Design of Communication and Control for

Swarms of Aquatic Surface Drones

Anders Lyhne Christensen

1,2,3

, Sancho Oliveira

1,2,3

, Octavian Postolache

1,2

, Maria Jo

ao de Oliveira

2,4

Susana Sargento

1,5

, Pedro Santana

1,2

, Lu

ıs Nunes

1,2

, Fernando Velez

1,6

, Pedro Sebasti

1,2

Vasco Costa

1,2,3

, Miguel Duarte

1,2,3

, Jorge Gomes

1,3,7

, Tiago Rodrigues

1,2,3

and Fernando Silva

1,3,7

Instituto de Telecomunicac¸

oes, 1049-001 Lisbon, Portugal

Instituto Universit

ario de Lisboa (ISCTE-IUL), 1649-026 Lisbon, Portugal

BioMachines Lab, 1649-026 Lisbon, Portugal

Vitruvius FabLab-IUL, 1649-026 Lisbon, Portugal

Universidade de Aveiro, 3810-193 Aveiro, Portugal

Universidade da Beira Interior, 6201-001 Covilh

a, Portugal

LabMAg, Faculdade de Ci

encias da Universidade de Lisboa, 1749-016 Lisbon, Portugal

Keywords:

Robotics Platform, Digital Manufacturing, Mesh Networks, Evolutionary Robotics, Decentralized Control.

Abstract:

The availability of relatively capable and inexpensive hardware components has made it feasible to consider

large-scale systems of autonomous aquatic drones for maritime tasks. In this paper, we present the CORATAM

and HANCAD projects, which focus on the fundamental challenges related to communication and control in

swarms of aquatic drones. We argue for: (i) the adoption of a heterogeneous approach to communication in

which a small subset of the drones have long-range communication capabilities while the majority carry only

short-range communication hardware, and (ii) the use of decentralized control to facilitate inherent robust-

ness and scalability. A heterogeneous communication system and decentralized control allow for the average

drone to be kept relatively simple and therefore inexpensive. To assess the proposed methodology, we are cur-

rently building 25 prototype drones from off-the-shelf components. We present the current hardware designs

and discuss the results of simulation-based experiments involving swarms of up to 1,000 aquatic drones that

successfully patrolled a 20 km-long strip for 24 hours.

1 INTRODUCTION

Maritime tasks are usually expensive to carry out due

to the use of manned vehicles with large operational

crews. While effort has been made to adapt unmanned

vehicle technology for use in maritime tasks, such

systems are currently relatively expensive to acquire

and operate, and only a single or a few vehicles are

typically deployed (Yan et al., 2010).

An alternative approach is the use of autonomous

systems composed of large numbers of relatively sim-

ple and inexpensive drones (swarms). The use of

swarms is advantageous given that many maritime

tasks such as environmental monitoring, sea life local-

ization, and sea-border patrolling require distributed

sensing. The goals of our ongoing HANCAD and

CORATAM projects are to overcome fundamental

challenges related to communication and control in

large-scale swarms of aquatic surface drones. In the

HANCAD project, we propose to use a heteroge-

neous network architecture in which only a subset

of the drones are required to carry long-range com-

munication equipment. As part of the project, we

will study and develop novel routing algorithms to

achieve effective communication in such ad-hoc het-

erogeneous networks. In the CORATAM project,

we propose to use a novel hybrid approach (Duarte

et al., 2014a) to the semi-automatic synthesis of self-

organized behavior for swarms of aquatic drones. The

potential beneﬁts of decentralized control based on

self-organization include scalability and robustness to

faults (Brambilla et al., 2013), both of which are es-

sential in many real-world scenarios.

In this paper, we present the major components of

our ongoing work, namely: (i) the design of our pro-

totype hardware, (ii) the heterogeneous communica-

548

Lyhne Christensen A., Oliveira S., Postolache O., João de Oliveira M., Sargento S., Santana P., Nunes L., Velez F., Sebastião P., Costa V., Duarte M.,

Gomes J., Rodrigues T. and Silva F..

Design of Communication and Control for Swarms of Aquatic Surface Drones.

DOI: 10.5220/0005281705480555

In Proceedings of the International Conference on Agents and Artiﬁcial Intelligence (ICAART-2015), pages 548-555

ISBN: 978-989-758-074-1

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

tion approach, and (iii) the methodology adopted for

the synthesis of self-organized control. The rest of the

paper is organized as follows. In Section 2, we discuss

the potential application of swarms of aquatic drones

to real-world tasks. We then discuss the challenges,

describe our proposed solutions, and the studies con-

ducted so far in terms of hardware (Section 3), com-

munication (Section 4), and control (Section 5). Fi-

nally, in Section 6, we discuss our ongoing work and

the future prospects for large-scale swarms of aquatic

drones.

2 ROBOTS FOR MARITIME

TASKS

Aquatic robots have been studied for a wide range

of applications including hydrography (Plueddemann

et al., 2008), environmental monitoring (Pinto et al.,

2014), geology (Lane et al., 1997), archeology (Clark

et al., 2008), defense (Clegg and Peterson, 2003), and

search and rescue (Furukawa et al., 2006). Over the

past decade, signiﬁcant public and private investment

has been made in the areas of autonomous underwa-

ter vehicles and autonomous surface vehicles (Man-

ley, 2008; Douglas-Westwood, 2012). The focus has

largely been on single-robot systems with a high de-

gree of hardware and software complexity. While

these systems have been applied to a variety of sce-

narios and have proven commercially viable (see,

for instance, offerings by Kongsberg

, Autonomous

Surface Vehicles Ltd

, and Blueﬁn Robotics

), they

are expensive and limited in terms of the tasks they

can undertake. In particular, tasks involving mon-

itoring, searching, or data collection over large ar-

eas typically require distributed sensing capabilities.

Distributed sensing can only be achieved by systems

composed of multiple, physically independent units

such as swarms of aquatic drones.

Land-based and air-based swarm robotics sys-

tems have been studied extensively, see Dorigo et al.

(2013); Lindsey et al. (2012) for examples, but the

same does not hold true for swarms for aquatic en-

vironments. While there have been numerous con-

ceptual studies on systems composed of multiple au-

tonomous vehicles for aquatic environments, only a

limited number of such systems have been realized.

The EU-ICT project CoCoRo (Schmickl et al., 2011)

is among a few exceptions. CoCoRo concerns the de-

sign and development of a swarm of autonomous un-

http://www.kongsberg.com/

http://www.asvglobal.com/

http://www.blueﬁnrobotics.com

derwater vehicles for deep-sea exploration. CoCoRo

uses custom-built robots and bio-inspired algorithms

for underwater tasks. One of the goals of CoCoRo

is to contribute to ﬁelds such as biology and meta-

cognition. The aim of our HANCAD and CORATAM

projects, on the other hand, is to address the key chal-

lenges related to the application of large-scale swarms

of aquatic drones in real-world scenarios.

Swarms of autonomous aquatic drones have sev-

eral potential real-world applications. Coastal coun-

tries have, for instance, faced an increased spending

on maritime missions over the years. In Italy, the

problem of illegal immigration (Monzini, 2007) and

organized crime (Lutterbeck, 2006) has contributed

to the growth of the Guardia di Finanza’s operation.

In 2013, the operation’s budget

amounted to $4.08B

and it employed 302 boats, 86 helicopters, and 16

airplanes, as well as a total of 59,335 military per-

sonnel (Carta, 2013). In Spain, immigrants cross-

ing the Gibraltar Strait through Morocco led to the

implementation of the Sistema Integrado de Vigilan-

cia Exterior (SIVE) in the late 1990s (Lutterbeck,

2006). SIVE relies on military-grade technology such

as ﬁxed and mobile radars, infrared sensors, and tra-

ditional aquatic and aerial vehicles. Such surveillance

and intruder detection tasks could potentially bene-

ﬁt from large-scale autonomous robotic swarms. Fur-

thermore, autonomous drones could be used for non-

military operations, such as aquaculture inspection,

environmental monitoring, and disaster relief.

In the HANCAD and CORATAM projects, we

will design and implement communication and con-

trol strategies for swarms of relatively simple and in-

expensive surface drones. Drones will be able to com-

municate through an ad-hoc heterogeneous wireless

network. The communication system will allow a re-

mote human operator to maintain a connection with

the swarm at all times through a subset of drones

equipped with long-range communication hardware.

We will use a novel approach that combines evo-

lutionary robotics techniques with manual engineer-

ing (Duarte et al., 2014a) to synthesize control semi-

automatically. The hybrid technique has the poten-

tial to combine the respective strengths of artiﬁcially

evolving control (Nolﬁ and Floreano, 2000), namely

the automatic synthesis of self-organized behavior,

with the beneﬁts of ﬂexible, engineering-oriented ap-

proaches in which the experimenter has ﬁne-grained

control over behavior.

The primary contribution of the projects will be

threefold: (i) we will develop a scalable, heteroge-

neous, and fault-tolerant ad-hoc network architecture

http://www.rgs.mef.gov.it/VERSIONE-I/Dati/

OPENDATA/SpeseBS/

DesignofCommunicationandControlforSwarmsofAquaticSurfaceDrones

549

39 cm

75 cm

58 cm

33 cm

39 cm

62 cm

39 cm

62 cm

Prototype I Prototype II Prototype III Prototype IV

40 cm

65 cm

Prototype V Prototype V Prototype V Prototype V

(above) (front) (below) (in water)

Figure 1: Top: Prototype I-IV, bottom: Prototype V, the current iteration of the aquatic drone prototype.

for swarms of aquatic drones, (ii) we will explore our

novel approach to control synthesis in a variety of

real-world maritime tasks, such as patrolling and in-

truder detection, environmental monitoring, or infras-

tructure inspection, and (iii) we will release all soft-

ware and hardware as open-source, which will allow

other researchers to build their own aquatic drones,

and to advance the state-of-the-art with respect to the

application of autonomous drones for maritime tasks.

3 HARDWARE

One of the main goals of the HANCAD and

CORATAM projects is to build prototype hardware

to serve as a platform for research and development

of swarms of aquatic drones. The platform is planned

to be orders of magnitude cheaper to build (< 1000

EUR per unit) than current commercial unmanned

surface vehicle, relatively small (< 1 meter in length),

and easy to manufacture to allow for large-scale de-

ployment of drones in swarms of hundreds of units

or more. We use widely available hardware, and

off-the-shelf sensors and motors. Prototype drones

will be developed based on open-source hardware and

open-source software. Digital manufacturing tech-

niques will be used in order to keep costs low and

facilitate adaptation and replication by third-parties.

Schematics, 3D models, and source code are available

at http://biomachineslab.com/aquaticdrone.

We use the Raspberry Pi (Upton and Halfacree,

2013) as the main computing device of each drone.

GPS receivers and compasses will provide each drone

Table 1: List of hardware in Prototype V.

Control

Raspberry Pi Model B, 512Mb RAM

Kingston 16Gb SD card Class 10

TP-Link TL-WN722N high-gain 150Mbps wireless dongle

Propulsion

Two 2213N 800Kv Brushless Motor motors

Two Turnigy TrackStar 25A speed controllers

Two 4mm Drive Shaft and 255mm Boat Shaft Sleeve

Power

ZIPPY Flightmax 8000mAh 3S1P 30C (for motors)

ZIPPY Flightmax 5000mAh 3S1P 40C (for everything else)

Turnigy 5A SBEC switching DC-DC regulator

Sensors

Sparkfun MAG3110 triple axis magnetometer

Adafruit GPS breakout (based on the MTK3339 chipset)

with localization and orientation information. The

drones can be further augmented with sensors for en-

vironmental monitoring, sea life detection, and other

task-speciﬁc equipment, which can be used in parallel

with the drone’s camera to demonstrate the potential

of collectives of aquatic drones.

3.1 Preliminary Results and Ongoing

Work

We have experimented with different designs of the

drone hull to achieve a good balance between hydro-

dynamic properties, size, and manufacturability. Cur-

rently, we have gone through ﬁve iterations of the

hull design, see Figure 1. In each iteration, we ﬁrst

designed and modeled a prototype in Rhinoceros

http://www.rhino3d.com/

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

550

and the design was then manufactured in extruded

polystyrene foam (XPS) using a 3-axis computer nu-

merical control milling machine. The use of XPS has

three main advantages, namely: (i) it is a relatively

inexpensive material, (ii) it has a low mass density

and it is not porous, and (iii) XPS is easily machin-

able. Furthermore, we also tested a variety of off-the-

shelf components to be used in the drones’ propulsion

system, such as shafts, motors, and speed controllers.

Other components of the robot are 3D printed, such

as motor mounts and propellers. In the current pro-

totype (Prototype V), we use the hardware listed in

Table 1. The total cost of a Prototype V unit is ∼ 260

EUR. In our ongoing work, we are equipping the

drones with additional hardware such as retractable

nets and water quality sensors, as well as a number of

communication technologies, namely Wi-Fi, ZigBee,

and WiMAX. In the longer term, we will give drones

the capacity to extend their operational autonomy by

equipping them with onboard photovoltaic panels.

4 COMMUNICATION

One of the main goals of the projects is to enable

swarms of drones to operate as a robust wireless sen-

sor network (WSN) where each node is embedded on

a low-cost aquatic drone. Mobile ad-hoc networks

(MANETs) have been widely studied, see Ko and

Vaidya (2000); Chlamtac et al. (2003); Akyildiz et al.

(2005) for examples. Studies on MANETs are typ-

ically conducted on systems in which nodes are ei-

ther conﬁned to a relatively small area, or present

at densities high enough to practically ensure net-

work connectivity between any two nodes regard-

less of movement, e.g. Sibley et al. (2002). Sce-

narios in open environments in which mobile nodes

must move to ensure connectivity have been studied.

However, the tasks were limited to either maintain-

ing network connectivity, or establishing connectiv-

ity between two ﬁxed points. In Winﬁeld (2000),

for instance, a swarm of robots must remain aggre-

gated based on connections formed over a low-range

wireless network, while in Hauert et al. (2009), aerial

robots must self-organize to establish and maintain

a wireless network between operators located on the

ground.

Distributed sensing over extended periods of time

is necessary for several maritime tasks such as pa-

trolling and environmental monitoring. In our design

of the communication system, we therefore prioritize

autonomy and robustness. To keep the per-unit cost

low, we will furthermore use a heterogeneous sys-

tem of drones: the majority of the drones will only

be equipped with relatively short-range communica-

tion equipment, while a few, more complex drones,

will also be equipped with long-range communica-

tion equipment and larger, more expensive batteries.

All drones will participate in task execution. The

drones with long-range communication capabilities

will serve as gateways through which observations

can be communicated to human operators and through

which operators can issue new instructions. We ex-

pect the ratio between the number of simple drones

and the number of complex drones to be between 10:1

and 25:1 depending on mission requirements.

Long-range communication can be achieved us-

ing high-gain Wi-Fi, 3G/GPRS, WiMAX, LTE, or

similar technologies. For missions in which drones

need to operate outside the range of terrestrial net-

works, satellite links could be used. In the HANCAD

and CORATAM projects, we are experimenting with

high-gain Wi-Fi and WiMAX.

Local drone-to-drone communication can be

achieved through a number of existing technologies.

In our projects, we will experiment with ZigBee and

Wi-Fi, but the speciﬁc technology used to establish

links between neighboring drones and to push data

across can be chosen depending on mission require-

ments. The key challenge related to the local drone-

to-drone communication is to attain mesh-networking

topologies that support the dynamic routing between

drones with only local communication capabilities

and gateway drones. Although there are several ad-

hoc routing protocols for networks with changing

topologies, we furthermore have to implement op-

portunistic routing and dissemination: some drones

can temporarily be outside the range of the rest of

the swarm due to mission requirements and limited

communication range. The aquatic environment also

poses delivery challenges due to the signal reﬂec-

tions, which must be taken into account in the routing

and dissemination protocols. We will develop a cus-

tom software layer on top of the low-level wireless

protocols to increase robustness and to support reli-

able connections across an ad-hoc network with con-

stantly changing topology and occasional signal re-

ﬂections. A comprehensive survey of the challenges

and proposed solutions for using wireless communi-

cation technologies in sensor networks deployed in

marine environments is presented in Xu et al. (2014).

Scenarios in which aquatic drones operate in open

maritime environments are challenging because their

task is not only to maintain network connectivity, but

also to carry out missions that can put constraints on

the spatial conﬁguration and motion of the drones.

The heterogeneous nature of the system must be taken

into account to ensure that drones with long-range

DesignofCommunicationandControlforSwarmsofAquaticSurfaceDrones

551

communication capabilities are accessible to the en-

tire swarm. The distance between any drone with

only local communication capabilities and a drone

with long-range communication capabilities should

be kept as short as possible to ensure reliable and

timely communication with human operators. If an

intruder is detected in a patrolling scenario, for in-

stance, it is important that human operators are noti-

ﬁed and can issue new instructions immediately. In

our ongoing work, we are studying the interplay be-

tween behavior and communication, as well as con-

ducting communication tests in real hardware.

5 CONTROL

Centralized control of multirobot systems, such as

swarms of aquatic drones, is attractive because plan-

ning, coordination and monitoring can be done based

on global knowledge of the system. However, com-

putational and/or communication constraints on the

robots may prevent centralized control (Crespi et al.,

2008). Moreover, centralized systems tend to be sub-

ject to scalability constraints and to be vulnerable

given that the central coordinator represents a single

point of failure. Systems based on decentralized con-

trol, on the other hand, do not have a single point of

failure, and when coordination is achieved through lo-

cal interactions, they tend to scale well, see for ex-

ample Christensen et al. (2009). Decentralized con-

trol based on self-organization is, however, difﬁcult

to design by hand because the behavioral rules for

individual robots cannot be derived from a desired

macroscopic behavior (Dorigo et al., 2004). In the

domain of large-scale, decentralized robot collectives,

the complexity stemming from the intricate dynam-

ics required to produce self-organized behavior fur-

ther complicates the hand-design of control systems.

In the ﬁeld of evolutionary robotics (ER), evolu-

tionary techniques are applied to automate the de-

sign of control systems for robots (Nolﬁ and Flore-

ano, 2000). ER techniques can be employed to syn-

thesize decentralized control for multirobot systems,

see Floreano and Keller (2010); Sperati et al. (2008)

for examples. Over the years, researchers have identi-

ﬁed certain challenges associated with the application

of ER. One of the most prevalent challenges concerns

bootstrapping the evolutionary process in complex

tasks. If controllers for a relatively complex task are

sought, evolution may be unable to ﬁnd a ﬁtness gra-

dient that leads to adequate solutions (Nelson et al.,

2009). Another challenge is the use of evolved con-

trol in real hardware. Except for a few cases in which

evolution is conducted directly on real hardware (see,

for instance Watson et al. (1999)), the evolution of

robotic controllers is conducted ofﬂine, in simulation,

due to the large number of evaluation necessary to

obtain capable controllers. Simulation-speciﬁc fea-

tures, which are not present in the real world, may be

exploited by evolution. As a consequence, the pro-

cess of transferring evolved controllers to real robotic

hardware, known as crossing the reality gap, typically

fails to preserve the level of performance achieved in

simulation (Jakobi, 1997).

Unless very simple tasks are considered, it is dif-

ﬁcult to foresee which evolutionary setup might be

suitable for solving a particular task (Christensen

and Dorigo, 2006). Between the controller, ﬁtness

function, and evolutionary algorithm, many differ-

ent combinations of settings are possible. It then

becomes necessary to run the computationally in-

tensive evolutionary process, often multiple times

with different initial conditions, to assess if a par-

ticular setup produces useful solutions. This leads

to a time-consuming trial-and-error process. While

a few studies have been conducted in which evolu-

tion was applied in a more engineered-oriented man-

ner, such attempts have, so far, been ad-hoc (Silva

et al., 2014). Techniques such as incremental evo-

lution (Gomez and Miikkulainen, 1997), incremen-

tal robot shaping (Urzelai et al., 1998), and task-

decomposition (Lee, 1999; Whiteson et al., 2005)

have been proposed, but such approaches do not ad-

dress the semi-automatic synthesis of behavior in a

systematic way.

In our projects, we use a novel, hybrid ap-

proach (Duarte et al., 2014a) in which the above-

mentioned challenges are addressed by systemati-

cally combining artiﬁcial evolution, manual engineer-

ing, and hierarchical decomposition of behavior. By

adopting such an approach, we expect to be able to

obtain scalable and robust decentralized control for

large-scale swarms of aquatic drones.

5.1 Preliminary Results and Ongoing

Work

We have studied a task in which a swarm of up to

1,000 simulated aquatic drones (Duarte et al., 2014b)

had to patrol a 20 km-long coastal strip of the island of

Lampedusa, see Figure 2. We applied our hybrid ap-

proach (Duarte et al., 2014a) for synthesis of control,

in which a complete mission can be broken down into

a number of simpler sub-tasks until evolution or a hu-

man designer can ﬁnd suitable behaviors. Individual

behavior primitives are simple behaviors that are syn-

thesized to solve particular sub-tasks. These behavior

primitives are then combined hierarchically using be-

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

552

Pursue Intruder

Behavior PrimitiveBehavior Primitive

PatrolGo To Waypoint

Behavior Primitive

Top Level

Behavior Arbitrator

Recharge

low

battery

low

battery

charged

Intruder

Pursue

Waypoint

Go To

no intruder

for 5 min

within 10 meters

of waypoint

intruder

detected/alert

Patrol

battery

(a) (b)

Figure 3: Representation of: (a) the hierarchical controller, with one preprogrammed behavior arbitrator and three evolved

behavior primitives, and (b) the behavior arbitrator, a preprogrammed ﬁnite state machine.

Figure 2: A 20 km-long patrolling zone was used in

simulation-based experiments (Duarte et al., 2014b) where

up to 1,000 aquatic drones had to execute an intruder de-

tection task. The chosen patrolling zone would be enough

to cover the south coast of the island Lampedusa, a major

illegal immigration hub.

havior arbitrators, which are decision nodes that del-

egate control to their sub-controllers. For the patrol

task, we evolved three behavior primitives: “Go To

Waypoint”, “Patrol”, and “Pursue Intruder”. After the

behavior primitives had been evolved, we combined

them using a simple state-based preprogrammed arbi-

trator, see Figure 3. Our initial experiments demon-

strated that scalable and self-organized control could

be evolved for large swarms operating in an open mar-

itime environment (Duarte et al., 2014b).

Controllers for more complex tasks can have mul-

tiple hierarchical layers of both evolved and prepro-

grammed nodes, allowing for detailed control over

behavior. A collection of evolved behaviors can thus

be built and potentially reused in different missions.

Robotic control for complex tasks can be synthesized

in an incremental and hierarchical manner by combin-

ing successfully previously evolved/preprogrammed

behaviors, while issues related to performance on real

hardware can be addressed at each increment. In this

way, our approach (Duarte et al., 2014a) circumvents

two fundamental issues associated with the applica-

tion of evolutionary robotics, namely: (i) bootstrap-

ping the evolutionary process, and (ii) crossing the

reality gap.

6 CONCLUSIONS

The aim of the HANCAD and CORATAM projects

is to study how the fundamental challenges related

to communication and control in swarms of aquatic

drones can be overcome. We have built ﬁve hard-

ware prototypes, we have proposed a heterogeneous

networking architecture, and we have conducted the

ﬁrst study on the synthesis of scalable, self-organized

behaviors or drones operating in an open maritime en-

vironment.

We are currently ﬁnalizing the design of the hard-

ware, and we expect to conduct the ﬁrst experiments

on a swarm of real drones before the end of 2014.

By the summer 2015, we expect to have demon-

strated a system up to 25 operational drones car-

rying out proof-of-concept tasks such as patrolling

and environmental monitoring in water. All soft-

ware and hardware will be made freely available at

http://biomachineslab.com/aquaticdrone.

6.1 Future Work

In our ongoing work, we continue to study ways in

which to increase the capabilities of the drones while

still keeping them simple. The simpler the drone, the

lower per-unit cost, which in turn allows for wider

adoption and/or larger deployments. One of the ways

in which the capabilities of drones with relatively sim-

ple sensory equipment can be augmented is through

DesignofCommunicationandControlforSwarmsofAquaticSurfaceDrones

553

the mutual sharing of sensory data through local, sit-

uated communication (Rodrigues et al., 2015). The

sharing of onboard sensory data between neighbor-

ing robots allows individual drones to combine in-

formation from multiple sources to obtain informa-

tion about the environment that would otherwise not

be available. The received information can even be

used to implement collective virtual sensors, which

an evolved controller can use as if they were regular

onboard sensors. Initial experiments with collective

sensors have produced promising results (Rodrigues

et al., 2015).

The limited onboard sensing and processing capa-

bilities may make it difﬁcult for drones to navigate

in cluttered environments such as lakes and rivers.

To overcome this limitation, the drones could use of-

ﬂine generated semantic maps of the environment.

A typical semantic map will indicate which regions

of the aquatic environment are closer to the margin,

whereas another will pinpoint which regions corre-

spond to shallow waters. The semantic maps could

be constructed at the beginning or prior to a mission

using a single or a few, sophisticated vessels such as

the Riverwatch (Pinto et al., 2014).

As part of our more long-term efforts, we are de-

veloping state-of-the-art methods such as the applica-

tion cooperative co-evolution driven by behavioral di-

versity instead of a traditional ﬁtness function (Gomes

et al., 2014) to synthesize behaviors that enable het-

erogeneous swarms of drones to maintain connectiv-

ity while executing tasks, and online learning in large-

scale decentralized systems (Silva et al., 2012).

ACKNOWLEDGEMENTS

This work was supported by Fundac¸

ao para a

encia e a Tecnologia (FCT) under the grants,

SFRH/BD/76438/2011, SFRH/BD/89573/2012,

SFRH/BD/89095/2012, PEst-OE/EEI/LA0008/2013,

and EXPL/EEI-AUT/0329/2013.

REFERENCES

Akyildiz, I. F., Wang, X., and Wang, W. (2005). Wire-

less mesh networks: a survey. Computer Networks,

47(4):445–487.

Brambilla, M., Ferrante, E., Birattari, M., and Dorigo, M.

(2013). Swarm robotics: a review from the swarm

engineering perspective. Swarm Intelligence, 7(1):1–

41.

Carta, C. L. (2013). Rapporto annuale. Technical report,

Corpo della Guardia di Finanza, Rome, Italy.

Chlamtac, I., Conti, M., and Liu, J. J.-N. (2003). Mobile ad

hoc networking: imperatives and challenges. Ad Hoc

Networks, 1(1):13–64.

Christensen, A. L. and Dorigo, M. (2006). Incremental

evolution of robot controllers for a highly integrated

task. In 9th International Conference on Simulation of

Adaptive Behaviour (SAB), pages 473–484. Springer,

Berlin, Germany.

Christensen, A. L., O’Grady, R., and Dorigo, M. (2009).

From ﬁreﬂies to fault tolerant swarms of robots. IEEE

Transactions on Evolutionary Computation, 13(4):1–

12.

Clark, C. M., Olstad, C. S., Buhagiar, K., and Gambin, T.

(2008). Archaeology via underwater robots: Map-

ping and localization within maltese cistern systems.

In 10th International Conference on Control, Automa-

tion, Robotics and Vision (ICARCV), pages 662–667.

IEEE Press, Piscataway, NJ.

Clegg, D. and Peterson, M. (2003). User operational evalua-

tion system of unmanned underwater vehicles for very

shallow water mine countermeasures. In OCEANS

2003, pages 1417–1423. IEEE Press, Piscataway, NJ.

Crespi, V., Galstyan, A., and Lerman, K. (2008). Top-

down vs bottom-up methodologies in multi-agent sys-

tem design. Autonomous Robots, 24(3):303–313.

Dorigo, M., Floreano, D., Gambardella, L. M., Mondada,

F., Nolﬁ, S., Baaboura, T., Birattari, M., Bonani, M.,

Brambilla, M., Brutschy, A., et al. (2013). Swar-

manoid: a novel concept for the study of heteroge-

neous robotic swarms. IEEE Robotics & Automation

Magazine, 20(4):60–71.

Dorigo, M., Trianni, V., S¸ahin, E., Groß, R., Labella, T. H.,

Baldassarre, G., Nolﬁ, S., Deneubourg, J.-L., Mon-

dada, F., Floreano, D., et al. (2004). Evolving self-

organizing behaviors for a swarm-bot. Autonomous

Robots, 17(2-3):223–245.

Douglas-Westwood (2012). The world AUV market report

2012-2016. Technical report, Douglas-Westwood Ltd.

Faversham, UK.

Duarte, M., Oliveira, S. M., and Christensen, A. L. (2014a).

Evolution of hybrid robotic controllers for complex

tasks. Journal of Intelligent and Robotic Systems. In

press.

Duarte, M., Oliveira, S. M., and Christensen, A. L. (2014b).

Hybrid control for large swarms of aquatic drones.

In 14th International Conference on the Synthesis &

Simulation of Living Systems (ALIFE), pages 785–

792. MIT Press, Cambridge, MA.

Floreano, D. and Keller, L. (2010). Evolution of adaptive

behaviour in robots by means of Darwinian selection.

PLoS Biology, 8(1):e1000292.

Furukawa, T., Bourgault, F., Lavis, B., and Durrant-Whyte,

H. F. (2006). Recursive Bayesian search-and-tracking

using coordinated UAVs for lost targets. In 2006 IEEE

International Conference on Robotics and Automation

(ICRA), pages 2521–2526. IEEE Press, Piscataway,

NJ.

Gomes, J., Mariano, P., and Christensen, A. L. (2014).

Avoiding convergence in cooperative coevolution with

novelty search. In 13th International Conference

on Autonomous Agents and Multiagent Systems (AA-

MAS), pages 1149–1156, IFAAMAS, Richland, SC.

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

554

Gomez, F. and Miikkulainen, R. (1997). Incremental evolu-

tion of complex general behavior. Adaptive Behavior,

3-4(5):317–342.

Hauert, S., Zufferey, J., and Floreano, D. (2009). Evolved

swarming without positioning information: an appli-

cation in aerial communication relay. Autonomous

Robots, 26(1):21–32.

Jakobi, N. (1997). Evolutionary robotics and the radi-

cal envelope-of-noise hypothesis. Adaptive Behavior,

6(2):325–368.

Ko, Y.-B. and Vaidya, N. H. (2000). Location-aided routing

(lar) in mobile ad hoc networks. Wireless Networks,

6(4):307–321.

Lane, D. M., Davies, J. B. C., Casalino, G., Bartolini, G.,

Cannata, G., Veruggio, G., Canals, M., Smith, C.,

O’Brien, D. J., Pickett, M., et al. (1997). Amadeus:

advanced manipulation for deep underwater sampling.

IEEE Robotics & Automation Magazine, 4(4):34–45.

Lee, W.-P. (1999). Evolving complex robot behaviors. In-

formation Sciences, 121(1-2):1–25.

Lindsey, Q., Mellinger, D., and Kumar, V. (2012). Con-

struction with quadrotor teams. Autonomous Robots,

33(3):323–336.

Lutterbeck, D. (2006). Policing migration in the mediter-

ranean. Mediterranean Politics, 11(1):59–82.

Manley, J. E. (2008). Unmanned surface vehicles, 15 years

of development. In OCEANS 2008, pages 1–4. IEEE

Press, Piscataway, NJ.

Monzini, P. (2007). Sea-border crossings: The organization

of irregular migration to italy. Mediterranean Politics,

12(2):163–184.

Nelson, A. L., Barlow, G. J., and Doitsidis, L. (2009).

Fitness functions in evolutionary robotics: A survey

and analysis. Robotics and Autonomous Systems,

57(4):345–370.

Nolﬁ, S. and Floreano, D. (2000). Evolutionary robotics:

The biology, intelligence, and technology of self-

organizing machines. MIT Press, Cambridge, MA.

Pinto, E., Marques, F., Mendonc¸a, R., Lourenc¸o, A., San-

tana, P., and Barata, J. (2014). An autonomous

surface-aerial marsupial robotic team for riverine en-

vironmental monitoring: Beneﬁting from coordinated

aerial, underwater, and surface level perception. In

2014 IEEE International Conference on Robotics and

Biomimetics (ROBIO). IEEE Press, Piscataway, NJ. In

press.

Plueddemann, A., Packard, G., Lord, J., and Whelan, S.

(2008). Observing arctic coastal hydrography using

the REMUS AUV. In 2008 IEEE/OES Conference on

Autonomous Underwater Vehicles (AUV), pages 1–4.

IEEE Press, Piscataway, NJ.

Rodrigues, T., Duarte, M., Oliveira, S. M., and Chris-

tensen, A. L. (2015). Beyond onboard sensors in

robotic swarms: Local collective sensing through sit-

uated communication. In 7th International Confer-

ence on Agents and Artiﬁcial Intelligence (ICAART).

SciTePress, Lisbon, Portugal. In press.

Schmickl, T., Thenius, R., Moslinger, C., Timmis, J.,

Tyrrell, A., Read, M., Hilder, J., Halloy, J., Campo,

A., Stefanini, C., et al. (2011). CoCoRo – The self-

aware underwater swarm. In 5th IEEE Conference

on Self-Adaptive and Self-Organizing Systems Work-

shops (SASO), pages 120–126. IEEE Press, Piscat-

away, NJ.

Sibley, G. T., Rahimi, M. H., and Sukhatme, G. (2002).

Robomote: A tiny mobile robot platform for large-

scale ad-hoc sensor networks. In 2002 IEEE In-

ternational Conference on Robotics and Automation

(ICRA), pages 1143–1148. IEEE Press, Piscataway,

NJ.

Silva, F., Duarte, M., Oliveira, S. M., Correia, L., and Chris-

tensen, A. L. (2014). The case for engineering the

evolution of robot controllers. In 14th International

Conference on the Synthesis & Simulation of Living

Systems (ALIFE), pages 703–710. MIT Press, Cam-

bridge, MA.

Silva, F., Urbano, P., Oliveira, S., and Christensen, A. L.

(2012). odNEAT: An algorithm for distributed online,

onboard evolution of robot behaviours. In 13th Inter-

national Conference on the Simulation & Synthesis of

Living Systems (ALIFE), pages 251–258. MIT Press,

Cambridge, MA.

Sperati, V., Trianni, V., and Nolﬁ, S. (2008). Evolving co-

ordinated group behaviours through maximisation of

mean mutual information. Swarm Intelligence, 2(2-

4):73–95.

Upton, E. and Halfacree, G. (2013). Raspberry Pi user

guide. John Wiley & Sons, Hoboken, NJ.

Urzelai, J., Floreano, D., Dorigo, M., and Colombetti, M.

(1998). Incremental robot shaping. Connection Sci-

ence, 10(3-4):341–360.

Watson, R., Ficici, S., and Pollack, J. (1999). Embodied

evolution: Embodying an evolutionary algorithm in a

population of robots. In 1999 IEEE Congress on Evo-

lutionary Computation (CEC), pages 335–342. IEEE

Press, Piscataway, NJ.

Whiteson, S., Kohl, N., Miikkulainen, R., and Stone, P.

(2005). Evolving keepaway soccer players through

task decomposition. Machine Learning, 59(1):5–30.

Winﬁeld, A. F. (2000). Distributed sensing and data collec-

tion via broken ad hoc wireless connected networks

of mobile robots. In Distributed Autonomous Robotic

Systems 4 (DARS), pages 273–282. Springer, Berlin,

Germany.

Xu, G., Shen, W., and Wang, X. (2014). Applications of

wireless sensor networks in marine environment mon-

itoring: A survey. Sensors, 14(9):16932–16945.

Yan, R., Pang, S., Sun, H., and Pang, Y. (2010). Develop-

ment and missions of unmanned surface vehicle. Jour-

nal of Marine Science and Application, 9:451–457.

DesignofCommunicationandControlforSwarmsofAquaticSurfaceDrones

555