Académique Documents
Professionnel Documents
Culture Documents
PROBLEM STATEMENT
110
100
When we deploy an UWSN using DBR routing proto-
ol, the setting of the network layer parameters, i.e.,
90 the holding time and the depth threshold, is helpful to
minimize the energy
onsumption but may be not su-
80
ient. In fa
t, the sele
tion of an optimum transmission
70
range at the physi
al layer may drasti
ally redu
e the
network energy
ost (and hen
e its lifetime) while main-
60 taining a reasonably high pa
ket delivery ratio. Trans-
0 100 200 300 400 500 600 700 800 900 1000
Distance (m) mission range plays a pivotal role in determining the
energy
onsumption and the pa
ket delivery ratio in a
Figure 1: Required transmission power for various dis- UWSN implementing DBR. Let us fo
us on the energy
tan
es
ost dened as the expe
ted energy required to su
ess-
fully send a pa
ket to the sink node. Short transmis-
sion ranges
ause problems in the network
onne
tiv-
eters, namely the and the depth threshold
, holding time ity and hen
e frequently require pa
ket retransmissions
plays a pivotal role for obtaining high performan
e with that
ause a high energy
onsumption. On the other
a low energy
onsumption. Intuitively, the forwarder hand very long transmission ranges require more energy
sele
tion is based on the pa
ket s
heduled sending time per pa
ket and
ause the in
rease of the number of re-
whi
h is de
ided on the basis of
omputation of the hold- dundant transmissions
aused by hidden terminals. In
ing time. The pa
ket holding time is proportional to this work, we seek the optimal value of the transmis-
the depth dieren
e between the sender and the
andi- sion range given a
ertain node density that results in
date forwarder and hen
e it favors the nodes that al- a low energy
onsumptions and maintains a reasonable
low the pa
kets to
over longer distan
es towards the high pa
ket delivery ratio. Moreover, an appropriate
sinks. The depth threshold is used to prevent nodes with
hoi
e of the transmission range redu
es the busy ter-
low depth dieren
e to be
ome
andidate forwarders. minal problem (Zhu et al. 2014) by limiting the burden
During the holding time duration, nodes dis
ard the on more stressed nodes from the network tra
.
enqueued pa
ket upon nding its transmission from a
lower depth neighbor. DBR targets lowest depth neigh- SIMULATION EXPERIMENTS
bor of sender as an optimal pa
ket forwarder whi
h is
also helpful in suppressing transmissions of other eligi- In this se
tion we address the problem of identifying the
ble neighbors of sender node. Thanks to its stateless optimal transmission range of sensor nodes with respe
t
and distributed nature, DBR is
apable of handling the to the energy
ost of the network by resorting to the
routing in UWSNs with high node mobility and main- simulation model introdu
ed in Se
tion . Together with
tains a low resour
e usage (there is no need to store this optimization we also study the pa
ket delivery ratio
routing tables) and easiness of implementation. for the optimal transmission ranges. The pa
ket deliv-
A
ording to (Yan et al. 2008) in DBR the holding time ery ratio is a good measure for observing the impa
t
is obtained as follows: of the busy terminal espe
ially for what
on
erns long
transmission ranges.
2τ
fDBR (d) = ∗ (T − d) ,
δ
Simulation s
enarios and performan
e indi
es
where T is the maximal transmission range of a node,
τ is the maximum propagation delay of one hop, i.e., We study UWSNs with various numbers of nodes de-
τ = T /v0 (where v0 is the sound propagation speed in ployed in a xed spa
e of 500m × 500m × 500m a
ord-
the water), d is the depth dieren
e between the sender ing to a uniform random distribution. The number of
and the re
eiver and δ is a s
aling fa
tor of the holding nodes varies from 100 to 800 and hen
e we re
reate the
times whi
h is
hosen in order to a
hieve the optimal s
enarios that are similar to those that have been pre-
performan
e of the network and to minimize the hidden viously studied for other purposes in (Yan et al. 2008).
terminal problem. The analysis of the impa
t of these The depth-threshold is 1/4 of the maximum transmis-
onguration parameters on the network performan
e sion range, and the mobility pattern is a random walk.
has been done in (Yan et al. 2008). Nevertheless, in For MAC layer, we implement Broad
ast MAC proto-
this paper we fo
us on the impa
t of a
onguration
ol (Mirza et al. 2009) whi
h e
iently supports the
parameter at the physi
al layer, namely the transmission fun
tioning of ooding-based routing proto
ols. The
Parameter Value
1000
Network size 500m ×500m ×500m Energy cost for 800 nodes
f 3kHz
400
sour
e node is pla
ed in the bottom of the network. Figure 2: Energy
ost of the network as a fun
tion of
Multiple on-surfa
e sinks have been deployed and the the transmission range.
sour
e node transmits a single pa
ket after every two
se
onds. Table 1 summarizes the experiment setting.
In order to identify the optimal transmission range, we in
reases with the sharp in
rease in the transmission
ompute the following performan
e indi
es: (i) Energy range thanks to availability of multiple paths between
ost of network dened as the expe
ted energy required sour
e node and the sinks. However, after rea
hing at
to su
essfully deliver a pa
ket measured in Joule per the maximum point, it de
lines due to the redundant
pa
ket, (ii) Pa
ket delivery ratio and (iii) Total number transmissions and problems
aused by the busy termi-
of transmissions of network. nals. Interestingly, the transmission range asso
iated
For ea
h measurement we performed 20 independent with the optimal pa
ket delivery ratio is
oherent with
experiments and build the
onden
e intervals at 95% the value whi
h optimize the energy
ost. It is also worth
whose width is always below 7% of the measured value. of noti
e that as observed in (Zhu et al. 2014) there is
a strong
orrelation between high pa
ket delivery ratio
Impa
t of transmission range on the energy
ost and redu
tion of the busy terminal problem. It is worth-
of network, pa
ket delivery ratio and total num- while of noti
e that the pa
ket delivery ratios de
rease
ber of transmissions after rea
hing the maximum but appear to be
ome more
stable. Also for what
on
erns the optimal pa
ket de-
In this experiment we study the network energy
ost as livery ratio, the experiments suggest that the networks
fun
tion of the transmission range of the sensor. with density of 500 nodes outperform those with higher
Figure 2 shows the results of our experiments, i.e., the densities in
ase of transmission ranges longer than 200
estimates of the energy
ost of the network as a fun
- and this may suggest that nding the optimal densities
tion of the transmission range for networks with 500 to
ould be an interesting problem for future works. Nev-
800 nodes. We observe that for very low transmission ertheless, we should observe that a network with high
ranges the
ost of retransmissions due to broken routes node density tends to be more robust to failures and
be
omes prohibitive from the point of view of the energy hen
e other performan
e indi
es should be analyzed be-
onsumed by the networks, whereas as the transmission fore drawing
on
lusions.
range in
reases we have both to fa
e the problem of Figure 4 shows the total number of transmissions per-
the higher
ost for the transmission of the single pa
ket formed in the network for 200s of simulation time. We
and the explosion of the number of retransmissions due
an observe an initial in
rease of the amount of the
to the hidden terminal problem and the
onsequent in- transmissions due to the in
reased number of eligible for-
reased number of
ollisions. We
an also observe that warders of the sender nodes. However, this value tends
as the density of the nodes in
reases, the
ost for redun- to qui
kly stabilize although it shows an irregular pat-
dant transmissions and the
onsequent
ollisions be
ome tern that probably depends on the average depth of the
dominant in in
reasing the energy
ost of the network forwarding nodes.
even in its optimal working point. For the four
onsid-
ered network densities we have an optimal transmission Optimal transmission range as fun
tion of the
range of approximatively 180 meters. We will see later node density
on that above a
ertain density of nodes the optimal
transmission range tend to stabilize to this value under In order to experimentally study the
onne
tion between
the assumptions of Table 1. the optimal transmission range and the network node
Consulting Figure 3, the pa
ket delivery ratio qui
kly density we have run a large set of simulations for ea
h
given density and identied the optimal value for the
1 energy
ost. This has been done by assuming the
on-
vexity of the fun
tion Ec = f (r), where Ec is the energy
0.9
ost as fun
tion of the transmission radius r . Then we
0.8 have pro
eeded by using a bise
tion method.
Figure 5 shows the optimal transmission range for var-
Packet delivery ratio
0.7
ious numbers of deployed nodes. We observe that for
0.6
networks with a number of nodes higher than 500 the
0.5 optimal transmission range stabilizes at approximatively
0.4
180 meters. As observed in Se
tion , this value optimizes
both the network energy
ost and its pa
ket delivery ra-
0.3 Packet delivery ratio for 800 nodes
Packet delivery ratio for 700 nodes tio. As the number of deployed nodes de
reases, the
0.2
Packet delivery ratio for 600 nodes
Packet delivery ratio for 500 nodes optimal transmission range in
reases to 240 meters as-
50 100 150 200 250 so
iated with 100 nodes as number of intermediate for-
Transmission range (meters) warders de
reases
ausing the de
rease in total energy
onsumption of network.
Figure 3: Pa
ket delivery ratio with dierent node den- A
ording to our experiments if ρ is the network node
sities. density expressed in expe
ted number of nodes for km3 ,
we
an say that the optimal transmission range r ∗ for
DBR de
reases with higher ρ as:
14000 r ∗ ∝ ρ1/6 .
12000 In Figure 5 we plot the fun
tion 745/ρ1/6 and we
an see
that it provides a good approximation of the estimates
Number of total transmissions
240
essors provides an a
urate modelling of the modem
operational modes, the
ross-layer intera
tions required
220
by this proto
ol and the busy terminal problem. The
simulator
an be downloaded at the o
ial repository
200 of AquaSim-NG (Martin 2016).
Spe
i
ally, we have addressed the problem of determin-
180 ing the optimal transmission range providing the lowest
energy
ost given the network density. To this aim we
160 rst studied the behavior of the energy
ost as a fun
tion
of the transmission range for networks with given node
densities and empiri
ally veried that this optimal value
0 100 200 300 400 500 600 700 800 900
Number of nodes
exists. Then, we have looked for this optimum value for
Figure 5: Optimal transmission range for dierent node dierent node densities. We observed that, a
ording
densities with minimum energy
ost. to our experiments, the transmission ranges that mini-
mize the energy
osts are also those that maximize the
pa
ket delivery ratio. Finally, we studied the relation Martin R.; Zhu Y.; Pu L.; Dou F.; Peng Z.; Cui J.H.;
between the network density and the optimal transmis- and Rajasekaran S., 2015. Aqua-Sim Next Genera-
sion range. As expe
ted, we found that sparse networks tion: A NS-3 Based Simulator for Underwater Sen-
require higher optimal transmission ranges, but that this sor Networks. In Pro
eedings of the 10th Interna-
values tends to de
rease slowly with denser networks. tional Conferen
e on Underwater Networks & Sys-
From the experiments that we run, we observed that the tems. ACM, 1822.
optimal transmission range de
reases as 1/ρ1/6 where ρ
is the expe
ted number of nodes for km3 . Future works Mirza D.; Lu F.; and S
hurgers C., 2009. E
ient
in
lude the development of an analyti
al model to vali- broad
ast MAC for underwater networks. Pro
edings
date this empiri
al law. of ACM WUWNet, Berkeley, CA, USA.
We believe that the out
omes of this work,
ombined Pompili D. and Akyildiz I.F., 2009. Overview of net-
with the previously developed optimizations at the net- working proto
ols for underwater wireless
ommuni-
work layer studied in (Yan et al. 2008),
an be helpful in
the optimization of the power
onsumption in UWSNs
ations. IEEE Communi
ations Magazine, 47, no. 1,
97102.
adopting DBR routing proto
ol.
Optimizing the trans-
Porto A. and Stojanovi
M., 2007.
REFERENCES mission range in an underwater a
ousti
network. In
OCEANS 2007. IEEE, 15.
Brekhovskikh L.M. and Lysanov, 2003. Fundamentals Uri
k R.J., 1983. Prin
iples of underwater sound.
of o
ean a
ousti
s
. Springer.
M
Graw-Hill.
Carneiro G., 2010. NS-3: Network simulator 3. In UTM
Lab Meeting April. vol. 20. Xie P.; Zhou Z.; Peng Z.; Yan H.; Hu T.; Cui J.H.; Shi
Z.; Fei Y.; and Zhou S., 2009. Aqua-sim: an NS-2
Coutinho R.W.; Bouker
he A.; Vieira L.F.; and based simulator for underwater sensor networks . In
Loureiro A.A., 2016.Design guidelines for opportunis- Pro
eedings of the IEEE O
eans 2009 . IEEE, 17.
ti
routing in underwater networks IEEE Communi-
.
Yan H.; Shi Z.J.; and Cui J.H., 2008. DBR: Depth-
ations Magazine , 54, no. 2, 4048.
Based Routing for underwater sensor networks. In
Domingo M.C. and Prior R., 2008. Energy analysis of Adho
and Sensor Networks, Springer. 7286.
routing proto
ols for underwater wireless sensor net- Zhu Y.; Cui J.H.; Peng Z.; and Zhou Z., 2014. Busy Ter-
works. Computer
ommuni
ations, 31, no. 6, 1227 minal Problem and Impli
ations for MAC Proto
ols in
1238. Underwater A
ousti
Networks. In Pro
eedings of the
Gao M.; Foh C.H.; and Cai J., 2012. On the sele
tion International Conferen
e on Underwater Networks &
of transmission range in underwater a
ousti
sensor Systems. ACM, 111.
networks. Sensors, 12, no. 4, 47154729. Zorzi M. and Pupolin S., 1995. Optimum transmission
Harris III A.F. and Zorzi M., 2007. Modeling the under-
ranges in multihop pa
ket radio networks in the pres-
water a
ousti
hannel in ns2. In Pro
eedings of the en
e of fading. IEEE Transa
tions on Communi
a-
2nd international
onferen
e on Performan
e evalua- tions, 43, no. 7, 22012205.
tion methodologies and tools. ICST, 1826.
Jafri M., 2017. AquaSim Next Generation : Li-
braries, DBR implementation by Mohsin Jafri.
https://github.
om/rmartin5/aqua-sim-ng/
blob/master/model/aqua-sim-routing-ddbr.
.
A
essed: 2017-04-04.
Kim D.; Cho Y.M.; Kim C.; Kim S.; Park S.H.; and
Kang T.W., 2007. E-ITRC Proto
ol with Long & Ad-
justable Range on Underwater A
ousti
Sensor Net-
work. In Advan
ed Information Networking and Ap-
pli
ations Workshops, 2007, AINAW'07. 21st Inter-
national Conferen
e on. IEEE, vol. 2, 665672.
Martin R., 2016. AquaSim Next Generation : Libraries,
by Robert Martin. https://github.
om/rmartin5/
aqua-sim-ng. A
essed: 2016-12-05.