AUVs as mobile nodes in acoustic communication

networks: field experience at the UAN10 experiment

Andrea Caiti, Vincenzo Calabr, Gianluca Dini, Angelica Lo Duca, Andrea Munaf
ISME - Interuniversity Res. Ctr. Integrated Systems for the Marine Environment, Italy
Centro E. Piaggio University of Pisa, Italy
Largo Lazzarino 1, 56122 Pisa, Italy

Abstract The UAN Underwater Acoustic Network project

has among its goals the investigation of networked
communication modalities in which mobile nodes are located in
the environment in order to optimize the communication
characteristics of the acoustic channel. The mobile network nodes
in UAN are placed on Autonomous Underwater Vehicles (AUVs)
carrying appropriate acoustic modems and capable of
autonomous decisions. With this contribution we focus on the
architectural design and implementation on AUVs of the
Folaga class of a mission supervisor handling the
communication, environmental sampling and decision making
tasks, and reports the first results in terms of network
communication capabilities as obtained in the UAN10
experimental activities that took place in the Tuscan Archipelago,
Mediterranean Sea, in September 2010.
Sensor networks; Autonomous Underwater Vehicles; Acoustic
Communication.

I.

INTRODUCTION

Underwater acoustic communication is a hot research topic

nowadays, since it is clearly seen as the enabling technology
to establish a network of fixed and/or mobile underwater
sensors. Application of such networks range from
environmental, climate and pollution monitoring to off-shore
industry to patrol and surveillance in security systems (both
military and civilian). A range of scientific issues in
underwater communication are investigated, for instance, in
the special issue [1], and are reviewed in [2], where an
extensive list of references is given. Within this research
stream, the UAN Underwater Acoustic Network [3] project
was established in 2008, having among its goals the
investigation of networked communication modalities in
which mobile nodes are located in the environment in order to
optimize the communication characteristics of the acoustic
channel. The mobile network nodes in UAN are placed on
Autonomous Underwater Vehicles (AUVs) carrying
appropriate acoustic modems and capable of autonomous
decisions. In particular, the AUVs must be able to
autonomously react to changes in the environment (e.g., in
oceanographic conditions) that affect the communication
channel characteristics and to appropriately relocate
themselves in order to guarantee some quality of service
performance in the communication task. Moreover, the
autonomous decisions must be taken in a distributed way, also
in presence of lack of communication among the network
nodes, and must in any case guarantee the fulfilment of a

general mission task, which in case of UAN is the surveillance

around a critical asset. The above goals are indeed ambitious,
and involve the development of appropriate both
methodological and hardware/software architectural tools.
From the point of view of distributed decision making in the
UAN context, a cooperative algorithm has been designed, and
its characteristics in terms robustness to node failures,
communication delays, incomplete information are reported in
a companion paper [4]. This contribution focuses on the
architectural design and implementation on AUVs of the
Folaga class [5] of a mission supervisor handling the
communication, environmental sampling and decision making
tasks, and reports the first results in terms of network
communication capabilities as obtained in the UAN10
experimental activities that took place in the Tuscan
Archipelago, Mediterranean Sea, in September 2010. UAN10
is an intermediate project activity with the goal of testing the
different network functionalities in order to prepare for the
conclusive experimental work scheduled for May 2011.
The paper is organized as follow: Section 2 describes the
eFolaga AUV a specific class of Autonomous Underwater
Vehicles used as mobile node of the network. In Section 3 we
go into the detail of the mission supervisor software
architecture giving details on its integration within the vehicle.
Section 4 reports the field experience during the UAN10
experiment providing field data on the network performance.
Finally, conclusions are given in Section 5.
II.

UAN MOBILE NODES: THE EFOLAGA AUV

This paragraph gives a brief overview on the general

characteristics of the eFolaga vehicles and on the
modifications of the AUVs done in order to integrate them
within the UAN network.
The most recent prototype of the vehicle class, the eFlaga,
has an actuation system combining those of autonomous
gliders with those of propulsion-driven AUVs. The vehicle
navigates on the sea surface with its own propulsion system
when in transit from one measuring station to another; it
submerges vertically, exploiting ballast and attitude changes,
when on station to perform the measurement. When at the
surface, the vehicle has continuous GPS (Global Positioning
System) contact and land-station contact through a mobile
phone link or a dedicated radio link. The land station link
allows for on-line modification of the mission requirements
and for almost real-time data transmission. The motion in the

surge direction is due to propulsion jet-pumps or to a propeller

(both options are available) at the vehicle stern, while steering
in surge sway plane is obtained through two lower power
jet-pumps at the vehicle bow, transversal to the surge
direction. Vehicle diving is obtained by combination of a
buoyancy change and attitude change through the presence of
a ballast chamber and a wormscrew mechanism to internally
displace the position of the battery pack. The diving actuation
system is similar to that of oceanographic gliders, though it
has to be noted that the Flaga has no wings or other
hydrodynamic surfaces and its intended dive is in the vertical
plane at zero pitch. A summary of the main technical
characteristics of the eFlaga is reported in Table 1.

Figure 2. Folaga AUV with the UAN module with the acoustic modem and
CTD mounted at mid-vehicle.
TABLE II.

PAYLOAD HARWARE: MAIN BOARD CHARACTERISTICS

Item
TABLE I.
Diameter (m)
External
Internal
Length (m)
Mass (Kg)
Mass variation range (Kg)
(assuming water density 1031
Kg/m3)
Range of moving mass
displacement (m)
Energy storage
Autonomy (hrs.)
Diving scope (m)
Break point in depth (m)
Speed
knots
m/s

EFOLAGA MAIN CHARACTERISTICS

0.140
0.125
2.000
30.0
0.500
0.050
Lead acid batteries, 12 V, 72 Ah
8
0 - 50
100
2 (jet pumps) / 4 (propeller)
1.01 / 2.02

To integrate the eFolagas within the UAN network a

specific payload with dedicated hardware has been designed
and implemented in order to connect the acoustic modem to the
vehicle electronics. Figure 1 shows a view of the payload
external structure with the protruding acoustic modem and
CTD sensor. The main hardware of the payload is shown in
Figure 1, right, and it is represented by a PC-104 board with
serial lines to communicate with the modem and with the CTD
probe which is available for continuous monitoring of the
water conditions. Figure 2 shows the Folaga AUV with the
UAN module mounted at mid-vehicle.

CPU
DRAM
Chipset
Serial Ports
USB Ports
Hard disk

III.

Description
1GHz, VIA EDEN, Ultra Low Voltage
1GB, DDR2, 533/400 on SO-DIMM
socket
VIA CX700M
1 RS232 Full modem
1 RS232FM/422/485 Configurable
2 x USB 2.0
4 GB Internal Flash Disk

SOFTWARE ARCHITECTURE

From an architectural perspective, the goal is that of

implementing a mission supervisor capable of interpreting and
generating messages to the other network nodes, and to give
commands to the vehicle native guidance, navigation and
control (GNC) system. The approach followed is the backseat driver paradigm, pioneered by the MIT group and coworkers [6 8]: the mission supervisor must be able to take
decisions and give high level commands to the native GNC
vehicle system, which is the sole responsible of the low-level
execution of the commands. Similarly, the supervisor has to
handle the communication tasks at the application level, while
the lower level communication layers are left to the
communication software implemented in the acoustic modem
itself. In this way, conceptually depicted in Figure 3, it is
possible to integrate in a modular way all the system
components reducing (ideally regardless) the specific nature of
the vehicle, of the acoustic modem and of the MAC and
routing strategy of the communication network.
Communication at the mission supervisor level is handled
through the MOOS system [5], a pub/sub middleware. The
mission supervisor is implemented as an event-driven Mealy
finite state machine, which generates an output based on its
current state and input, and which is in itself a MOOS process
able to establish a connection to a remote MOOS Database for
sending and receiving acoustic messages.

Figure 1. Left: UAN module with acoustic modem (from Kongsberg

Maritime) and CTD. Right: Paylod PC104 main board

A. Interface to the Folaga

One of the most critical parts in the architecture described
is represented by the interface between the mission supervisor
and the Flaga existing software. For this reason, to maintain
the maximum flexibility and separation between the mission
supervisor and the Folaga GNC native system, a dedicated
interface has been set up. Figure 5 shows the block diagram of
the main processes running on the whole vehicle, both in the
UAN payload and in the core of the Folaga. In particular, the
following process and their counter parts on the vehicle support
the mission supervisor and allows for the integration between
the payload and the vehicle itself:

Figure 3. Conceptual architectural scheme of the implemented system

In particular, the mission supervisor, called UANAgent,

extends the CMOOSClient class provided by the MOOS
framework:
Class UANAgent: public CMOOSClient {
bool OnNewMail(MOOSMSG_LIST& newMail);
bool Iterate();
}

The method OnNewMail() collects all the acoustic

messages received, and updates the data structures used by
UANAgent; the method Iterate() specifies the main thread of
the class, and implements the main operations, such as the
overall control (from an application standpoint) of the acoustic
communication and of the Folaga vehicle.

Folaga Controller: connects with the User Control Server

on the Folaga. It is responsible for the delivery to the Folaga
Control System of the commands issued by the MOOS Client
main process for the control of the vehicle, and vice versa to
communicate Folaga answers and/or errors to the mission
supervisor.
Radio Controller: connects to the Radio Modem Client on
the Folaga. It is responsible for receiving user commands when
the Folaga is on surface. Through this process the payload can
be configured and monitored in real time by the user.
Logger Client: connects to the Logger Server on the
Folaga. It sends logs from the MOOS process to the Folaga,
which in turn will send them via radio to the user, when the
vehicle in on surface.

The logic of the mission supervisor is realized through a state

machine composed by three states (IDLE, TASK, TEST) and
two macro-states (MISSION, AUTO), as depicted in the state
machine diagram shown in Figure 4. Each one of the states of
the machine is related to a single task to be executed by the
Folaga vehicle (e.g. Navigation task), a task to be executed by
the MOOS pub/sub system (e.g. send a specific acoustic
message), or a waiting condition. The two macro states allow
the mission supervisor to execute complex and autonomous
tasks, and in particular:
MISSION macro-state: the supervisor goes into this state
to execute a desired mission, i.e. an ordered sequence of tasks
for the Folaga AUV or for the MOOS pub/sub system. For
example, a typical mission can include: a diving task to reach a
desired depth followed by an underwater navigation task for a
desired period of time while sending periodic notification on
the vehicle status (e.g. battery level, depth, etc.) through the
acoustic link.
AUTO macro-state: when in this state, the mission
supervisor is able to adapt its behavior to changing
environmental conditions, executing periodic samplings of the
water column and moving to optimize the acoustic
communication.

Figure 4. Conceptual representation of the mission supervisor state machine

diagram. The red arrow indicates the initial state of the machine. The main
events for each state are also indicated. The two macro-states, AUTO and
MISSION, allow the machine to execute complex and autonomus missions.

TABLE III.
Packet
size

54 bytes

84 bytes

Figure 5. Interface between the Folaga vehicle and the UAN payload.
99 bytes

IV.

UAN10 EXPERIMENT

This paragraph report results from field data as obtained

during the UAN10 experiment where, in shallow water
conditions, the vehicle and network functionalities were tested
with up to two vehicles and two fixed nodes.
The experiment, held in Pianosa Island, Tuscan
Archipelago, Mediterranean Sea, in September 2010 tested the
underwater network for several hours, from the physical layer
up to the MOOS application level. In particular, the physical
layer of the network was represented by acoustic modems built
by Kongsberg Maritime capable of transmitting up to 500bps
and which also implemented the routing layer and multi-hop
strategies. The network architecture was hence completed by
the implementation of the PPP protocol to establish the IP
connection, and on the use of UDP or TCP as transport
protocols to allow the integration of the MOOS framework as
application level. Seen from an application level, each node of
the network was hence represented by a MOOS client
(UANAgent) which was responsible for monitoring and
managing the node status and behaviour and which connected
to a central database for sending and receiving data.

Average
Min
Max
Standard
deviation
Average
Min
Max
Standard
deviation
Average
Min
Max
Standard
deviation

SEPTEMBER 17, 2010

RTT (s)

Packet loss (%)

7.107
3.721
23.150

17%
-

4.485

9.151
4.935
26.865

14%
-

6.081

12.795
5.491
25.991

33%
-

63.99

Figure 6. Master modem location and Folaga path on September 17, 2010.
Arrows indicate Folaga direction along the route.

During UAN10 experiment, which was an intermediate

project activity with the goal of testing the different network
functionalities, it was possible to record data sets relative to
communication with MOOS pub/sub system configured and
active, in the open sea. In particular, in this work we report
results obtained on September 17 and on September 22 where
all the network layers were tested with increased level of
complexity.
The first test executed, on September 17 was a complete
test in terms of network layers but with limited complexity in
terms of network topology as only two nodes, one master and
one slave were used. Figure 4 shows the nodes locations: the
fixed node 1 (master modem) was located at 10 meters depth
while the Folaga was navigating at full speed (2 knots) at
surface, at 5m and 10m at varying ranges from 150m to
1000m. Figure 5 shows the CTD measurement obtained at the
fixed master station location. Table 3 shows the network
performance obtained on September 17, using the network in
the master-slave configuration in terms of Round Trip Time
and average packet loss percentage.

Figure 7. CTD casts taken on September 17, 2010, at location 42.5940N,

10.1071E

It is interesting to correlate these communication results

with the environmental conditions: as it can be seen from Fig.
9, there was no significant change of the sound speed structure
in the water column during the whole day. The sound speed
profile measured during the test, with a strong thermocline
approximately between 25 and 45 m depth, is always causing
downward refraction to acoustic propagation, and interaction
with the bottom. Looking at the bathymetry (Fig. 8), one can
note that the fixed node has at its West a rather gentle upslope,
and at its East a rather abrupt down slope. The Master node
was always eastward with respect to the fixed node; when the
Folaga was placed westward of the fixed node, it could not
relay messages to the fixed node, while successful
communication was achieved with Folaga on the eastward side
of the fixed node. One possible explanation for this behavior is
that in the eastward Folaga configuration there are less bottom
multipaths from the Folaga to the fixed node with respect to the
Folaga at the Westward situation. While this is only given as a
conjecture at the moment of writing, requiring supplementary
analysis which has not been completed yet, it is in any case
evident that, at least in this case, combined effects of
bathymetry and sound speed have indeed to be taken into
account in order to optimize channel characteristics.
Finally, in the second test, all the nodes were sending and
receiving in parallel, and again mobile node 2 was also acting
as a relay between the master and the fixed node 3. Table 5
reports the network statistics in this case. It is possible to see an
increase in the RTT and in the packet loss if compared with the
first test due to packet collisions created by the traffic increase.

Network nodes locations, September 22, 2010

42.596
Master modem (node 1)
42.595
50

42.594
L a titu d e N (d e g )

On September 22, true network connectivity was

established among the master node, located again at 10 meters
depth, and two slave nodes, one fixed, moored close to the sea
bottom at 51m depth (anchored at the bottom, in a total of 58 m
of water) and one moving node installed on the Flaga. Figure
8 shows the network topology during the experiment and the
bathymetric contours. Figure 9 shows the CTD profile obtained
at the master node location. The master node and the moving
node were above the thermocline (the Folaga was actually
always at the surface) for the whole test, while the fixed node,
placed at the bottom, was always right below it. Two tests were
performed. In the first test, which corresponds to the top-left
and top-right part of the Folaga path, the network traffic was
generated from the master node (node 1) to the fixed node 3,
with node 2, initially positioned West of the line connecting the
two fixed node, only acting as a bridge to establish a multi-hop
connection. In this initial condition, it was not possible to
correctly establish the network connectivity because of a too
high packet loss on the physical layer. After this first attempt
the mobile node was moved to the East of the line between the
master modem and the fixed node 3 and there was immediate
success of the physical relaying of packets, and the
consequence correct activation of the whole network stack.
Table 4 reports in this last configuration the network
performance in terms of RTT and average packet loss.

55

60

65 70
80

Fixed node 3

42.593
S2
42.592
42.591

E2

42.59
42.589
10.106 10.1065 10.107 10.1075 10.108 10.1085 10.109 10.1095 10.11 10.1105 10.111
Longitude E (deg)

Figure 8. Network nodes locations on September 22, 2010. The Folaga path
is shown overimposed on the bathymetric countours (main depth profiles in
the area of interest are tagged) with the dashed line. Folaga started at location
S2 and ended the mission at E2. Its orientation is indicated with the arrows.

Figure 9. CTD casts taken on September 22, 2010, at location 42.5940N,

10.1060E

TABLE IV.

SEPTEMBER 22, 2010 NETWORK PERFORMANCE (TWO NODES

CONNECTED VIA MULTI-HOP)

Packet
size

54 bytes

Average
Min
Max
Standard
deviation

RTT (s)

Packet loss (%)

15.432
9.618
31.675

20%
-

5.864

TABLE V.
Packet
size

54 bytes

54 bytes

SEPTEMBER 22, 2010 NETWORK PERFORMANCE WITH THREE

NODES.
Network path

1->2

1->2->3

Average
Min
Max
Standard
deviation
Average
Min
Max
Standard
deviation

RTT
(s)
23.794
4.514
75.253

Packet loss
(%)
58%
-

ACKNOWLEDGMENT
This work was supported in part by European Union, 7th
Framework Programme, Project UAN Underwater Acoustic
Network under Grant no. 225669
REFERENCES

20.563

[1]

45.947
7.192
99.976

34%
-

[2]

29.553

[3]
[4]

CONCLUSIONS
In the paper we described the architectural design and
implementation on AUVs of the Flaga class of a mission
supervisor handling the communication, environmental
sampling and decision-making tasks for the integration of the
autonomous underwater vehicle as a mobile node of an
underwater acoustic network.
The paper reports the first results in terms of network
communication capabilities as obtained in the UAN10
experimental activities that took place in the Tuscan
Archipelago, Mediterranean Sea, in September 2010. UAN10
is an intermediate project activity with the goal of testing the
different network functionalities in order to prepare for the
conclusive experimental work scheduled for May 2011.

[5]

[6]

[7]

[8]

[9]

J.Heidemann, U. Mitra, J. Preisig, M. Stojanovic, M. Zorsi, L. Cimini

(Eds.), Special issue on Underwater wireless communication network,
IEEE J. on Selected Areas in Communication, vol. 6, n. 9, Dec. 2008.
M. Chitre, S.Shahabuden, L.Freitag. M.Stojanovic, Recent advances in
underwater acoustic communication and networking, Proc. IEEE
Oceans08, Quebec City, Canada, 2008.
FP7 UAN Project Underwater Acoustic Network www.ua-net.eu
A.Caiti, A.Munaf, Cooperative distributed behaviours of an AUV
network for asset protection with communication constraints, Proc.
IEEE Oceans11 Europe, Santander, Spain, 2011.
A.Caffaz, A.Caiti, G.Casalino, A.Turetta, The hybrid glider/auv
FOLAGA, IEEE Robotics and Automation Magazine, vol. 17, n. 1, pp.
31 44, 2010.
A.Balasuriya, H.Schmidt, M.Benjamin Nested Distributed Autonomy
Architecture for Undersea Sensor Networks, Proc. IEEE/MTS
Oceans09, Biloxi, U.S.A., 2009.
D.P. Eickstedt, S.R. Sideleau, The Backseat Control Architecture for
Autonomous Robotic Vehicles: A Case Study with the Iver2 AUV,
Proc. IEEE Oceans09 Pacific, Kobe, Japan, 2008
M. Benjamin, H. Schmidt, P. Newman, J. Leonard , "Nested Autonomy
for Unmanned Marine Vehicles with MOOS-IvP", Journal of Field
Robotics, Volume 27, Issue 6, pp. 834-875, 2010.
P. M. Newman, MOOS - A Mission Oriented Operating Suite, MIT Department of Ocean Engineering, Tech. Rep. OE2003-07, 2003.