Anda di halaman 1dari 3

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the IEEE INFOCOM 2008 proceedings.

A Localization and Routing Framework for Mobile


Underwater Sensor Networks
Melike Erol, Sema Oktug
Istanbul Technical University, Computer Engineering Department, Istanbul, Turkey
melike.erol@itu.edu.tr, oktug@itu.edu.tr

AbstractIn this paper, we propose a framework to establish


localization and routing in mobile underwater sensor networks.
Localization and routing are done at two consecutive rounds.
Localization messages include localization specific data and
additional fields which are used in the routing decision. The
proposed catch up or pass forwarding algorithm is a greedy
geo-forwarding scheme. It benefits from the uncontrolled motion
of the underwater nodes. The underwater nodes use position and
velocity information to decide whether to carry the packet until
they catch up with a sink or pass it to a faster or slower
relay.

I. I NTRODUCTION
Underwater Sensor Networks (USN) are emerging as popular tools in ocean observations. Currently, there are few
testbeds. The wired ocean bottom network, NEPTUNE [1]
and the wireless acoustic Seaweb [2] are two examples. As
USNs are in the early development stage, most of the real life
experiments are concerned with the physical layer issues. On
the other hand, research is widely based on simulation studies
for synchronization, localization, medium access and routing
protocols.
In this paper, we focus on two fundamental problems:
localization and routing. Sensed data are meaningful with time
and location information, e.i, localization is essential for data
tagging. Sending the data efficiently to a sink is another major
issue in sensor networks. In USNs, more than one sink may
be employed and the underwater nodes may act as relays for
routing the data from source to sink. In this work, we combine
these two significant tasks in a framework where they mutually
help each other. We use a limited number of Mobile Beacon
and Sink (MBS) nodes. MBSs are able to move vertically in
the water column. The rest of the nodes stay under the water
and drift with the currents. MBSs periodically descent to the
same depth with the underwater sensors to localize them and
collect their data. Then, MBSs surface to receive coordinates
from the GPS and upload the data that has been collected by
the underwater sensors to a ground station. MBSs can ascent
and descent with modifying density. Real life implementation
of such floats are used in Argo project [3]. A similar technique
can be applied to MBS nodes to achieve vertical movement.
Localization is done iteratively. Initially MBS nodes are
localized via GPS. They periodically announce their coordinates while diving to the deepest level of the underwater
network. Upon hearing from several beacons (at least three
beacons for 2D and four beacons for 3D) a node is localized.

A localized underwater node becomes an active node and


helps in localization. It acts as a beacon and distributes self
coordinates. This localization phase has a fixed duration which
is also announced in the localization message. It can be
calibrated according to the depth of the underwater network
and the speed of the MBS. Alternatively, after each dive, the
duration of the interval may be uploaded via satellite to the
MBS nodes.
Routing starts after the localization round. Sensors that have
data to send, pick an MBS and forward their data towards that
sink. The proposed routing scheme benefits from the messages
sent in the localization phase. The location and the velocity
of the MBS node and the neighbours are learnt during the
localization phase. The routing algorithm chooses the best
relay according to the position and relative motion of the MBS
and the sensor nodes.
II. N ETWORK A RCHITECTURE AND THE M OBILITY
M ODEL
We design our framework for a mobile underwater sensor
network with acoustic communications. Underwater nodes
stay at a certain depth and drift with the currents. MBS nodes
ascent and descent to provide location service and to deliver
data from the sensors to the ground station via satellite or
radio links. The example network is illustrated in Fig. 1.
Most of the existing mobility models for ad hoc networks
assume that the sensors move independently from each other
[4]. However, in a fluid medium strong correlations near
the same velocity field are expected. In the ocean, mobility
is determined by advection from the ocean currents. The
subsurface current is a jet-like current, meandering between
recirculating vortices. In Fig. 1(a), the sensors either move
with the meandering jet or they are captured to eddies and
almost stay at the same place. This model is verified in [5]
for Gulf Stream measurements and used for underwater sensor
networks in [6]. In this work, we simplify the motion to a
horizontal drift. In Fig. 1(b), the simplified mobility model
is shown. Here, we assume that the sensors in the jet are
separated from the ones in the eddies with straight lines
and there is no crossing between two regions. We define the
velocities in the horizontal and vertical axis with vx and vy ,
respectively. For the nodes in the jet, vx is chosen between
vjet min < vx < vjet max and for the nodes outside the jet
veddy min < vx < veddy max where veddy max < vjet min . The

978-1-4244-2219-7/08/$25.00 2008 IEEE.

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE INFOCOM 2008 proceedings.

Fig. 2.

Fig. 3.

Fig. 1.
model

(a) Network architecture and the mobility, (b) simplified mobility

simplified model assumes that vy i = 0, for i = 1, 2, .., N,


where N is the total number of underwater nodes. Moreover,
the jet is divided into velocity zones similar to highway lanes.
The nodes in the central part drift with the highest velocity
and the nodes in the two surrounding velocity zones drift with
lower velocity values.
This USN architecture employs passive moving (drifting)
underwater nodes and MBS nodes which have controlled
movement along the vertical axis. Note that, MBSs also drift
with the currents along x-y axis.
The framework includes two consecutive rounds: localization and routing. These rounds start after MBS nodes descent
to the level of the underwater nodes. The MBS nodes are
calibrated to stay underwater for enough time to at least
accomplish localization. After localization, the remaining time
is used for sending data. We assume that all the nodes are clock
synchronized. Synchronization can be assumed for short term
missions. Sensor networks that need to work underwater for
several weeks require an additional mechanism for synchronization.
III. L OCALIZATION
MBS nodes learn their coordinates via GPS while they
are floating on the surface. Then, they descent and send
localization messages [7]. Localization message includes the
ID, the coordinates, the velocity of the MBS node and a
timestamp field. The timestamp field is used to determine the
distance via Time of Arrival (ToA) method. Nodes measure

MBS localization packet format

Active Underwater node localization packet format

the range with ToA and do lateration to estimate their location.


The underwater nodes are able to locate themselves with
at least three localization messages sent from non-collinear
positions. z-coordinate is the depth and it is derived from the
pressure sensor. In our scheme we use four MBS nodes to
decrease the estimation error. The format of the localization
message sent by the MBS is given in Fig. 2. Localization
message also includes the length of each round. This may
be pre-configured or downloaded to MBS nodes via satellite.
Nodes forward localization messages during the localization
period. After this period, nodes that are localized and have
data to sent start sending their data. They are allowed to send
until the end of the routing period.
To decrease the number of the MBSs, we use iterative
localization. Successfully localized underwater nodes become
active and announce self coordinates. Localization message of
active nodes include additional fields: the total number of hops
to MBS nodes, the estimated error in localization, MBS ID,
MBS coordinates, MBS time stamp and MBS velocity of the
four MBS nodes that the node has used in localizing itself.
The format of the localization message of the active nodes is
given in Fig. 3. The total number of hops to the MBS nodes
and the estimated error are used in iterative localization. If
a node hears from more than four MBS nodes then the one
with least hop count and error is selected. The position and
the velocity of the MBS nodes are used in routing.
The localization success and the error of the iterative
localization scheme is given in Fig. 4. for a static underwater
network with the number of underwater sensor nodes varying
between 50-200 and 25 MBS nodes. The preliminary results
on mobile scenarios show that more than 80% of the nodes
may be localized with this technique. The simulations are
performed in Qualnet Simulator [8] using an acoustic physical
layer.

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE INFOCOM 2008 proceedings.

Localized nodes
1

the nodes may be so low that given a limited routing round,


the nodes may not have the chance to meet.
C/P works in corporation with the localization scheme.
It does not need extra messages for position or velocity
information exchange. Moreover, C/P delays the packets on
the nodes until they meet with the sink. This is cost and energy
efficient for USNs since the low bandwidth, high bit error rate
and propagation delay [9] enforces limited messaging.

Mean error (m)


30

0.9
25

Localized nodes/nodes

0.8
0.7

20

Mean Error

0.6
0.5

15

0.4

10

0.3
0.2

V. C ONCLUSION

0.1
0
50

100

150

0
200 50

nodes

100

150

200

nodes

Fig. 4. Localization success and error for the number of underwater nodes
between 50-200 and 25 MBSs.

IV. ROUTING
The underwater nodes learn the beginning of the routing
round and the ID of the potential sinks at the localization
round. Each node independently selects an MBS and forward
its data towards that MBS. The forwarding decision is taken
according to the relative position and velocity of the MBS
node. The underwater nodes benefit from mobility and try to
catch up with the MBSs. According to the catch up or pass
(C/P) algorithm, a node either carries the packet until it catches
up with the selected MBS or, if the MBS and the node is not
likely to catch up then it forwards the packet to a slower or
faster relay.
The algorithm works as follows. The source node compares
its position and velocity with the MBS. If MBS is behind and
the velocity of the MBS is larger than the node (vM BS >
vN ), then, MBS has a chance to catch up with the node. The
packet is delayed for some time. Delay is a function of the
distance in between and the difference between the velocities,
i.e., delay = f (distance, vM BS , vN ). If MBS is behind and
vM BS vN , the MBS will never catch up, so the packet is
passed to a slower neighbour. Note that, the velocity values
of the neighbours are learnt during localization. If MBS is
ahead of the forwarding node and vM BS < vN , C/P delays
the packet on the node again because this time the node can
catch up with MBS. However, if vM BS vN then, the node
is unable to reach MBS and it passes the packet to a faster
neighbour.
When the packet is carried, waiting for a possibility of
catching up, it is delayed until the forwarding zone of the
MBS (see Fig1(b)). When the node is within the forwarding
zone, geographic coordinates are used for forwarding. Because
the velocities indicate the horizontal displacement (vyi = 0)
and the packet is delayed until the x coordinates of the node
and the MBS are close enough, in the vertical axis they may
be further away. In the forwarding zone, only y coordinates
are used to reach the MBS. The node with the least vertical
distance is selected as the relay.
Note that, the delay value is bounded by some threshold
because even though the direction and the velocities of the
nodes may allow each other to catch up, the relative speed of

In this work, we propose a framework for localization and


routing in mobile underwater sensor networks. We use mobile
beacons both as location servers and sinks. Their vertical
movement enable GPS driven coordinates to be distributed
in underwater. Moreover, they deliver data from the depths
of the network to the surface of the ocean. Our routing
algorithm uses position and velocity of the sensors. This
approach can be used in VANETs for a highway scenario,
as well. In this paper, we introduce the simplest framework
and enhancements will be made as future work. Due to high
propagation delay in underwater, using predicted positions
rather than the coordinates sent in packets will be considered.
Moreover, routing loops, voids and ping-pong effect will be
analyzed.
R EFERENCES
[1] Neptune project, http://www.neptunecanada.ca/about/index.html
[2] J. A. Rice, Us navy seaweb development, in Proc. of WuWNet, pp.
3-4, Montreal, QC, Canada, 2007.
[3] Argo project, http://www.argo.ucsd.edu
[4] C. Bettstetter, Mobility Modeling, Connectivity, and Adaptive Clustering in Ad Hoc Networks, Utz Verlag, 2004.
[5] A. S. Bower, A simple kinematic mechanism for mixing fluid parcels
across a meandering jet, J. Phys. Ocean., vol. 21(1), pp. 173180, 1991.
[6] A. Caruso, F. Paparella, L. Vieira, M. Erol, and M. Gerla, Meandering
current model and its application to underwater sensor networks, in to
appear in INFOCOM08, 2008.
[7] M. Erol, L. F. M. Vieira, and M. Gerla, Localization with divenrise
(dnr) beacons for underwater acoustic sensor networks, in Proc. of
WuWNet, pp. 97-100, Montreal, QC, Canada, 2007.
[8] Qualnet Network Simulator, http://www.scalable-networks.com/
[9] J. Heidemann, W. Ye, J. Wills, A. Syed, and Y. Li, Research challenges
and applications for underwater sensor networking, in IEEE Wireless
Communications and Networking Conference, Las Vegas, Apr. 2006.

Anda mungkin juga menyukai