Professional Documents
Culture Documents
AB
(x) = max(
A
,
B
)+(1)
1
2
(
A
+
B
) (1)
AB
(x) = min(
A
,
B
) +(1)
1
2
(
A
+
B
) (2)
where
A
and
B
denote the fuzzy memberships in fuzzy
sets A and B respectively, whereas is a parameter in the
range [0, 1] that controls the degree to which OWA operator
resembles a pure or or a pure and, respectively.
4. The System Model
We present a brief description of our WSN system model,
and also formulate the online maximum lifetime routing
problem. We consider a static WSN deployment, and model
it as a directed graph G(V, E), where V is the set of nodes
and E is the set of edges. All the nodes have equal initial
energy . The node batteries are neither replaceable nor
remotely rechargeable. Each node v
i
V has a set of
neighbor nodes (denoted as neighborhood set N
i
) that v
i
can
each by a single hop transmission using a certain maximum
transmission radius r
t
. An edge e(v
i
, v
j
) between the two
nodes is dened to exist only if v
i
and v
j
are within each
others radio transmission range, i.e., if d
ij
r
t
where d
ij
is the Euclidian distance between the two nodes.
The energy consumption model used in our simulations
is based on the rst order radio propagation model used
by many related works [15], [25], [26]. According to
this model, the energy expended by a sensor node during
transmission and reception of a k-bit packet is given by
Equations (3) and (4), respectively.
TX
ij
= (A+B d
m
ij
) k (3)
RX
ij
= A k (4)
where A is distance-independent and accounts for the
energy consumed in running transmitter or receiver circuitry,
B denotes the energy required by the transmitters amplier,
whereas m is a eld constant typically in the range [2,4] and
depends on certain characteristics of the wireless medium.
We assume a point-to-point communication model for
our WSN scenarios where we have a set of source nodes
performing sensing task as well as a set of sink nodes
(base stations) that receive data from the source nodes. At
any time, a source node v
m
may initiate a routing request
r
h
(v
m
, v
n
), h = {1, 2, }, for sending its sensed data
to a sink node v
n
. A routing request does not imply a
single data packet, rather it represents a sequence of data
packets to be sent from the source node to a sink node. We
assume there are numerous routing requests {r
1
, r
2
, r
3
, . . . }
during the lifetime of WSN. The goal of the proposed online
routing algorithms is to efciently route each routing request
r
h
, without knowledge of future routing requests r
q
(where
q > h), in such a manner that maximizes the number of
successful routing requests before the end of WSN lifetime.
We use a simple, but commonly used, WSN lifetime
denition: The WSN lifetime is equal to the minimum of
the lifetime values of all the node in the network, i.e., the
network lifetime ends as soon as any node in WSN runs
out of its battery [15][18]. If the lifetime of a WSN node
is denoted by T
vi
, the WSN lifetime may be expressed as
given by Equation (5).
T = min
i
{T
vi
} v
i
V (5)
5. The Fuzzy Multiobjective Routing Algo-
rithm for Maximum Lifetime and Minimum
Delay (FMOLD)
The proposed FMOLD algorithm makes use of the fuzzy
membership functions to compute the cost of each of the
125
routing objectives, namely the system lifetime and the
source-to-sink delay. In addition, a multiobjective cost func-
tion is used to aggregate the costs of the two objectives
into a scalar value. The details of the fuzzy functions are
given in the following sections, but for the sake of a better
understanding, a brief description of fuzzy membership
function design ow is provided below.
A fuzzy linguistic variable and a corresponding fuzzy set
are dened. The value of a linguistic variable is expressed
in words in a natural or an articial language. The fuzzy
membership function is used to map a given value of the
linguistic variable to its fuzzy membership value in the cor-
responding fuzzy set [7], [8]. Unlike crisp sets, in context of
fuzzy logic, a certain element may have a partial membership
(in range [0,1]) in a fuzzy set. A higher membership usually
reects a higher goodness level. A membership function may
be designed in a exible way to assign a certain membership
corresponding to a given value of the linguistic variable
in order to shape a desired behavior of an optimization
objective.
5.1. The Fuzzy Membership Function for Maxi-
mum Lifetime
The fuzzy membership function for the maximum lifetime
objective (depicted in Figure (1)) is the same that was used
in our earlier work, i.e., the FML routing scheme (interested
readers are referred to [5] for a more detailed description).
TX
.
re(
l
t
i
j
Figure 1. A depiction of the fuzzy membership function
for lifetime.
An expression for the fuzzy lifetime membership function
can be derived using the equation of a line, and is given by
Equation (6).
ij
lt
=
1 (
1
1
) (1
re(vi)
) :
if < re(v
i
)
TXij
(re(v
i
) TX
ij
) :
if TX
ij
< re(v
i
)
0 : if re(v
i
) TX
ij
(6)
Here re(v
i
) = ce(v
i
)TX
ij
, and ce(v
i
) denote the resid-
ual and current energy values of the node v
i
, respectively,
whereas [0, 1], [0, 1] are two algorithmic
parameters.
5.2. The Fuzzy Membership Function for Mini-
mum Delay
In order to incorporate the minimum delay objective in our
routing algorithm, a linguistic variable partial delay pD(v
i
)
up till the node v
i
(partial delay pD(v
i
) is dened as the
length of the partially constructed path from the source node
v
m
to the node v
i
), and a corresponding fuzzy set short delay
are dened. A fuzzy membership function (Figure (2)) is
designed to map a value of the variable partial delay to its
corresponding fuzzy minimum delay membership
ij
md
.
(D)(D)
pD (v
)
1
m
d
i
j
Figure 2. A depiction of the fuzzy membership function
for minimum delay.
As may be seen, the function assigns the lowest (highest)
membership value to an edge whose starting node has the
longest (shortest) partial delay among all the nodes at the
partial path search front (i.e., the set of all the leaf nodes of
the Dijkstras path search tree). This behavior of membership
function encourages the selection of such edges that will
result is a shorter path length. The lowest membership may
be altered by adjusting the value of algorithmic parameter
. An expression for the fuzzy minimum delay membership
function is given by Equation (7).
ij
md
= 1 +
( 1) pD(v
i
)
max(pD)
(7)
where
max(pD) = max
f
{pD(v
f
)} f s.t. v
f
(8)
min(pD) = min
f
{pD(v
f
)} f s.t. v
f
(9)
126
5.3. The Fuzzy Multiobjective Membership Func-
tion
In order to formulate a fuzzy multiobjective aggregation
function for an edge, the following fuzzy rule is proposed:
IF an edge
has start node with high lifetime AND
has short partial path delay
THEN it is a good edge.
The above fuzzy rule translates to the following and-like
function by employing the OWA operator [24]:
ij
= min(
ij
lt
,
ij
md
) + (1 ) (
ij
lt
+
ij
md
2
) (10)
where
ij
is the fuzzy multiobjective membership of
the edge e(v
i
, v
j
) and [0, 1] is a constant. As can
be observed, the above OWA function due to the term
min(
ij
lt
,
ij
md
), asserts a preference on the objective
having the least membership value. Also it may be noted,
that due to nature of our carefully designed membership
functions, the minimum value of
ij
lt
is 0, whereas that of
ij
md
is > 0. Therefore it is easy to infer, that if the
residual energy (and the corresponding lifetime membership
ij
lt
) is high, a similar preference level is given to both the
objectives; otherwise if
ij
lt
< , the preference shifts to the
maximum lifetime objective. As a conclusion, the value of
parameter affects the relative preference of the two routing
objectives.
C J, E is the given directed graph
For each routing request r
h
(v
m
, v
n
)
For each edge e(v
i
v
j
in J
Compute Iuzzy liIetime membership value
lt
ij
Compute Iuzzy short delay membership
md
ij
Compute multiobjective membership
ij
Assign weight w
ij
ij
End For
Find minimum weight path p
h
Irom v
m
to v
n
Send data along path p
h
Compute the minimum node energy in C J, E
IF a node has run out oI energy, stop.
End For
Figure 3. The fuzzy multiobjective routing algorithm
(FMOLD) for simultaneous optimization of the lifetime
and the delay objectives.
The proposed FMOLD algorithm (Figure (3)) nds the
multiobjective path as follows: When a routing request
r
h
(v
m
, v
n
) is initiated, the fuzzy lifetime and minimum
delay membership values for each edge are computed using
Equations (6)and (7), respectively. Then the fuzzy multiob-
jective membership value is computed using Equation (10),
and a weight is assigned to each edge using Equation (11).
w
ij
= 1
ij
(11)
Following the weight assignment, the multiobjective path
p
h
between v
m
and v
n
is found using the Dijkstras shortest
path algorithm [27].
6. Performance Evaluation and Discussions
In this section, we compare the results of the FMOLD
scheme to those of the FML scheme with a view of study-
ing the effect of incorporating the minimum source-to-sink
delay objective on the lifetime maximization process. The
performance metrics used for comparison are the network
lifetime and the average source-to-sink delay.
The simulation setup consists of a 2-dimensional grid of
size 25 25, populated with randomly deployed n sensor
nodes. In our simulations, sink node(s) have innite energy
(xed power source) and predetermined location(s), whereas
all the other nodes (including the source nodes) have an
equal level of initial residual energy equal to . We assume
a perfect MAC layer with no energy losses due to retransmis-
sion attempts (it may appear to be an optimistic assumption
considering contention-based wireless MAC protocols, but
some recently proposed MAC protocols are able to offer
deterministic channel access guarantees [28], [29]).
There are number of parameters to which, we need
to assign suitable values. The best values for the FML
parameters and , respectively, are 0.2 and 0.9 as was
determined in [5], whereas the best value for the parameter
is determined empirically. For this purpose, a series
of experiments were conducted, and it was observed that
= 0.2 has the best effect on the maximum lifetime as well
as the minimum delay objective. Hence we set its value at
0.2 in all the subsequent runs. As far as the parameter is
concerned, it was noted that its value has a critical effect
on the values of both the objectives. Therefore we decided
to vary its values in the range [0.2, 0.8] in order to obtain
a desirable tradeoff between the two objectives. A list of
all the simulation parameters and their values is given in
Table 1.
We conduct experiments to study the effect of in a
scenario of varying node density. For a fair and reliable
comparison, in case of both the approaches, every result
shown is averaged over a 100 runs - 10 network topologies
were used, and 10 random request sequences were generated
for each topology. Figures (4) (a) and (b) show the lifetime
and the average path delay (in terms of the ratio of the length
of found path to the shortest path), respectively, obtained by
the FML and the FMOLD for a varying node density n. In
127
0
2000
4000
6000
8000
10000
12000
14000
30 40 50 60 70 80 90 100
Number of nodes ()
L
i
f
e
t
i
m
e
FML FMOLD - 0.2
FMOLD - 0.4 FMOLD - 0.8
0
0.5
1
1.5
2
2.5
3
3.5
4
30 40 50 60 70 80 90 100
Number of nodes ( )
R
a
t
i
o
t
o
t
h
e
s
h
o
r
t
e
s
t
p
a
t
h
FML FMOLD - 0.2
FMOLD - 0.4 FMOLD - 0.8
(a) (b)
Figure 4. A comparison of FMOLD and FML in terms of (a) the lifetime and (b) the average source-to-sink delay
for a varying node density (n). The FMOLD is not only successful in improving on the path delay metric, but is also
able to offer a exible tradeoff control between the two routing objectives through a careful choice of a value for the
algorithmic parameter .
Table 1. A list of values used for various simulation
parameters.
1 J
A 100 nJ/bit
B 50 pJ/bit/m
4
m 4
0.2
0.9
0.2
[0.2, 0.8]
n {30, 40, , 100}
case of the FMOLD, three curves are plotted corresponding
to three different values of the parameter {0.2, 0.4, 0.8}
(For instance, FMOLD - 0.2 denotes the results obtained
by the FMOLD algorithm for = 0.2).
As expected, the FML algorithm obtains the highest life-
time for all the values of n since it is optimized solely for the
lifetime objective. But on the other hand, the FML performs
worst in terms of the path delay objective, as the paths found
are on the average more than two folds longer compared to
the shortest possible path. The interesting thing to notice
is that the proposed minimum-delay-driven multiobjective
scheme FMOLD is successful in considerably reducing the
delay values, though the improvement margin is different
depending on the chosen value of the parameter . The value
of has a critical effect on the simultaneous multiobjective
routing process - a lower value of yields shorter path delay
values, but also casts an adverse effect on the lifetime values:
For instance, the least values for the path delay is obtained
when = 0.2, but this also results in the least lifetime
values.
From the above results, it is easy to infer that a lifetime-
delay tradeoff is clearly visible in the FMOLD results. This
tradeoff is a powerful feature of the multiobjective FMOLD
algorithm since it offers a exible control over choosing a
desired balance between the two routing objectives.
7. Conclusions
A fuzzy multiobjective algorithm for online routing in
WSN is presented. Simulation results are presented to show
that the proposed FMOLD algorithm is able to achieve a
desirable tradeoff between the maximum lifetime and the
short end-to-end delay objectives. Thus the multiobjective
routing scheme may offer an effective maximum lifetime
routing solution in the context of delay-sensitive sensing
WSN applications.
However, not all possible network scenarios of interest
were simulated in this work. For instance, we did not study
the effect of deploying multiple sink nodes. Also we did
not experiment with various geographic placements of the
sink node(s). Moreover, the proposed routing algorithm is
centralized in its path search mechanism, and we did not
make an effort to compute the control trafc overhead.
In the future, we plan to extend our multiobjective ap-
proach by considering a multipath search mechanism. In
addition, for better scalability and lesser control overhead,
a fully distributed implementation of the FMOLD scheme
needs to be investigated.
References
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci,
Wireless Sensor Networks: A Survey, Elsevier Computer
Networks, vol. 38, pp. 393422, March 2001.
[2] D. Culler, D. Estrin, and M. Srivastava, Overview of Sensor
Networks, IEEE Computer, vol. 37, no. 8, pp. 4149, August
2004.
128
[3] J. N. Al-Karaki and A. E. Kamal, Routing Techniques
in Wireless Sensor Networks: A Survey, IEEE Wireless
Communications, vol. 11, no. 6, pp. 628, December 2004.
[4] M. Perillo and W. Heinzelman, Fundamental Algorithms and
Protocols for Wireless and Mobile Networks. CRC Hall,
2005, ch. Wireless Sensor Network Protocols.
[5] M. R. Minhas, S. Gopalakrishnan, and V. C. M. Le-
ung, Fuzzy Algorithms for Maximum Lifetime Routing
in Wireless Sensor Networks, in Proceedings of IEEE
Global Telecommunications Conference (Globecom), Decem-
ber 2008, pp. 16.
[6] , Online Routing Schemes for Maximizing
Lifetime of Wireless Sensor Networks, Department
of Elec. and Comp. Enigeering, University of British
Columbia, Tech. Rep., December 2008, available at
http://www.ece.ubc.ca/mrminhas/publications/tech-
report.pdf.
[7] H. J. Zimmerman, Fuzzy Set Theory and Its Applications,
3rd ed. Kluwer Academic Publishers, 1996.
[8] R. R. Yager, Fuzzy Sets and Approximate Reasoning in De-
cision and Control, in Proceedings of the IEEE International
Conference on Fuzzy Systems, 1992, pp. 415427.
[9] J. Park and S. Sahni, An Online Heuristic for Maximum
Lifetime Routing in Wireless Sensor Networks, IEEE Trans-
actions on Computers, vol. 55, pp. 10481056, August 2006.
[10] K. Kar, M. Kodialam, T. Lakshman, and L. Tassiulas,
Routing for Network Capacity Maximization in Energy-
Constrained Ad-Hoc Networks, in Proceedings of the IEEE
Infocom, 2003.
[11] Q. Li, J. Aslam, and D. Rus, Online Power-Aware Rout-
ing in Wireless Adhoc Networks, in Proceedings of IEEE
MobiComm, July 2001.
[12] A. Misra and S. Banerjee, MRPC: Maximizing Network
Lifetime for Reliable Routing in Wireless Environments,
in Proceedings of the IEEE Wireless Communication and
Networking Conference (WCNC), March 2002.
[13] J. Aslam, Q. Li, and D. Rus, Three Power-Aware Routing
Algorithms for Sensor Network, Wireless Communication
and Mobile Computing, vol. 3, pp. 187208, March 2003.
[14] S. Kim, S. Park, S. Park, and S. Kim, Energy Efcient On-
line Routing Algorithm for QoS-Sensitive Sensor Networks,
IEICE Transactions on Communications, vol. E91-B, no. 7,
pp. 24012404, July 2008.
[15] J. Chang and L. Tassiulas, Maximum Lifetime Routing
in Wireless Sensor Networks, IEEE/ACM Transactions on
Networking, vol. 12, no. 4, pp. 609619, August 2004.
[16] R. Madan and S. Lall, Distributed Algorithms for Maxi-
mum Lifetime Routing in Wireless Sensor Networks, IEEE
Transactions on Wireless Communications, vol. 5, no. 8, pp.
21852193, August 2006.
[17] V. Shah-Mansouri and V. W. S. Wong, Distributed Maximum
Lifetime Routing in Wireless Sensor Networks Based on
Regularization, in Proceedings of the IEEE Global Telecom-
munications Conference (Globecom), November 2007, pp.
598603.
[18] Y. Cui, Y. Xue, and K. Nahrstedt, A Utility-Based Dis-
tributed Maximum Lifetime Routing Algorithm for Wire-
less Networks, IEEE Transactions on Vehicular Technology,
vol. 55, no. 3, pp. 797805, May 2006.
[19] C. Rahul and J. Rabaey, Energy Aware Routing for Low
Energy Ad-Hoc Sensor Networks, in Proceedings of the
IEEE WCNC, vol. 1, March 2002, pp. 350355.
[20] C. Intanagonwiwat, R. Govindan, and D. Estrin, Directed
Diffusion: A Scalable and Robust Communication Paradigm
for Sensor Networks, in Proceedings of the ACM Mobicom,
August 2000, pp. 5667.
[21] D. Ganesan, R. Govindan, S. Shenker, and D. Estrin, Highly
Resilient, Energy-Efcient Multipath Routing in Wireless
Sensor Networks, ACM SIGMOBILE Mobile Communica-
tion Review, vol. 5, no. 4, pp. 1125, 2001.
[22] Y. Lu and V. W. Wong, An Energy-efcient Multipath
Routing Protocol for Wireless Sensor Networks, Wiley Inter-
national Journal of Communication Systems, vol. 20, no. 7,
pp. 747766, July 2007.
[23] T. Srinivasan, R. Chandrasekar, and V. Vijaykumar, A Fuzzy,
Energy-efcient Scheme for Data Centric Multipath Routing
in Wireless Sensor Networks, in Proceedings of the IFIP
International Conference on Wireless and Optical Communi-
cation Networks, April 2006.
[24] R. R. Yager, On Ordered Weighted Averaging Aggregation
Operators in Multicriteria Decision Making, IEEE Trans-
actions on Systems, MAN, and Cybernetics, vol. 18, no. 1,
January 1988.
[25] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan,
Energy-Efcient Communication Protocol for Wireless Mi-
crosensor Networks, in Proceedings of the 33rd Hawaii
International Conference on Sys. Sci., January 2000.
[26] K. Kalpakis, K. Dasgupta, and P. Namjoshi, Efcient Al-
gorithms for Maximum Lifetime Data Gathering and Aggre-
gation in Wireless Sensor Networks, Computer Networks,
vol. 42, no. 6, pp. 697716, August 2003.
[27] E. W. Dijkstra, A Note on Two Problems in Connexion with
Graphs, Numerische Mathematik, vol. 1, pp. 269271, 1959.
[28] T. L. Crenshaw, S. Hoke, A. Tirumala, and M. Caccamo,
Robust Implicit EDF: A Wireless MAC Protocol for Collab-
orative Real-Time Systems, ACM Transactions on Embedded
Computing Systems, vol. 6, no. 4, pp. 28/128/29, September
2007.
[29] U. Roedig, A. Barroso, and C. J. Sreenan, F-MAC: A
Deterministic Media Access Control Protocol Without Time
Synchronization, in Proceedings of the Third IEEE European
Workshop on Wireless Sensor Networks (EWSN). IEEE
Computer Society Press, February 2006, pp. 276291.
129