a nite number of SSs having a buffer
sufcient to store exactly one request. We have set the
message size used by all SSs equal to one packet. In terms
of bandwidth requesting process, each SS may be either in
one of two states: thinking or backlogged. A thinking SS has
no message ready for transmission and may generate one
during a frame with the probability q
a
. Once a new message
is generated, the SS enters the backlogged state where no
new arrivals are possible. This corresponds to the real system
where after the rst arrival a SS starts the contention process
and the subsequent arrivals are irrelevant to establish the
sought contention delay. Once the transmission of a BW-Req
is successful, the SS enters the thinking state and is able to
generate new messages. In each slot the backlogged SS i
attempt to retransmit a BW-Req with probability q
i
r
. Similarly
to [20], we consider that the probability q
i
r
is given as
function of backoff window minimal W
min
, q
i
r
=
2
W
min
+ 1
.
The determination of the above random time can be
considered as a stochastic control problem. The information
structure, however, it is not a classical one: the SSs do not
have knowledge of the full state information. They do not
know the number of backlogged SSs, nor they do know the
number of messages involved in a collision. We study this
control problem as a team problem, where there is a common
goal to all SSs in the network: to maximize the network
throughput or to minimize the message transmission time.
We shall restrict in our problem to simple policies in which
q
i
r
does not change over time. Since SSs are symmetric, we
shall further restrict to nding a symmetric optimal solution,
i.e., all stations attempt to retransmitting with the same
probability, denoted from now on by q
r
We use as the state of the system the number of backlogged
SSs (or equivalently, of backlogged messages) at the
beginning of a slot, and denote it by n. For any choice
of values q
r
[0, 1], the state process is a Markov chain
that contains a single ergodic chain and possibly transient
states as well. Dene q
r
to be the vector of retransmission
probabilities where all SSs whose jth entry is q
j
r
. Let (q
r
)
be the corresponding vector of steady state probabilities
where its nth entry,
n
(q
r
), denotes the probabilities of n
backlogged SSs. Let Q
r
(i, n) be the probability that i out of
the n backlogged SSs retransmit at a given slot. Then
Q
r
(i, n) =

n
i

(1 q
r
)
(ni)
q
i
r
(1)
Let M denote the number of SSs in the network. The proba-
bility that i unbacklogged SSs transmits a new message in a
given slot, Q
a
(i, n) is then given by:
Q
a
(i, n) =

M n
i

(1 q
a
)
(Mni)
q
i
a
(2)
The transition probabilities of the Markov chain are given by
P
(n,n+k)
(q
r
) =

Q
r
(1, n)Q
a
(k + 1, n 1)+
(1 Q
r
(1, n))Q
a
(k, n) 1 k M n
(1 Q
r
(1, n))Q
a
(0, n)+
Q
r
(1, n)Q
a
(1, n 1) k = 0
Q
r
(1, n)Q
a
(0, n 1) k = 1
0 Otherwise
A. Expected Delay of Transmitted Messages (E.D.T.M.)
The throughput, dened as the sample average of the
number of messages that are successfully transmitted, is given
by:
thp(q
r
) =
M

n=1

n
(q
r
)[P
n,n1
(q
r
) +Q
a
(1, n 1).Q
r
(0, n)]
+
0
(q
r
)Q
a
(1, 0) (3)
Note: The rst term of the equality follows from the fact
that if the state at the beginning of the slot is n > 0 then
a departure of a backlogged message occurs during that slot
with probability P
n,n1
, while the departure of a new arriving
message occurs with probability Q
a
(1, n 1).Q
r
(0, n). The
second term takes into account that if n = 0 then a new arriv-
ing message can be successfully transmitted with probability
Q
a
(1, 0).
The team problem is therefore given as the solution of the
optimization:
max
qr
thp(q
r
) s.c

(q) = (q
r
)P(q
r
)

n
(q
r
) 0, n = 0, ...., M

M
n=0

n
(q
r
) = 1
(4)
A solution to the team problem can be obtained by computing
recursively the steady state probabilities [21], and thus obtain
an explicit expression for thp(q
r
) as function of q
r
.
Existence of a solution . The steady state probabilities (q
r
)
are continuous over 0 < q
r
1. Since this is not a close
interval, a solution does not exist. However, as we restrict to
the closed interval q
r
[, 1] where > 0, an optimal solution
indeed exists. Note also that the limit lim
qr0
(q
r
) exists
since (q
r
) is a rational function of q
r
at the neighborhood
of zero. Therefore for any > 0, there exists some q
r
> 0
which is -optimal. (q

r
> 0 is said to be -optimal if it
satises thp(q

r
) thp(q
r
) for all q
r
[0, 1].
In order to calculate the expected delay of the transmitted
messages of the team problem, we should rst compute the
average number of backlogged messages which is given by:
S(q
r
) =
M

n=0

n
(q
r
)n (5)
where the throughput of the number of messages that are
successfully transmitted is given by 3
The delay of transmitted messages D, dened as the average
time slots that a message takes from its source to the receiver
can then be obtained by applying Littles result:
D(q
r
) = 1 +
S(q
r
)
thp(q
r
)
(6)
Note that the rst term accounts for the rst message trans-
mission from the SS.
The team problem E.D.T.M. is therefore given as the solution
of the optimization:
min
qr
D(q
r
) s.c

(q
r
) = (q
r
)P(q
r
)

n
(q
r
) 0, n = 0, ....M

M
n=0

n
(q
r
) = 1
(7)
Existence of a solution The same reasoning as above is
valid for any other objective function and in particular
herein towards the minimization of the delay of transmission
message ).
B. Expected Delay of Backlogged Messages (E.D.B.M)
Another relevant quantity in this context is the expected
delay of backlogged message D
c
(E.D.B.M) which is dened
as the average time, in slots, that a backlogged message takes
to go from the SS to the BS. To compute it, we need to
calculate the total throughput of backlogged messages which
is dened as the sample average of the number of new arriving
messages that become backlogged. This is given by
thp
c
(q
r
) =
M1

n=0
Mn

i=1
P
n,n+i
(q
r
)i
n
(q
r
) (8)
Applying Littles result, the expected delay of message that
arrive and become backlogged is given by
D
c
(q
r
) = 1 +
S(q
r
)
thp
c
(q
r
)
(9)
The team problem E.D.B.M is therefore given as the solution
of the optimization:
min
qr
D
c
(q) s.c

(q
r
) = (q
r
)P(q
r
)

n
(q
r
) 0, n = 0, ....M

M
n=0

n
(q
r
) = 1
(10)
Fig. 1. Optimal delay as a function of the arrival probabilities qa for
M=2, 4, 10, 20
Fig. 2. The optimal retransmission probabilities as a function of the
arrival probabilities qa for M = 2, 4, 10, 20
V. NUMERICAL RESULTS
In this section we shall obtain the optimal values of the
retransmission probabilities and the optimal size of the backoff
window W
min
. We further investigate the dependence of the
throughput and delay on the arrival probabilities q
a
and on the
number of SSs.
A. Expected Delay Of Transmitted Messages
We consider the team problem in which the performances
to optimize are the throughput and the expected delay of
transmission messages. In Figures 1 and 2 the expected
delay of transmitted messages and the optimal retransmission
probabilities are plotted versus the arrival probability q
a
for
the team problem.
The results show that the optimal retransmission probabili-
ties, q
r
, should be reduced as the arrival probabilities increase
or as the number of SSs is large. In fact, the results show
that for arrival probabilities as low as 0.05 and for M=20, the
retransmission probability should be reduced drastically.
Figure 3 shows the throughput for M = 2,4,10,20 for the
team problem, as a function of the arrival probability q
a
. Under
heavy trafc, the throughput decreases drastically as a function
of the trafc intensity and much faster in the case of a network
consisting of a larger number of SSs. Figure 4 shows the
optimal window backoff W
min
for M = 2,4,10,20 and as a
function of the arrival probabilities. As expected, when the
network is exposed to heavier load, larger number of active
SSs or higher arrival rates, the window has to be widen in
order to reduce the collision probability.
Fig. 3. Optimal throughput as a function of the arrival probabilities
qa for M=2, 4, 10, 20
It results interesting that under overload conditions, the
optimal throughput converges independently of the number
of active SSs. For instance, Figures 5 shows the optimal
throughput versus the number of SS for q
a
= 0.7, 0.9. From
the gure, it is clear that the throughput indeed converges to
0.1438 (resp. 0.0510) for q
a
= 0.7 (resp. q
a
= 0.9). In fact,
under heavy trafc and a large number of SS, the optimal
retransmission probability is seen to be . Thus, the steady
probabilities are then close to
M
= 1, and
n
= 0,
n < M. In this case the expected delay is given by
D = 1 +
1
M(1 q
r
)
M1
q
r
(1 q
a
)
and the throughput is given by:
thp(q
r
) = M(1 q
r
)
M1
q
r
(1 q
2
a
)
This explains the convergence of delay to innity, and the
convergence of throughput to 0 when the arrival probability
approach to 1 see Figures 1 and 3, respectively .
Fig. 4. The optimal window backoff W
min
as a function of the arrival
probabilities qa for M = 2, 4, 10, 20
Fig. 5. Optimal throughput as a function of the number of SS for
qa = 0.7, 0.9
B. Expected Delay of Backlogged Messages
We consider the team problem in which the performance
metrics to optimize are the throughput and the expected delay
of backlogged messages given by 8 and 10, respectively.
We compare the performance for different scenarios: team
problem with E.D.T.M and team problem with E.D.B.M. In
Figure 7, we plot the throughput using the optimal backoff
window size W
min
for the team problem with E.D.T.M
and E.D.B.M versus the arrival probability q
a
for M=10,
respectively. As shown in Figure 7, the throughput of the
E.D.B.M. team problem decreases drastically at high arrival
probabilities. The backoff window W
min
remains almost
constant as q
a
increases, see Figure 4.
Another interesting observation can be made when compar-
Fig. 6. The expected delay at optimal backoff window W
min
with the
problems of minimizing E.D.T.M and E.D.B.M respectively, as a function
of the arrival probabilities qa for M = 10
ing the throughput of for the E.D.T.M and E.D.B.M cases.
Figure 6 shows the optimal throughput for the E.D.T.M and
E.D.B.M cases as a function of q
a
for M=10. We observe that
a large percentage of messages go through the backlogged
process even under light load conditions. Furthermore, Figure
7 shows the expected delay of messages and non-backlogged.
As expected, the delay of backlogged messages when the
arrival probability increases 1 is higher than the delay of
transmitted messages.
Fig. 7. The throughput at optimal backoff window W
min
for the
problem of minimizing E.D.T.M and E.D.B.M, respectively, as a function
of the arrival probabilities qa for M = 10
VI. QOS CRITERIA
As shown in last section, the delay of backlogged messages
become very large in heavy trafc when our objective is to
minimize the expected delay of all transmitted messages (or
equivalently, when maximizing the throughput). Thus we can
distinguish two separate QoS criteria: the total expected delay
of transmitted messages as well as the delay of backlogged
messages only. We then consider the team problem of mini-
mizing the expected delay of transmitted messages subject to
a constraint on the expected delay of backlogged messages.
The new team problem that has a new QoS constraint is given
by:
min
qr
D(q
r
) s.t D
c
(q
r
) < d (11)
d is some constant. We shall denote by d
max
the smallest
value of d that insures that there is a solution to 11 for any
value of q
a
. There are indeed values of d for which , for some
q
a
, the problem 11 dont admit a solution .
Note that due to 6, maximizing the throughput is equivalent
to minimizing S(q
r
). Hence, we can deduce the fact that the
maximizing the throughput is equivalent to minimizing the
expected delay. Therefore the retransmission probabilities that
solve 11 also solve the problem:
max
qr
thp(q
r
) s.t D
c
(q
r
) < d (12)
Figure 8 shows the expected delay of transmitted messages
Fig. 8. The optimal delay for the team problem as a function of the
arrival probabilities qa, with/without delay constraint for M=10.
with and without constraints for M = 10. We observe that for
all values of q
a
, the expected delay of transmitted messages
without QoS constraint equal to that with QoS constraints,
the constraints does not allow the system to decrease the
delay at the expense of the backlogged messages, there is no
performance degradation in the expected delay of transmitted
messages. For d = 120, we dont have any feasible strategy
satisfying the QoS for some value of q
a
, the feasible region
of strategies is not empty under the condition q
a
0.4, but
for d = d
max
= 563.5395, we have a feasible region for all
values of q
a
because d
max
is the maximum expected delay
of backlogged messages at optimal retransmission policy for
q
a
[0, 1] which explains the existence at leat one feasible
strategy for any values of q
a
.
VII. CONCLUSION
We have used a game theory approach for setting up
the retransmission probabilities and window size used in the
collision resolution protocol of the IEEE 802.16 standard. Our
objective has been to minimize the expected delay and max-
imize the throughput during the bandwidth request process.
Our results show that as the request arrivals increase, the SSs
should reduce the retransmission probability by widening their
backoff windows. We have also observed that the delay of the
backlogged message rapidly increases under heavy trafc load
conditions, and the new constraint does not allow the system to
decrease the delay at the expense of the backlogged messages,
there is no performance degradation in the expected delay of
transmitted messages.
REFERENCES
[1] IEEE Std. 802.16-2004, IEEE standard for local and metropolitan area
networks, Air Interface for Fixed Broadband Wireless Access Systems.
[2] E. Altman, D.Barman, R. El Azouzi, T. Jimenez, A game theoretic
approach for delay minimization in slotted aloha, ICC, 20-24, Paris,
France, June 2004.
[3] E. Altman, R El Azouzi and T. Jimenez, Slotted Aloha as a Stochastic
Game with Partial Information, WiOpt03, Sophia Antipolis, France,
March 3-5, 2003.
[4] S. Redana , M. Lott Performance Analysis of IEEE 802.16a in Mesh
Operation Mode, The 13th IST SUMMIT, Lyon, France, June 2004.
[5] M. Lott, S. Redana and M. Carlozzo Reducing the delay of IEEE
802.16a in multi-hop scenarios World Wireless Congress, USA, May
2005.
[6] S. Guo J. Sydor, A. Brandao Performance Analysis of the P2MP
Metropolitan Area Broadband Wireless Access Network, IEEE PIMRC
2003.
[7] Y. Fei and G. Pujolle, Performance evaluation of polling mechanisms in
IEEE 802.16 networks, in proc of IEEE ISWPC 2009, pp. 1-5, (2009).
[8] C. Cicconetti, A. Erta, L. Lenzini, and E. Mingozzi, Performance
Evaluation of the IEEE 802.16 MAC for QoS Support, IEEE Trans
on Mobile Computing, pp. 26 - 38 (2007).
[9] A. Ebian, M. Shokair, and K. H. Awadalla, Performance evaluation
of Wimax system using conolutional product code (CPC), Progress In
Electromagnetics Research C, Vol. 5, 125-133, (2008).
[10] I. Rubin, Access-control disciplines for multi-access communication
channels: Reservation and tdma schemes. IEEE Transactions on Infor-
mation Theory 25(5), 516-536 (1979).
[11] D. Aldous, Ultimate instability of exponential back-off protocol for
acknowledgment based transmission control of random access commu-
nication channels. IEEE Transactions on Information Theory 33(2),
219-233 (1987)
[12] J. Goodman, A. Greenberg, N. Madras, P. March, Stability of binary
exponential backoff. Journal of the ACM 35(3), 579-602 (1988)
[13] M. Paterakis, L. Georgiadis,P. Papantoni-Kazakos, On the relation
between the nite and the innite population models for a class of raas.
IEEE Transactions on Communications 35, 1239-1240 (1987)
[14] B. Chlebus, Randomized Communication in Radio Networks. In: P.
Pardalos S. Rajasekaran, J. Reif, Rolim, Handbook of Randomized
Computing, vol. 1, pp. 401-456 (2001)
[15] B. Tsybakov, Survey of user contributions to random multiple-access
communications. IEEE Transactions on Information Theory 31(2), 143-
165 (1985)
[16] A. Vinel , Y. Zhang , M. Lott , A. Turlikov Performance Analysis of
the Random Access in IEEE 802.16, in Proc. of the IEEE PIMRC05,
Berlin, Germany, September 2005.
[17] D. Staehle, R. Pries, A. Vinel and A. Mder, Performance evaluation
and parameterization of the IEEE 802.16 contention-based CDMA
bandwidth request mechanism for the OFDMA physical layer In Proc
of the 12th ACM, pp: 374-383 (2009).
[18] G. Bianchi Performance Analysis of the IEEE 802.11 Distributed Coor-
dination Function, IEEE Journal on Selected Areas in Communications,
V18, N3, 2000, pp 535-548.
[19] D. Cortizo, J. Garcia, C. Blondia , B. VanHoudt FIFO by Sets ALOHA
(FS-ALOHA): A Collision Resolution Algorithm for the Contention
Channel in Wireless ATM Systems, IFIP WG 7.3 Performance, Istan-
bul, 1999.
[20] S. Andreev, A. Turlikov, A. Vinel, : Contention-based polling efciency
in broadband wireless networks. In: International Conference on Ana-
lytical and Stochastic. (2008)
[21] D. Bertsekas, R. Gallager, Data Networks, Prentice Hall, Englewood
Cliffs, NJ, 1987.