Content uploaded by Lars Landmark
Author content
All content in this area was uploaded by Lars Landmark on Jun 30, 2015
Content may be subject to copyright.
iOLSR: OLSR for WSNs Using Dynamically
Adaptive Intervals
Erlend Larsen, Joakim Flathagen, Vinh Pham
FFI, Norway
Email: {erl, jfi, vph}@ffi.no
Lars Landmark
Q2S, NTNU, Norway
Email: larslan@q2s.ntnu.no
Abstract—Proactive link state routing protocols, as used within
the Mobile Ad hoc NETwork framework, have not been as
successful in wireless sensor networks. This is mainly due to the
extensive energy usage by control traffic transmissions and state
requirements. However, such protocols are in many situations
a more suitable candidate than their counterparts. The benefits
are their topology overview, and more importantly the already
available spanning trees for information distribution. The high
signaling overhead associated with proactive protocols can be
reduced by taking advantage of the static nature of wireless
sensor networks. In this paper, we investigate how the Optimized
Link-State Routing (OLSR) protocol, as a proactive routing
protocol candidate, can be adapted to work better in a wireless
sensor network environment. The basis for the solution is that
control messages are sent with a low frequency when the network
is stable, and more often if topology changes occur. The proposed
solution is investigated using simulations from no loss to lossy
link environments showing promising results.
Keywords-Ad hoc networks; Routing; Wireless sensor net-
works.
I. INTRODUCTION
Mobile Ad Hoc Networks (MANETs) and Wireless Sensor
Networks (WSNs) are regarded as two distinctly different
types of ad hoc networks, requiring routing protocols with
specialized attributes. While MANET routing protocols are
challenged with mobility, the main limitations for routing in
WSNs are energy and memory. The IETF 6LoWPAN (IPv6
over Low power Wireless Personal Area Networks) working
group [1] have made great efforts to bring IP to WSNs and
other low power wireless networks. Therefore, modifying a
MANET protocol that natively supports IP to better fit the
challenges of WSNs may help the introduction of IP in WSNs.
Radio communication is a major energy consumer in WSNs.
Shwe et al. [2] proposed among other measures to minimize
the number of control packet transmissions to reduce energy
usage. However, limiting the control traffic could make the
routing protocol less able to maintain routes and do route
repair. Thus, our goal is to reduce the number of control packet
transmissions while avoiding these negative consequences.
The traffic pattern types in WSNs are mainly the following,
arranged in probability of occurrence from high to low:
1) Sensors to sink
2) Sink or a specific controller to all or some sensors
3) Sensor to sensor
In the first and third case, the main challenges that have
been addressed are on optimizing energy preservation and
memory usage. Less attention has been given to traffic flows
from a controller to the sensors, for instance for software
updates. The process of software updating would require a
more optimized distribution tree. Hence, a proactive protocol
enabling optimized message distribution and routing would in
many cases be more advantages than a protocol optimized for
sensor to sink.
The Optimized Link State Routing (OLSR) protocol [3] is
a proactive link state MANET routing protocol. It sets up and
maintains routes regardless of application layer communica-
tion demands. The route maintenance is based on the regular
transmission of control traffic. A high number of control
packet transmissions will make the protocol less suitable
for WSNs. At the same time it offers several advantages
that are not as easily available with reactive protocols. For
example, it can provide: quick rerouting in case of topology
changes, spanning trees for information distribution, node
cooperation, and node localization. Sleep functionality may
be more challenging with a proactive protocol, but on the
other hand, a proactive protocol may offer a distribution tree
that may allow more efficient synchronization of sleep state.
WSNs often operates over lossy links. In networks where links
experiencing radio silence, a protocol enabling fast recovery
is a requirement. Reactive protocols tend to increase their path
distance as their have no mechanism to roll back after expiring
radio silence as proactive protocols.
The OLSR protocol is not specifically designed for fixed
topologies. Furthermore, due to its link state properties it also
has a larger state requirement, than other protocols tailored
for WSNs. There are no specified mechanisms to adapt the
emission interval of control messages, depending on the grade
of topology change. This means that the rate of control
messages must be decided before the network deployment,
based on the expected dynamics and the wanted reaction time
to such dynamics. WSNs can be perceived as static and fixed
without any dynamics. Nonetheless dynamics will occur, due
to fluctuating links, new deployed nodes or nodes disappearing
due to energy depletion or other malfunction. In a more
dynamic network, where links or nodes break frequently, the
routing protocol needs to perform control traffic dissemination
more often.
The main contribution of this paper is the adaption of OLSR
to exploit the static nature of WSNs through dynamically
18
SENSORCOMM 2011 : The Fifth International Conference on Sensor Technologies and Applications
Copyright (c) IARIA, 2011. ISBN: 978-1-61208-144-1
increasing the intervals of the control messages. This proposed
solution is named iOLSR. Scaling OLSR used in WSNs has
been criticized for the extensive use of state, which makes
scaling a challenge. However, for medium WSNs, the required
memory for holding OLSR state will likely be below current
memory limitations of many commercial nodes for WSNs. The
proposed solution shows its ability to reduce its control traffic,
and to deal well with in environments with error prone links.
The rest of the paper is structured as follows. Related work
is presented in Section II. The changes proposed to OLSR
are presented in Section III. The solution is investigated and
compared to alternatives in Section IV. Finally, the paper is
concluded along with an outline of further work in Section V.
II. RE LATE D WOR K
Earlier work that address adaptation of OLSR for WSNs
include [4], where Benslimane et al. propose a way to per-
form energy-aware routing using OLSR. Minet and Mahfoudh
presents an energy-aware version of the OLSR routing pro-
tocol in [5]. These two papers focus on routing traffic over
paths that minimize the energy consumed in the end-to-end
transmission of a packet flow and avoiding nodes with low
residual energy, increasing the network lifetime. iOLSR, on
the other hand, does not consider energy levels, but instead
focuses on the reduction of control traffic.
The OLSR standard allows for different nodes having dif-
ferent interval settings, but there are no described options or
methods to vary the intervals while operating. Fast-OLSR [6]
is a proposal to enable the broadcast of Fast-Hello messages
with a shorter interval in case high mobility is detected. It is
thus a proposal to change the control message intervals with
basis in information about the relative mobility of the node,
depending on if there is a high number of changes in the node’s
neighborhood. In addition, Fast-OLSR proposes a Fast-Hello
message with a reduced set of neighbors announced, to reduce
the increased routing traffic overhead. Our proposed solution
does not impose a new network message type and it is tested
in a link burst environment. As with Fast-OLSR, iOLSR is
compatible with the standard OLSR protocol.
An IPv6 routing protocol for Low Power and Lossy Net-
works (RPL) is currently being developed in the IETF. Clausen
and Herberg investigate RPL-Enabled Optimized Broadcast in
[7]. The authors argue that Multi-Point Relay (MPR)-based
efficient broadcast is a well performing mechanism for WSNs,
and the MPR mechanism is essential to the OLSR routing
protocol, upon which we base our proposed solution.
III. PROPOSED SOLUTION
The OLSR protocol uses two different control messages
for its most basic routing functionality, Hello and Topology
Control (TC). The Hello messages are generated by all nodes
and are periodically broadcasted to all 1-hop neighbors. Based
on the information exchanged in Hello messages, a subset of
the nodes in the network are selected as MPR nodes. These
nodes generate Topology Control (TC) messages, which are
flooded throughout the network using the other MPR nodes.
The Hello and TC emission intervals affect the reaction
latency to topology changes, and the intervals can be set
balancing between the energy usage and the topology change
discovery latency. The default lengths of the Hello and TC
intervals are 2 and 5 s, respectively, and the main motivation
for these low values is the ability to cope with high mobility
induced topology change.
There are two clear side effects of increasing the control
packet intervals. The first is the increased latency in detecting
link breaks. To reduce the link break detection time, dis-
appearing nodes causing link failures can be detected using
Link Layer Notification (LLN). The second side effect is the
latency in detecting new nodes. If new nodes are introduced
in the network when the network has been operating for a
while, these nodes will only be employed for routing when
the network has discovered them. However, until the new
node has received TC messages from all elected MPRs in
the network, there is a risk that the node will discard packets
to destinations it is unaware of. And worse, generate loops if
it has a different view of the shortest path than the upstream
node. Large message intervals will delay the discovery and the
use of more optimal paths.
In static WSNs, there are no topology changes caused by
mobility. The topology is stable and static for the most of the
time. After performing the initial discovery of the topology, the
routing protocol could stop disseminating control messages.
However, at any time, a node may disappear or make its
appearance in the network, and links may fluctuate. Links may
even be broken due to external causes, such as targets entering
the the network detection zone. Therefore, even proactive
protocols for WSNs must perform control with the network
links to detect and recover from topology changes. To reduce
the overhead of routing messages, which drains the nodes of
energy, the message intervals can be increased or turned off.
However, this is at the cost of slower detection of topology
changes detected by necessary control packets.
We propose to allow each node to adjust its Hello and
TC intervals depending on the local state of the network.
In the initial startup phase, where each new received Hello
message contains new information, the node keeps the default
low interval between each new originated message. As the
initialization phase draws to an end, and no more changes are
experienced in the neighborhood, the control message intervals
are increased. In this way, the energy usage is reduced while
the topology is stable and unchanged. If a change is detected
in the local neighborhood, the message intervals are reset, and
then incremented anew when no changes are detected (i.e., the
network is perceived as stable again). The topology changes
are detected by using Link Layer Notifications (LLNs) and
through Hello messages containing new information. As a
consequence, the protocol is able to adjust itself to operate
over both stable links and more lossy environments while
optimizing the overhead of routing messages through the
increasing intervals.
In our proposal, the intervals of the control messages must
vary between a lower and an upper limit. If the lower interval
19
SENSORCOMM 2011 : The Fifth International Conference on Sensor Technologies and Applications
Copyright (c) IARIA, 2011. ISBN: 978-1-61208-144-1
is too short, the number of control messages transmissions
will drain the network nodes of energy without improving the
routing protocol’s reaction time. If set very low, collisions of
control messages may even impair the network, causing loss of
data traffic due to loops or lack of routes. In the upper end, the
time fields of the OLSR message header limits the maximum
interval. The time is encoded in the header in a mantissa and
exponent format, each of 4 bits, into one byte. A time value
iis encoded as i=C·(1 + a
16 )·2bwhere ais the highest 4
bits of the field, and bis the lowest 4. The scaling factor Cis
proposed as 1
16 second, giving a time field range of 0.0625 s
– 3968 s. The scaling factor could be increased to achieve a
higher maximum time range, which could be advantageous for
our proposed solution, but this has not been looked into in this
work. In such a case, one would loose the resolution at lower
numbers.
The intervals and the corresponding message timeouts (valid
times) are increased each time the control message is trans-
mitted. Upon experiencing a change in the neighborhood the
intervals are reset to the default values, and the incrementation
process begins over again. The rate that the intervals are
increased by, can be discussed, and the simulation results show
three different takes on this increase.
A generic representation of the calculation of control mes-
sage intervals and their timeouts has been sought for. The
following formula represents the calculation of increasing
the intervals continually, either linearly or exponentially. The
generic interval vcan be calculated as follows:
v=vd·(αi+β·i)(1)
In (1), vis the resulting interval, vdis the starting (default)
interval, αis the base exponential value, βis the linear
increment and iis a counter of successfully transmitted control
messages. Upon a change in the local topology information,
this counter is reset to 0.
The basis for message information timeout has been to
follow the proposal of the OLSR RFC [3], using 3 times the
interval rate as the timeout value. However, with an expected
increasing interval, the timeout vt must be calculated as stated
in (2).
vt =
2
X
k=0
vd·(α(i+k)+β·(i+k)) (2)
In the equations, αand βare constants, set for the entire
duration of the simulation use or network deployment.
IV. SIMULATIONS
A. Setup
The proposed solution was investigated using simulations
on the ns-2 network simulator [8] version 2.34. The OLSR
protocol as described in [3] and implemented for ns-2 in [9]
was used for unicast routing, and the IEEE 802.11 protocol
[10] was used as MAC layer. LLN was enabled in all simu-
lations.
TABLE I
SIMULATION PARAMETER SETTINGS
Radio-propagation model TwoRayGround
Interface queue type FIFO with DropTail,
PriQueue for OLSR packets
Interface queue size 30 packets
Antenna Model OmniAntenna
Data/basic rate 2 Mbps / 1 Mbps
Data transmission/sensing radius 250 m / 550 m
Simulation/measurement time 6000 s / 600–5900 s
Random seed Heuristic
The solution was tested on a scenario with fluctuating links
where the link loss is a consequence of link failures using
an implementation of the Gilbert-Elliot link burst error model
[11], [12]. The link has a certain probability of going into a 0–
3 s burst error period for each received packet, and while in the
burst error period, the link experiences a 100% packet loss.
When not in a burst error period, the two-ray-ground radio
propagation model is employed with a 250 m transmission
radius.
The topology size was set to 40 nodes in a 1500 x 300 m2
area, and the nodes were placed randomly using software from
[13], to allow the examination of a wide network without
the very long simulation processing time that follows using
ns-2 with a high number of simulated nodes. The sink was
randomly positioned. The simulation time was 6000 seconds
unless otherwise stated. All nodes generated packets, except
those nodes that appeared or disappeared, and the packets
were set with the sink as destination, to test the paths toward
the sink. The traffic load was 1 packet per second from each
traffic-generating node. The traffic type was UDP unicast with
a packet size of 50 bytes, and the traffic flows were started at
500 s. All data points are an average of 10 simulation runs, and
are presented with a 95% confidence interval. The topologies
were the same 10 topologies for each of the simulations. Other
simulation parameter settings are presented in Table I.
For our proposed iOLSR solution, the interval incrementa-
tion counters were reset at the following events:
•Hello messaging causing link change or timeout.
•A new MPR selector or a timeout of an existing one.
•Link break causing a LLN.
When the static Hello intervals were increased, the TC
intervals were increased correspondingly, so that for example
a Hello interval of 10 had a TC interval of 25.
B. Results
1) Increasing intervals: First we investigated how the
iOLSR solution compared to the regular static interval be-
havior of OLSR in varying link stability conditions. Three
different static Hello intervals were simulated: 2, 20 and
100 s. The corresponding TC intervals are: 5, 50 and 250 s.
The iOLSR default intervals were 2 s for Hello and 5 s
for TC, and the interval increment was 2-base exponential.
Examining the goodput results (Fig. 1) we see that all the
variations manage to perform well when the topology is stable
without many link errors. When the link error probability
20
SENSORCOMM 2011 : The Fifth International Conference on Sensor Technologies and Applications
Copyright (c) IARIA, 2011. ISBN: 978-1-61208-144-1
10
20
30
40
50
60
70
80
90
100
1e-06 1e-05 0.0001 0.001 0.01
Goodput (%)
Link burst error probability
2s Hello
20s Hello
100s Hello
iOLSR
Fig. 1. Average goodput for iOLSR compared to OLSR with static intervals.
1000
10000
100000
1e+06
1e-06 1e-05 0.0001 0.001 0.01
Transmissions
Link burst error probability
2s Hello
20s Hello
100s Hello
iOLSR
Fig. 2. Number of control packet transmissions for iOLSR compared to
OLSR with static intervals.
increases, a higher interval between the Hello messages makes
the routing protocol less capable of taking advantage of the
links rebounding from a burst error, and this leads to a logical
partitioning that reduces the goodput. Interestingly, we mark
that iOLSR is able to offer the same performance as that of
standard OLSR with 2 s Hello intervals, even at the highest
link burst error probability.
While the goodput performance for all alternatives was
very good at the lower link break probabilities, it is the
number of transmitted control packets that is most interesting
in WSNs, since the number of transmissions directly affect
the energy use of the nodes. The number of control packet
transmissions (Fig. 2) show iOLSR as being highly adaptive
to the environment it operates in. When there is low probability
of burst errors, the number of routing packets is kept at a much
lower level than the comparable 2 s Hello interval results,
and even compared to the 100 s Hello results. As the burst
error probability increases, the routing protocol dynamically
increases the number of Hello and TC messages generated
locally in the area around each failing link.
2) Interval increment rate: The control message intervals’
rate of increase is important when evaluating iOLSR. We have
investigated three increment options where αand βrefer to
(1) and (2):
1) Linear (lin) (α= 1 and β= 1)
TABLE II
CON TROL M ES SAG E INT ERVAL PR OGR ESS IO N
Hello TC
Lin Exp2 Exp3 Lin Exp2 Exp3 Exp3 vt
2 2 2 5 5 5 65
4 4 6 10 10 15 195
6 8 18 15 20 45 585
8 16 54 30 40 135 1755
10 32 162 35 80 405 5265
··· ··· ··· ··· ··· ··· ···
65
70
75
80
85
90
95
100
1e-06 1e-05 0.0001 0.001 0.01
Goodput (%)
Link burst error probability
iOLSR lin iOLSR exp iOLSR exp 3
Fig. 3. Average goodput for linear contra exponential increase of iOLSR
intervals.
2) 2-base exponential (exp2) (α= 2 and β= 0)
3) 3-base exponential (exp3) (α= 3 and β= 0)
As we see in Table II, the linear option will increase the
interval by the default value for each successful transmission,
while the 2-base and 3-base exponential options increment the
message intervals exponentially according to (1).
The simulation results with varying control message interval
incrementation rate shows that the goodput (Fig. 3) is not
affected adversely by choosing a 2-base exponential increase
of the intervals, even in an environment with a high probability
of link burst errors. It follows the linear increment results
very closely. The 3-base exponential increase is more prone
to errors.
100
1000
10000
100000
1e+06
1e-06 1e-05 0.0001 0.001 0.01
Transmissions
Link burst error probability
iOLSR lin iOLSR exp iOLSR exp 3
Fig. 4. Number of control packet transmissions for linear contra exponential
increase of iOLSR intervals.
21
SENSORCOMM 2011 : The Fifth International Conference on Sensor Technologies and Applications
Copyright (c) IARIA, 2011. ISBN: 978-1-61208-144-1
0
500
1000
1500
2000
2500
3000
3500
0 1000 2000 3000 4000 5000 6000
Transmissions/500s
Simulation time (s)
iOLSR lin iOLSR exp2 iOLSR exp3
Fig. 5. Number of control packet transmissions for linear contra exponential
increase of iOLSR intervals at increasing simulation time, 0–6000 s.
Examining the results for the number of control packet
transmissions (Fig. 4), there is evidently great gain in using an
exponential interval increment compared to a linear increment.
However, the gain is much less in using a larger base exponent
such as 3 compared to 2. The reason is twofold. First, the
number of transmissions required to reach the maximum
time field limit is lower with an increasing increment. After
the incrementation phase, there is no difference between the
increment values, since the intervals are no longer increased.
For the exp3, the maximum interval for TC is reached at the
fifth transmission, since the vt (control message information
timeout) will be 5265 (Table II), thus exceeding the maximum
time (3968). Second, the beginning of the incrementation
phase is the phase where changes are most likely happen,
especially at the initialization of the network. An interval
increment that moves too quickly towards higher control
message intervals may actually harm the initialization and
convergence of the protocol.
Since there is great impact of how long the initialization
phase is, when it comes to the number of transmissions,
we have run some simulations without data traffic, and only
routing traffic present, where the number of control packet
transmissions is examined for the lin, exp2 and exp3 increment
options. For a simulation lasting for 6000 s (Fig. 5), clearly
the lin option is unable to reach the maximum interval. The
two other options, however, reach the maximum interval very
early.
In a longer simulation, 0–300000 s, the maximum time
limit is even more pronounced, examining the control packet
transmissions (Fig. 6). Except for the first measured step, the
exponential increase options operate at the maximum for the
entirety of the simulation.
The conclusion of the investigation into the increment alter-
natives is thus that the 2-base exponential increment represents
a middleway between a too slow move away from the default
intervals toward the maximum, and at the same time a more
slow move away from the default intervals in the first phase
of a stabilizing network.
0
2000
4000
6000
8000
10000
12000
14000
0 50000 100000 150000 200000 250000 300000
Transmissions/10000s
Simulation time (s)
iOLSR lin iOLSR exp2 iOLSR exp3
Fig. 6. Number of control packet transmissions for linear contra exponential
increase of iOLSR intervals at increasing simulation time, 0–300000 s.
60
65
70
75
80
85
90
95
100
1e-06 1e-05 0.0001 0.001 0.01
Goodput (%)
Link burst error probability
iOLSR AODV
Fig. 7. Average goodput for iOLSR contra AODV.
3) Comparison with AODV: The final comparison in this
paper is between the iOLSR (2-base exponential increase) and
Ad hoc On-demand Distance Vector (AODV) [14]. Protocols
for WSNs normally establish routes from sink to sensors
to reduce the control traffic. AODV establishes routes from
sensors to sink. In lossy environments it is likely that sensors
must take part in path or link recovery. Hence, AODV in this
sense resembles many protocols for WSNs in its behavior.
In this work we want to compare the impact of burst error
on iOLSR to a reactive protocol handling path/link recovery,
such as AODV. With AODV, the start of the traffic flows
were spaced up with 1 s intervals, to prevent effects of a
synchronized route setup process. The goodput results (Fig. 7)
show the interesting fact that AODV and iOLSR follow each
other closely.
Examining only the number of control packets transmitted
(Fig. 8) with the goodput results in mind, AODV is clearly
better at low link burst error rates, yielding a much lower
number of control packet transmissions than iOLSR.
However, the control packet results only tell part of the
story. Investigating further, the number of hops for the data
traffic (Fig. 9) is much higher for AODV than for iOLSR. This
is due to the way AODV sets up routes only once, flooding the
network with a route request from each source in the network.
Although the total number of control packets may be low, the
22
SENSORCOMM 2011 : The Fifth International Conference on Sensor Technologies and Applications
Copyright (c) IARIA, 2011. ISBN: 978-1-61208-144-1
10
100
1000
10000
100000
1e+06
1e+07
1e-06 1e-05 0.0001 0.001 0.01
Transmissions
Link burst error probability
iOLSR AODV
Fig. 8. Number of control packet transmissions for iOLSR contra AODV.
2
2.5
3
3.5
4
4.5
5
5.5
1e-06 1e-05 0.0001 0.001 0.01
Hops
Link burst error probability
iOLSR AODV
Fig. 9. Average number of hops for the data traffic for iOLSR contra AODV.
forwarding transmissions of the route requests are at risk of
collisions, resulting in a failure to propagate the shortest path
outwards to the destination. Furthermore, in case of low data
traffic and low error rate, the result indicates the benefit of a
reactive protocol. However, as traffic increases more traffic is
traveling over more hops, thus draining more resources.
The consequence of the higher number of hops for the data
traffic using AODV is the higher total number of transmis-
sions (Fig. 10) where both the data traffic and control traffic
100000
1e+06
1e+07
1e-06 1e-05 0.0001 0.001 0.01
Transmissions
Link burst error probability
iOLSR AODV
Fig. 10. Total number of transmissions (control and data traffic) for iOLSR
contra AODV.
transmissions are counted.
V. CONCLUSIONS AND FUTURE WORK
This paper has presented an adaptation of OLSR for WSNs
by introducing dynamically adaptive intervals. The advan-
tages of employing dynamic intervals for control packets
were demonstrated. We achieved less control packet overhead
than by using the default control packet interval. Also, we
demonstrated a faster detection and integration of new nodes
than by using a large control packet interval.
The solution induces costs in terms of less route main-
tenance. Even so, the proposed solution represents a much
better alternative for reducing the number of transmissions
than that of preset large intervals, since it will depend on the
real dynamics of the network whether the routing protocol
transmits many or few packets. Last, but not least, using a
proactive protocol provide the ability for a more optimized
traffic pattern from sink to sensors.
Next, we will elaborate on the benefit of turning off the TC
functionality to further reduce the control traffic. Destinations
located further than two hops away would be searched for by
a request using MPR for request forwarding.
REFERENCES
[1] IPv6 over Low power WPAN working group. (2010, May) IETF.
(Last accessed 2011-05-18). [Online]. Available: http://datatracker.ietf.
org/wg/6lowpan/charter/
[2] H. Y. Shwe, X.-H. Jiang, and S. Horiguchi, “Energy saving in wireless
sensor networks,” Journal of Communication and Computer, vol. 6,
no. 5, pp. 20–28, May 2009.
[3] T. Clausen, P. Jacquet (editors), C. Adjih, A. Laouiti, P. Minet, P. Muh-
lethaler, A. Qayyum, and L.Viennot, “Rfc 3626: Optimized link state
routing protocol (OLSR),” RFC 3626, pages 1–75, pp. 1–75, October
2003, network Working Group.
[4] A. Benslimane, R. El Khoury, R. El Azouzi, and S. Pierre, “Energy
power-aware routing in OLSR protocol,” in Mobile Computing and
Wireless Communication International Conference, 2006. MCWC 2006.
Proceedings of the First, 17-20 2006, pp. 14 –19.
[5] S. Mahfoudh and P. Minet, “EOLSR: an energy efficient routing protocol
in wireless ad hoc and sensor networks,” Journal of Interconnection
Networks, vol. 9, no. 4, pp. 389–408, 2008.
[6] M. Benzaid, P. Minet, and K. Al Agha, “Analysis and simulation of Fast-
OLSR,” The 57th IEEE Semiannual Vehicular Technology Conference
(VTC), vol. 3, pp. 1788–1792, April 2003.
[7] T. Clausen and U. Herberg, “Comparative study of rpl-enabled
optimized broadcast in wireless sensor networks,” INRIA, Research
Report RR-7296, May 2010, (Last accessed 2011-05-18). [Online].
Available: hal.inria.fr/inria-00488030/PDF/RR- 7296.pdf
[8] J. Heidemann and T. Henderson (Editors). (2009, October) Network
Simulator 2. (Last accessed 2011-05-18). [Online]. Available: http:
//nsnam.isi.edu/nsnam/
[9] F. J. Ros and P. M. Ruiz. MANET Simulation and Implementation at
the University of Murcia (MASIMUM). (Last accessed 2011-05-18).
[Online]. Available: http://masimum.dif.um.es/
[10] IEEE, “Wireless LAN medium access control (MAC) and physical layer
(PH Y) specification,” IEEE standard 802.11, June 1999.
[11] E. N. Gilbert, “Capacity of a burst-noise channel,” Bell System Technical
Journal, vol. 39, pp. 1253–1265, 1960.
[12] E. O. Elliot, “Estimates of error rates for codes on burst-noise channels,”
Bell System Technical Journal, vol. 42, pp. 1977–1997, 1963.
[13] S. PalChaudhuri. Ns-2 code for random trip mobility model. (Last
accessed 2011-05-18). [Online]. Available: http://monarch.cs.rice.edu/
∼santa/research/mobility
[14] C. Perkins and E. Belding-Royer, “Ad-hoc On-Demand Distance Vector
Routing,” in Proceedings of the 2nd IEEE Workshop on Mobile Comput-
ing Systems and Applications, February 1999, pp. 90–100, New Orleans,
LA.
23
SENSORCOMM 2011 : The Fifth International Conference on Sensor Technologies and Applications
Copyright (c) IARIA, 2011. ISBN: 978-1-61208-144-1
