Anda di halaman 1dari 6

54 (IJCNS) International Journal of Computer and Network Security,

Vol. 1, No. 3, December 2009

Optimal Channel Allocation in Wireless Sensor


Network using OFDMA
Mohd. Sabir1, Rakesh Kumawat2 and Dr. V.S.Chouhan3
1
Electronics & Communication Department, Sobhasaria Engineering College,
Sikar (Raj.), India
Sabii.sankhla@gmail.com
2
Electronics & Communication Department, Sobhasaria Engineering College,
Sikar (Raj.), India
rakesh130186@gmail.com
3
Prof. & Head, Electronics & Communication Department,
I.E.T. Alwar (Raj.), India
vsc_iet@rediffmail.com

other optimal algorithms, which consider perfect CSI, when


Abstract: Wireless Sensor Network is a Wireless network of
error in channel estimation is high. We then consider the
Spatially Distribution of Various Autonomous Device which
using Sensors to Co-operative Monitor Physical or real problem of Designing a centralized scheduler in data
environment conditions like temp., light etc. in wireless sensor transmission that maximizes the average sensor node
network, we require a fast data transfer between one node to throughput, and satisfies the average delay constraint of
another node in critical real-time applications, in present ,we individual node. We assume a data transfer between one
face slow data rate between one sensor node to another sensor node to another node in wireless sensor node is use single
node.In this Paper, we want to solve this problem and propose channel and multiple queues at the base station. We propose
unique approach for optimal channel allocation in wireless an online throughput optimal delay constraint satisfying
sensor network using OFDMA. We assume that sensor node (OTODCS) algorithm for this model, and we will show
knows channel state information (CSI) partially, and perfect through simulation that it performs better than multi-queue
queue state information (QSI). Our Objective is to minimize TDM system under the same set up. Further we extend the
long-term average packet delay over multiple time slots.
same problem to a problem with multicasting data transfer
Keywords: Wireless sensor network, channel state with optimal channel allocation. We will show via
information, data rate, OFDMA, Channel allocation. simulation that we will be get, in both throughput and delay
in comparison with the other online algorithm with single
1. Introduction server. OFDM is a promising technique for broadband
wireless communication in a multi-path environment with
A wireless sensor network (WSN) is a wireless network frequency selective fading. OFDM achieves high spectral
consisting of spatially distributed autonomous devices using efficiency in multi user environment by dividing the total
sensors to cooperatively monitor physical or environmental available bandwidth to narrow-bands in efficient way. This
conditions, such as temperature, sound, vibration, pressure, allows the mobile to spread information selectively in order
motion or pollutants, at different locations. The to avoid sub-bands where fading occurs. For the user
development of wireless sensor networks was originally downloading data, it is required to have as much as
motivated by military. Applications such as battlefield throughput as possible. OFDM achieves high spectral
surveillance. However, wireless sensor networks are now efficiency by converting a frequency selective channel into a
used in many civilian application areas, including set of parallel non-interfering frequency flat channels on
environment and habitat monitoring, healthcare which efficient power allocation can be done. When we
applications, home automation, and traffic control. In assume that each user has infinite amount of data stored in
addition to one or more sensors, each node in a sensor buffer water filling is throughput optimal. However, when
network is typically equipped with a radio transceiver or considering stochastic packet arrival with finite rate, water
other wireless communications device, a small filling is not throughput optimal. However in all previous
microcontroller, and an energy source, usually a battery. work they assume perfect CSI at the base station, but in
The envisaged size of a single sensor node can vary from practice only the inaccurate estimate is available and
shoebox-sized nodes down to devices the size of grain of optimal policies may not be robust to this estimation error. It
dust, although functioning 'motes' of genuine microscopic is interesting to study performance of these scheduling
dimensions have yet to be created. A sensor network algorithms under partial CSI.
normally constitutes a wireless ad-hoc network, meaning
that each sensor supports multi-hop routing algorithm
(several nodes may forward data packets to the base 2. Problem Definition and assumption
station).We propose a heuristic Approaches, which we show
We consider a multi-queue multi-server system with
through simulation performs better, in comparison with the
stochastic connectivity’s. There are N queues (users) and K
(IJCNS) International Journal of Computer and Network Security, 55
Vol. 1, No. 3, December 2009

servers (sub-carriers). Fixed-size packets arrive column matrix permutation Ππ corresponding to a


stochastically for each user and are transmitted over a set of permutation π is defined as, for any j and k Є {1, ...,N},
allocated servers. Each user has an infinite buffer to store π (xj) = xk ↔ Ππ (yj) = yk
the data packets that cannot be immediately transmitted. Using the above notation and definition, we make the
The system is time slotted. The users have the same priority following symmetry assumptions on the arrival and
and are symmetric, i.e., they have statistically identical connectivity processes:
arrival and channel connectivity processes. At the beginning A(1) The packet arrival processes [a(n)] to users’ queues
during each time slot are i.i.d across time slots. The packet
of each time slot, the assignment of servers to users is
arrival processes are symmetric or exchangeable, i.e., the
instantaneous and made by a centralized resource manager.
joint probability mass function is permutation invariant, i.e.,
The resource manager has perfect knowledge of the current
P[a(n) = π(x)] = P[a(n) = x]
queue backlogs and the connectivity’s which are assumed to for any n, vector x, and permutation π.
be constant during a time slot but varying independently A(2) The connectivity profiles [C(n)] are i.i.d across time
over time slots (e.g. block fading model). We do not allow slots and exchangeable across server, i.e. for any n, matrix
sharing of any server and assume no error in the Y , and column by column permutation matrix Ππ
transmission. Assumption A(2) is valid when channel and the mobility
We use the following convention: lower case for scalar, bold creates a homogeneous environment for all users. Note that
faced lower case letters for row vector, upper case letter for A(1) and A(2) imply independence across time slots but not
matrices and scripted upper case letter for space of matrices. across users, i.e., at a given time the arrival to various
b(n) = (b1, b2, ..., bN) : Backlogs (in units of packets) of each queues need not to be independent. Note that we will some
queue in the beginning of time slot n. times use the vector w(n) to mean the matrix product
• a(n) = (a1, a2, ..., aN) : Stochastic number of fixed-length 1W(n).
packets arrived to each queue during a time slot n. The new
packet arrivals at time n can be served only at time n + 1 or 4. Mathematical Formulation Problem (P)
after.
• C(n) = [ci,j ]: the K-by-N stochastic connectivity matrix at Consider the system described above. We wish to determine
time slot n where ci,j Є{0, 1, ..., cmax < ∞} denotes the a Markov server allocation policy σ that minimizes the cost
maximum number of packets sub-carrier i can serve from function at the finite horizon T:
queue j at time n, and we further assume that ci,j takes =E | ]
discrete values form 0 to cmax, where cmax is the maximum
number of packets that a user j can transmit in channel i. Where summarizes all information available at time zero
• ξ(n) = [ξi,j ]: the K- by- N matrix , each element of this and denotes the cost under Markov policy σ over horizon
matrix is a circularly symmetric complex Gaussian random T.
variable. =
• W(n) = [wi,j ]: K-by-N allocation matrix at the beginning where the cost function ø(b) = g(bj) and g is a
of time slot n where wi,j {0, 1} and wi,j = 1 denotes that convex and strictly increasing function. Note that when g is
sub-carrier i is assigned to serve queue j during time n. an identity function, problem reduces to a average total
In our model we assume that there is error in channel backlog (E [ bj(t)]) minimization problem
estimation. We denote the
over horizon T. From Little’s theorem, any optimal policy
Erroneous connectivity matrix by
that achieves minimum average backlog achieves minimum
H(n) = C (n) + ξ (n),
average packet delay as well. Thus, in this special case, the
If above is the equation for channel visible to base station,
study reduces to the study of the average delay
we assume that base station Does MMSE estimation on the
minimization.
received channel matrix H(n). And with MMSE estimate.
5. Proposed Solutions to the Problem
3. Problem Formulation and Assumption
In this section we present a heuristic policy as a solution to
The channel estimation at base station is given by [14]: the above problem. We modify maximum throughput load
Ĥ(n) = SNRest / 1+ SNRest {C(n) + ξ(n)} balancing (MTLB) algorithm and maximum weighted
Where matching (MWM) algorithm proposed by Kittipiyakul and
SNRest = E[|C|2] / E[|C|2] + E[|ξ|2] Javidi in [5] to take into account the error in channel
The dynamics of queue length vectors under allocation W estimation. In [13] authors proved that MTLB policy is a
(n) are described by equation special case of MWM when we assume channel is On/Off.
b(n+1) = [b(n)-1W(n)Θmin C(n),H(n)]+ +a(n),n=1,2… Our algorithm is valid for both On/Off channel and general
where element wise product (W(n) ‫ ּס‬min(C(n),H(n)) is a channel with cmax > 1 by the arguments in [13]. We
matrix [wi,j min(ci,j ,Hi,j)], 1 is a K dimensional row vector assume full queue state information (QSI) and partial
of K ones, and, for a vector, [v+] = [v+ 1 , ..., v+ N] with v+ channel state information (CSI).
j = max{0, vj}. Algorithm:
Definition 1: For a row vector x = (x1, ..., xN) and a matrix • X = {1, 2, ..., K}
Y = [y1, ..., yN] where yj is a column vector, a column by • loop (until stop):
56 (IJCNS) International Journal of Computer and Network Security,
Vol. 1, No. 3, December 2009

– If X = Φ, then stop via simulation that our algorithm performs better than
– (i*,j*) = arg maxi∈X,j∈{1,2,...,N} bj( (Ci,j\ Hi,j)) ; MTLB when error in channel estimation is high and
– if bj Hi,jP (ci,j |Hi,j)> 0 then w∗ i*,j* = 1 channel is On/Off. Figs. 1 to 3 above show the simulated
performance in terms of average queue backlogs for Poisson
– bj* = bj* − min(Hi*, j*, ci*,j*) and X = X − {i∗ }
distribution when the channel is On/Off. The simulation
• Assign W∗ = wi*, j*
results in Fig.2 demonstrate the superior performance of
MTLB algorithm with reliability information in comparison
6. Simulation parameter with simple MTLB algorithm with channel estimation error
Before we provide the details of simulation set up let us first and no reliability information. By reliability information we
discuss the MTLB and MWM algorithm: mean probability with which estimation is correct. In MTLB
(1) MTLB Algorithm: The subcarrier assignment uses the channel is On/Off. It is clear with simulation results in
full information about the queue lengths (full QSI) Fig. 1 that we do well in terms of the average queue backlog
and binary (ON-OFF) information about the when error in channel estimation is high. From Figs. 2 and
channel: a sub-carrier is considered ON if ci, j > 3, it is clear that we gain nothing by reliability information,
cthreshold. Then MTLB policy is used for sub carrier and the performance is almost same as the MTLB even with
allocation. channel estimation error. While Figs. 4 show the simulated
= performance under the same set up with the condition that
channel is more general, i.e., Hi, j ∈ {0, 1, 2, 3}. It is clear
For state (b,ĉ). Compute an MTLB allocation W∗
from the simulations that when channel is general rather
(2) MWM Algorithm: The sub carrier assignment uses
than just On/Off we gain nothing from reliability
full queue state information (full QSI) and full
information. Our heuristic policy performs almost same as
channel state information (full CSI)
the MWM with error in channel estimation.
• X = {1, 2, ..., K}
• loop (until stop):
– If X = Φ, then stop
– (i∗ , j∗ ) = arg maxi,j∈{1,2,...,N} bjci,j ;
– if bj*ci*,j* > 0 then w∗ i*,j* = 1 else stop;
– bj* = bj* − ci*,j* and X = X − i∗
• AssignW∗ = wi*,j*
where ci,j denotes the channel capacity for user j on server i,
K denotes the number of sub carrier and N denotes the
number of user (queues).We now compare the performance
of MTLB algorithm and MWM algorithm in [5] with our
heuristic policy. We consider downlink OFDM consisting of
32 statistically independent and identical users, and 128 sub
carriers. We assume equal power distribution among all
users, and further assume that the loss of throughput due to
equal power allocation is tolerable [15]. We can represent
the number of packets per time slot the sub carrier i can Figure 1. Average queue backlog for MTLB policy with
transmit to user j, in terms of the erroneous channel gain gi,j, SNR = 0 dB
as follows:

Hi, j =

where D is the number of QAM symbols per channel in a


time slot, β the fixed packet length (in bits) and N0 is the
noise power in the sub carrier. The parameters P, D, β and
N0 are chosen such that the connectivity Hi,j ∈ {0, 1, 2, 3}.
All simulations are conducted over 6,000 time slots. We
assume arrivals to each user have the Poisson distribution
and are independent from user to user.

7. Results and Comparisons

We compare the performance of our algorithm with the


maximum through- put load balancing (MTLB) and Figure 2. Average queue backlog for MTLB policy with
maximum weighted matching (MWM) algorithm pro-posed SNR = 6 dB
by Kittipiyakul and Javidi in [13] with error in channel
estimation error under different channel model. We show
(IJCNS) International Journal of Computer and Network Security, 57
Vol. 1, No. 3, December 2009

average delay constraint in terms of number of slots. All


users in a group have the same system parameters.
We do two experiments. In experiment 1, we vary the
arrival rate for group 1 and measure the user delay and
throughput, while in experiment 2 we vary the average SNR
of the users in a single group and measure the same
quantities as in experiment 1. For both the experiments, we
compare 4 schedulers, i.e., Opportunistic scheduler,
heuristic, OTODCS scheduler in [12] and OTODCS
scheduler for broadcast channel case. In the OTODCS
scheduler for broadcast channel we assume that at the most
2 users can be scheduled at a time, and also assume equal
power distribution among them. Figure 5shows the variation
of the delay of a user in group 1 against various arrival
rates. Figure 6 shows the variation of the system throughput
Figure 3. Average queue backlog for MTLB policy with against various arrival rates for group 1. In this experiment,
SNR = 9 dB group 1 has poor channel conditions; hence the pure
opportunistic scheduler allocates a small share of bandwidth
to this group. From figure 5 we observe that pure
opportunistic scheduler achieves a very high user delay. The
heuristic scheduler, OTODCS scheduler for the broadcast
case and its TDM counterpart all satisfy the delay
requirements, and we see in Figure 6 that throughput
achieved by all the scheduler except OTODCS for broadcast
are same. However the throughput for OTODCS broadcast
case is higher than the other scheduler. From Figure 5 it is
clear that OTODCS scheduler for the broadcast case do
much better than the others when we compare in terms of
delay.

Figure4.Average queue backlog for MWM with SNR = 0dB

8. Performance and Comparisons


Before we provide the details of simulation parameters let us
first discuss a Heuristic algorithm which is also a delay
satisfying algorithm.
Heuristic Algorithm:
We state here the heuristic algorithm proposed by Salodkar
et.al, in [12]. Let in a time slot n, i be the estimate of i(n),
the average rate at which data can be transmitted to a user i.
We calculate i using exponential averaging as follows,
i (n+1) = α i(n) + (1-α) yi(n),
Where yi (n) is the rate at which data can be transmitted to a Figure 5. Single user delay variation with arrival rate
user in time slot n and α ∈ (0, 1). The heuristic scheduler
chooses a user i in time slot n such that,

We perform the simulation using MATLAB. We simulate


the system for 50 users, then we divide user’s in 10 groups
and each group has 5 users. The arrivals to each users queue
are i.i.d., and we use the Poisson distribution of arrivals for
each user, bounded to 8. The distribution for each user
channel is Rayleigh, The mean of |hi(n)|2, i.e., is the
average SNR of a user (expressed in dB).
We normalize the ratio P/N0 to 1. The rate at which user
can receive data from a base station in a time slot and is
upper bounded by 25. The throughput achieved in a time
slot is expressed in packets/timeslot/Hz. Each user has an
Figure 6. System throughput variation with arrival rate
58 (IJCNS) International Journal of Computer and Network Security,
Vol. 1, No. 3, December 2009

some of the existing ones. We have shown via simulation


that our algorithm performs better than MTLB algorithm
when error in channel estimation is high, and channel is
considered to be On/Off. We then formulated the problem of
maximizing average throughput, with individual delay
constraints, and we considered a broadcast channel at the
downlink. We proposed an online algorithm as a solution to
this problem. We compared our algorithm with some of the
existing algorithms. We have shown via simulation that our
algorithm gives better performance in comparison with its
TDM counterpart in terms of throughput and delay. Further
we extended our problem to two server case. We have shown
that this has the same structure as single server problem. We
have shown via simulation that we gain in performance
when we increase the number of servers to two.
Figure 7. Single sensor node variations with channel
condition
References
[1] R. S. Cheng and S. Verdu, “Gaussian multiaccess
channel with isi: capacity region and multiuser
waterfilling,” IEEE Trans. on Inform. Theory, vol. 39,
no. 3, pp. 773–785, May 1993.
[2] R. V. Nee and R.Prasad, “OFDM for wireless
communications,” Artech House,Boston, 2000.
[3] T. Javidi, “Rate stable resource allocation in OFDM
systems: from waterfilling to queue balancing,”
Allerton conference on communication, control and
computing, September 2004.
[4] G. Li and H. Liu, “Dynamic resource allocation with
finite buffer constraint in broad-band OFDMA
networks,” IEEE Wireless Communication and
Networking, pp. 1037-1042, March 2003.
Figure 8. System throughput variations with channel [5] S. Kittipiyakul and T. Javidi, “Delay-optimal server
condition allocation in multi-queue multi-server systems with
Figure 7 shows the variation of the delay of user in group 10 time varying connectivities,”IEEETrans. on info.
against various channel gain (average SNR). Figure 8 shows theory.
the system throughput against various channel gain (average [6] L. Tassiulas and A. Ephremides, “Dynamic server
SNR) for group 10. In Figure 5.3 it is clear that OTODCS allocation to parallel queues with randomly varying
scheduler for the broadcast case, OTODCS scheduler with connectivity,” IEEE Trans. on Inform. Theory, vol. 39,
TDM and heuristic policy all are able to satisfy delay no. 2, pp. 466–478, May 1993.
requirements, but the delay performance of OTODCS for [7] A. Ganti, E. Modiano, and J. T. Tsitsiklis, “Optimal
broadcast case is the best among all. We can observe form transmission scheduling in symmetric communication
Figure 8 that OTODCS for the broadcast case do better than models with intermittent connectivity,” IEEE Trans.
the other in terms of throughput as well. On Inform. Theory, vol. 53, no. 3, pp. 998–1008,
March 2007.
9. Applications [8] X. Liu, E. Chong, and N. Shroff, “Transmission
The applications for WSNs are many and varied, but scheduling for efficient wireless re-source utilization
typically involve some kind of monitoring, tracking, and with minimum performance guarantees,” Proceeding of
controlling. Specific applications for WSNs include habitat IEEE VTC,vol. 2, pp. 824–828, October 2001.
monitoring, object tracking, nuclear reactor control, fire [9] S. Kulkarni and C. Rosenberg, “Opportunistic
detection, and traffic monitoring. In a typical application, a scheduling policies for wireless systema with short
WSN is scattered in a region where it is meant to collect term fairness constraint,” IEEE GLOBECOMM, vol. 1,
data through its sensor nodes. pp. 533–537, December 2003.
[10] R. Berryand and R. Gallager, “Communication over
10. Conclusion fading channel with delay constraint,” IEEE Trans on
Info. Theory, vol. 48, no. 5, pp. 533–537, May 2002.
The problem of delay optimal server allocation policy [11] E. Yeh and A. Cohen, “Dealy optimal rate allocation in
studied in literature is with full CSI and QSI. We considered multiaccess fading channel,”IEEE Worksop on
a more practical scenario where there is channel gain Multimedia Signal Processing, pp. 404–407, October
estimation error. We proposed a heuristic policy as a 2004.
solution to this problem. We then compared the policy with
(IJCNS) International Journal of Computer and Network Security, 59
Vol. 1, No. 3, December 2009

[12] N. Salodkar, A. karanidikar, and V. S. Borkar, “Delay


constrianed point to multipoint scheduling in wireless
fading channel,” IEEE Journal on selected areas in
communication.
[13] S. Kittipiyakul and T. Javidi, “Resource allocation in
OFDMA with time-varying channel and bursty
arrivals,” IEEE Communication Letters, vol. 11, no. 9,
pp. 708–710, September 2007.
[14] S. M. Kay, Fundamentals of statistical Signal
Processing: Estimation theory. Prentice Hall, 1993.
[15] A. Czylwik, “Adaptive OFDM for wideband radio
channels,” IEEE Globecomm, vol. 1, pp. 713–718,
1996.
[16] D. P. Bertsekas, Nonlinear Programming. MA: Athena
Scientific, 1999.
[17] V. S. Borkar, “Stochastic approximation with two time
scales,” System and Contol Letters, vol. 29, 1996.
[18] J. C. Spall, Introduction to Stochastic Search
Optimization. Wiley, 2003.

Mohd. Sabir received the B.E. degree


(Electronics & Communication) from
MAIET, Jaipur (Raj.) in 2003. And M.E.
degree (Digital Communication) from
M.B.M. Engineering College, Jodhpur
(Raj.) in 2009. He now with Sobhasaria
Engineering College, Sikar, as lecturer in
department of electronics.

Rakesh Kumawat received the B.E.


degree (Electronics & Communication)
from SEC, (Raj.) in 2007. And M.Tech.
degree (VLSI Design) from C-DAC
Mohali in 2009. He now with Sobhasaria
Engineering College, Sikar, as lecturer in
department of electronics.

Vijay S. Chouhan was born in India in


1960. He received B.E. degree in
Electronics & Communication
Engineering and M.E. degree in Digital
Communication Engineering from J. N.
Vyas University, Jodhpur, India. He
submitted his Doctoral thesis on
Biomedical Signal Processing in 2007.
Presently he is working as Professor in
the Department of Electronics &
Communication Engineering, Institute of Engineering &
Technology, Alwar, India. His research interest includes fields of
Biomedical Signal Processing, Soft Computing and Digital
Communications.

Anda mungkin juga menyukai