Networked Control of Multiple Marine Vehicles:

Theoretical and Practical Challenges in the Scope of the

EU GREX Project

A. Aguiar

, J. Almeida

, M. Bayat

, B. Cardeira

, R. Cunha

, A. Hausler

P. Maurya

, A. Oliveira

, A. Pascoal

, A. Pereira

, M. Rufino

L. Sebastião

, C. Silvestre

and F. Vanni

Dept. Electrical Engineering and Computers / Institute for Systems and Robotics (ISR)

Instituto Superior Técnico (IST), Lisbon, Portugal

Dept. Computer Science, University of Southern California (USC), U.S.A.

Abstract. This paper overviews some of the theoretical and practical issues that

arise in the process of developing advanced motion control systems for cooper-

ative multiple autonomous marine vehicles (AMVs). Many of the problems ad-

dressed were formulated in the scope of the EU GREX project, entitled Coordi-

nation and Control of Cooperating Heterogeneous Unmanned Systems in Un-

certain Environments. The paper offers a concise introduction to the general

problem of cooperative motion control that is well rooted in illustrative mission

scenarios developed collectively by the GREX partners. This is followed by the

description of a general architecture for cooperative autonomous marine vehicle

control in the presence of time-varying communication topologies and stringent

communication constraints. The results of simulations with the NetMarSyS

(Networked Marine Systems Simulator) of ISR/IST are presented and show the

efficacy of the algorithms developed for cooperative motion control. The last

part of the paper focuses on practical issues and describes the results of a series

of tests at sea in the Azores, in the Summer of 2008. The paper concludes with

a discussion of theoretical and practical implementation issues that warrant fur-

ther research and development effort.

1 Introduction

Worldwide, there has been increasing interest in the use of autonomous marine ve-

hicles (AMVs) to execute missions of increasing complexity without direct supervi-

sion of human operators. A key enabling element for the execution of such missions

is the availability of advance systems for motion control of AMVs. The past few

decades have witnessed considerable interest in this area. The problems of motion

control can be roughly classified into three groups: i) point stabilization, where the

goal is to stabilize a vehicle at a given target point with a desired orientation; ii) tra-

jectory tracking, where the vehicle is required to track a time parameterized reference,

Aguiar A., Almeida J., Bayat M., Cardeira B., Cunha R., Hauster A., Maurya P., Oliveira A., Pascoal A., Ruﬁno M., Sebasti

ao L., Silvestre C., Vanni F.

and Pereira A. (2009).

Networked Control of Multiple Marine Vehicles: Theoretical and Practical Challenges in the Scope of the EU GREX Project.

In Proceedings of the International Workshop on Networked embedded and control system technologies: European and Russian R&D cooperation,

pages 146-160

 SciTePress

and iii) path following, where the vehicle is required to converge to and follow a

desired geometric path, without a timing law assigned to it.

Current research goes well beyond single vehicle control. In fact, recently there

has been widespread interest in the problem of cooperative motion control of fleets of

AMVs. A particular important scenario that motivates the cooperation of multiple

autonomous vehicles and poses great challenges to systems engineers, both from a

theoretical and practical standpoint, is automatic ocean exploration/monitoring for

scientific and commercial purposes. In this scenario, one can immediately identify

two main disadvantages of using a single, heavily equipped vehicle: lack of robust-

ness to system failures and inefficiency due to the fact that the vehicle may need to

wander significantly to collect data over a large spatial domain. A cooperative group

of vehicles connected via a mobile communications network has the potential to

overcome these limitations. It can also reconfigure the network in response to envi-

ronmental parameters in order to increase mission performance and optimize the

strategies for detection and measurement of vector/scalar fields and features of par-

ticular interest. Furthermore, in a cooperative mission scenario each vehicle may only

be required to carry a single sensor (per environmental variable of interest) making

each of the vehicles in the formation less complex, thus increasing its reliability.

As an example, Fig. 1.a captures a conceptually simple mission scenario where an

autonomous surface craft (ASC) and an autonomous underwater vehicle (AUV) ma-

neuver in synchronism along two spatial paths, while aligning themselves along the

same vertical line, so as to fully exploit the good properties of the acoustic communi-

cations channel under these conditions. This is in striking contrast to what happens

when communications take place at slant range, for this reduces drastically the band-

width of the channel, especially due to multipath effects in shallow water operations.

Cooperative Autonomous Marine Vehicle Motion Control is one of the core ideas

exploited in the scope of the EU GREX project, entitled Coordination and Control of

Cooperating Heterogeneous Unmanned Systems in Uncertain Environments, see [1].

Both theoretical and practical issues are addressed in the scope of the project. It is

worth to stress that from a theoretical standpoint, the coordination of autonomous

robotic vehicles involves the design of distributed control laws in the face of dis-

rupted inter-vehicle communications, uncertainty, and imperfect or partial measure-

ments. This is particular significant in the case of underwater vehicles. It was only

recently that these subjects have started to be tackled formally, and considerable re-

search remains to be done to derive multiple vehicle control laws that can yield good

performance in the presence of severe communication constraints. For previous work

along these lines, the reader is referred to [3], [4], [14], [17], [18], [19], [22], [30],

[31], [32], [34].

The structure of the paper is as follows. In section 2 we give the practical motiva-

tion for the problem of cooperative multiple vehicle control with the help of a repre-

sentative scientific mission scenario that emerged naturally in the scope of the EU

Project GREX. Section 3 describes a general architecture for cooperative autonomous

marine vehicle control in the presence of time-varying communication topologies and

communication losses. Section 4 contains the results of computer simulations aimed

at assessing the efficacy of the algorithms developed for cooperative motion control.

Section 5 contains experimental results. Finally, Section 6 summarizes the main re

147

Fig. 1. a) Cooperative control of two (surface and underwater) autonomous marine vehicles for

data gathering at sea; b) Marine habitat mapping scenario.

sults obtained and discusses briefly issues that warrant further research and develop-

ment work.

2 Practical Motivation and Scientific Mission Scenarios

In what follows we describe one of a large number of mission scenarios that have

been discussed and defined in detail by the GREX partner group. The mission scena-

rios envisioned are rooted in challenging problems in the field of marine science.

They also bring out the ever increasing important role that marine technology is hav-

ing in terms of affording marine scientists the tools that are needed to explore and

exploit the ocean. We place the focus on missions for which two basic ingredients are

required: i) the missions require the use of several intelligent autonomous vehicles

equipped with appropriate instrumentation, and ii) inter-vehicle coordination and

mission control is dynamic and highly dependent on the type of information obtained

as the missions unfold.

Mission Scenario: Marine Habitat Mapping

Habitat maps of the marine environment, that is, maps containing data on the bathy-

metry and nature of the seabed as well as on the type and localization of biological

species, are the key to an in-depth understanding of the distribution and extent of

marine habitats. Knowledge of the distribution of marine habitats serves to establish

sensible approaches to the conservation needs of each habitat and to facilitate a better

management of the marine environment through an understanding of how particular

human activities are undertaken in relation to marine habitats. This will in turn allow

for the establishment of policies capable of ensuring sustainable development. This

subject is receiving widespread attention worldwide because of its far reaching impli-

cations and has led to the definition of a number of guidelines and directives for the

study and preservation of marine habitats. At an European level, for example, Annex

I of the celebrated EU Habitats Directive establishes that marine habitats classified as

Special Areas of Conservation (Natura 2000) need special assessment in order to

verify their accordance with the European Union requirements. The mission scenario

for marine habitat mapping proposed here was greatly influenced by and aims to

148

automate and improve classical procedures that are normally used by marine scien-

tists. The key ideas can be explained by referring to Fig. 1.b). For simplicity of expo-

sition, we start by focusing on the ASV/ROV ensemble in the figure, where the ROV

is connected to the ASV through a thin umbilical for fast data transmission.

In this scenario, the ASV executes a lawn mowing manoeuvre above the seabed

automatically, while the ROV executes a similar manoeuvre in cooperation with the

ASV. Using this set-up, the ROV transmits pictures of the seabed back to the support

ship (and thus to the scientist in charge) via a radio link installed on-board the ASV.

A number of AUVs stay dormant either on the seabed or at the sea surface. Upon

detection of interesting patterns on the seabed by the scientist in charge, a signal is

sent to a selected member of the AUV fleet (via an acoustic communication link

installed on-board the ASV), to dispatch it to the spot detected so as to map the sur-

rounding region in great detail. Meanwhile, the ASV/ROV ensemble continues to

execute the lawn mowing manoeuvre in search of other sites of interest. With the

methodology proposed, sites that are interesting from an ecological viewpoint are

easily detected along the transects.

To execute the abovementioned challenging missions, a number of autonomous

vehicles must work in cooperation, under high level human supervision. This entails

the development of advanced systems for cooperative motion control and navigation

in the presence of severe underwater communication constraints, together with the

respective software and hardware architectures.

3 A General Architecture for Multiple Vehicle Cooperation

This section describes a very general architecture for multiple vehicle cooperative

motion control that has emerged naturally out of a research effort in which the au-

thors have been participating. The section further describes key single and multiple

vehicle motion control primitives that were judged appropriate for practical imple-

mentation of the architecture developed on a set of multiple heterogeneous vehicles,

in the scope of the GREX project.

3.1 Multiple Vehicle Cooperative Motion Control

The systems that are at the root of the architecture adopted for multiple vehicle coop-

eration are depicted in Fig. 2. See [3] for a fast paced introduction to the subject. The

scheme depicted is quite general and captures basic trends in current research.

Each vehicle is equipped with a navigation and control system that uses local in-

formation as well as information provided by a subset of the other vehicles over the

communication network, so as to make the vehicle manoeuver in cooperation with the

whole formation. Navigation is in charge of computing the vehicle's state (e.g. posi-

tion and velocity). Control accepts references for selected variables, together with the

corresponding navigation data, and computes actuator commands so as to drive track-

ing errors to zero. The cooperation strategy block is responsible for implementing

cooperative navigation and control. Its role is twofold: i) for control purposes, it

issues high level synchronization commands to the local vehicle based on information

149

available over the network (e.g. speed commands to achieve synchronization of a

number of vehicles executing path following maneuvers); ii) For navigation purpos-

es, it merges local navigation data acquired with the vehicle itself as well as by a

subset of the other vehicles. This is especially relevant in situations where only some

of the vehicle can carry accurate navigation suites, whereas the others must rely on

less precise sensor suites, complemented with information that is exchanged over the

network. Finally, the system named logic-based communications is responsible for

supervising the flow of information (to and from a subset of the other vehicles),

which we assume is asynchronous, occurs on a discrete-time basis, has latency, and is

subject to transmission failures.

Fig. 2. A general architecture for multiple vehicle cooperation.

Central to the above scheme is the fact that each vehicle can only exchange infor-

mation with a subset of the remaining group of vehicles. Furthermore, and because of

the intrinsic nature of the underwater communications channel, communications

should be parsimonious and take place at a very low data rate. This calls for the im-

plementation of systems to decide when and what minimum information should be

transmitted from each of the vehicles to its neighbours. Interestingly enough, analog-

ous constraints appear in the vibrant area of networked control systems, from which

interesting and fruitful techniques can be borrowed.

Close inspection of the general architecture for multiple vehicle cooperation de-

scribed above reveals the plethora of problems to be solved:

• Cooperative Control (CC) (e.g. cooperative path following and cooperative

trajectory tracking),

• Cooperative Navigation (CN),

• CC and CN under strict communication constraints over a faulty, possibly

150

switched network.

In the scope of the GREX project, considerable work was done to advance design

tools to tackle the above problems. For a description of key technical aspects in-

volved in the development of advanced schemes for single and multiple vehicle con-

trol, the reader is referred to [2], [3], [4], [6], [7], [8], [10], [11], [12], [17], [18], [19],

[20], [33], [34], [35]. See also [13], [14], [15], [22], [23], [25], [26], [28], [29], [30],

[31], [32], [36], and the references therein for an overview of the state of progress in

the area.

The results obtained so far hold potential for application. To the best of our know-

ledge, some of the work reported is pioneering in that it effectively addresses explicit-

ly time-varying communication networks with temporary failures and latency in the

transmissions, and logic-based communications aimed at reducing the amount of

discrete-time data to be transmitted among the vehicles. The results obtained were

instrumental in defining, together with the GREX partners, a library of Single and

Multiple Vehicle Primitives (MVPs) for motion control that are described in the next

section.

3.2 Single and Multiple Vehicle Primitives

The work envisioned in the scope of the GREX project aims at affording system

designers the tools to develop, using a “bottom-up” approach, the modules that are

needed to implement a true Multi Vehicle Mission Control System for a fleet of auto-

nomous vehicles.

Based on the mission scenarios and the general architecture for multiple vehicle

cooperation described in the previous sections, a set of Multiple Vehicle Primitives

(MVPs) for coordinated motion control were developed. The definition of the primi-

tives and the algorithms for their implementation take into account the fact that the

vehicles considered have complex dynamics, exhibit large parameter uncertainty, are

often underactuated, and must perform well in the presence of unknown, shifting

ocean currents. During the first part of the project, the attention was focused on the

development of primitives enabling the following tasks:

• Point Stabilization, Path Following, and Trajectory Tracking of single

marine vehicles with complex dynamics.

• Path Planning for multiple vehicles.

• Cooperative Path Following of multiple vehicles.

• Cooperative Target Following and Cooperative Target Tracking of multiple

vehicles.

• Cooperative Manoeuvring in the presence of tight communication

constraints by exploiting recent research results on Networked Control

Systems.

• All of the above in the presence of sensor and actuator faults.

In what follows we provide a brief description of each of the tasks listed above and

point out relevant bibliography that describes the motion control algorithm solutions

developed by the ISR/IST team.

151

Point Stabilization (also referred to as Go to Point) refers to the problem of steering a

vehicle to a point with a desired orientation (in the absence of currents), or simply to

a desired point without a desired orientation (in the presence of currents). The algo-

rithms derived are reported in [4] and [6].

Path Following. In this task, the objective is to steer a vehicle towards a path and

make it follow that path with an assigned speed profile. Notice that there are no ex-

plicit temporal specifications, that is, the vehicle is not required to be a certain point

at a desired time. Rather, what is relevant is for the vehicle to traverse the path, albeit

with a speed that may be path dependent. Algorithms are reported in [2], [17], [33],

and [34].

Trajectory Tracking. In contrast with the Path Following objectives, what is now

required is that the vehicle track a desired temporal/spatial trajectory. Timing con-

straints become important for this task. In practice, trajectory tracking systems are

harder to design (when compared with Path Following systems) and may yield

“jerky” maneuvers and large actuator activity. This is because of tight temporal con-

straints; see [5] and [9]. In this respect, Path Following strategies usually lead to more

benign maneuvers. However, there are instances in which one is forced to adopt tra-

jectory tracking strategies (for example, when one wishes to investigate a phenome-

non that is strongly time-dependent). Algorithms are summarized in [2].

Path Planning for Multiple Vehicles. Multiple vehicle path planning methods build

necessarily on key concepts and algorithms for single vehicle path following. How-

ever, they go one step further in that they must explicitly take into account such is-

sues as inter-vehicle collision avoidance and simultaneous times of arrival. See [21]

and the references therein.

The literature on path planning is vast and the methodologies used are quite di-

verse. Classical methodologies aim at computing feasible strategies off-line that mi-

nimize a chosen cost criterion. More recently, new methodologies have come to the

forum where the objective is to generate paths on-line, in response to environmental

data, so as to optimize the process of data acquisition over a selected area. In the

scope of GREX we focused on the problem that arises when multiple vehicles are

scattered in the water and it is required that they safely reach the starting location of a

cooperative mission with a desired formation pattern and assigned terminal speeds

(Go-To-Formation manoeuver). The cost criteria of interest may include minimizing

travel time or energy expenditure. The key objective was to obtain path planning

methods that are effective, computationally easy to implement, and lend themselves

to real-time applications.

The techniques that we developed for multiple vehicle path planning are based

upon and extend the work reported in [24] for unmanned air vehicles. See [20] and

[21] for recent work on the subject, with applications to autonomous marine vehicles.

Explained in intuitive terms, the key idea exploited is to separate spatial and temporal

specifications, effectively decoupling the process of spatial path computation from

that of computing the desired speed profiles for the vehicles along those paths. The

first step yields the vehicles' spatial profiles and takes into consideration geometrical

constraints; the second addresses time related requirements that may include, among

others, initial and final speeds, deconfliction in time, and simultaneous times of arriv-

152

al. Decoupling the spatial and temporal constraints can be done by parameterizing

each path as a set of polynomials in terms of a generic variable

and introducing a

polynomial function

(

) that specifies the rate of evolution of

with time, that is,

/dt =

(

). By restricting the polynomials to be of low degree, the number of para-

meters used during the computation of the optimal paths is kept to a minimum. Once

the order of the polynomial parameterizations has been decided, it becomes possible

to solve the multiple vehicle optimization problem of interest (e.g., simultaneous time

of arrival under specified deconfliction and energy expenditure constraints) by resort-

ing to any proven direct search method; see [27].

Cooperative Path Following. In this case, a fleet of vehicles is required to track a

series of pre-defined spatial paths, while holding a desired formation pattern at a

desired “formation speed”. The implementation of the corresponding MVP calls for

the execution of a path following algorithm for each of the vehicles, together with a

synchronization algorithm that changes the nominal speeds of the vehicles so as to

achieve the desired temporal synchronism. The algorithms used are described in [3],

[7], [10], [11], [17], [18], [19], and [34] and take into account explicitly the topology

of the inter-vehicle communication network.

Cooperative Target Following (CTF) and Cooperative Target Tracking (CTT). The

CTF and CTT tasks enable a group of vehicles to follow (in space) and track (in

space and time) a moving target, respectively. The CTF refers to the situation where

the group of vehicles follows the path traversed by the target, without stringent tem-

poral constraints. This is done by “observing” the target motion, extracting from it a

spatial reference path, and following it. No further objective is attempted, and the

distance between the group of vehicles and the target is left uncontrolled. As an ex-

ample, we cite the situation where a manned vessel leads (“shows the way” to) a

group of marine craft through a harbour area where obstacles are present. By observ-

ing the motion of the manned vessel, the group of vehicles learns a safe path across

the harbour and follows it accurately (“doing by imitation”). The CTT is similar to

CFT, except that it is now required for the group of vehicles to maintain a desired

along-path distance from the target. Instead of traversing the path defined by the

target “at leisure”, the group of vehicles is required to adjust its overall speed so as to

keep a desired distance to the target. These two problems are far from trivial in the

case when the trajectory to be tracked is not available apriori, but is instead defined

implicitly by the unknown motion of a target vehicle. Interestingly, enough, both

problems can be solved by converting them into an equivalent path following prob-

lem. This is done by having at least one vehicle in the formation “observe the mo-

tion” of the target and fit a parameterized path to it over a short, receding time win-

dow. The parameters of the consecutive segments of paths thus obtained are then

broadcast to the other vehicles, and a coordinated path following algorithm executed.

Cooperative Manoeuvring in the Presence of Tight Communication Constraints.

This task refers to the problem of developing MVPs for Cooperative Path Following

and Cooperative Target Following and Target Tracking in the presence of varying

communication topologies, communication losses, and delays. The latter is especially

relevant in view of the small speed of propagation of sound in the water. Solutions

are proposed in [3], [18], and [19]. In [3], the solutions address explicitly the fact that

underwater communications occur at discrete intervals of time, thus reducing drasti-

153

cally the frequencies at which the vehicles communicate. As far as we could ascer-

tain, previous work along these lines is not available in the literature for multiple

underwater vehicle control. The new solution adopted borrows from related work in

networked control and holds potential for further refinement aimed at striking an

adequate balance between performance and energy spent to communicate.

4 Simulation Results

In this section we show results of simulations that illustrate the performance that can

be achieved with the motion control algorithms mentioned before. The simulations

were done using the Networked Marine Systems Simulator (NetMar

SyS), a software

suite developed at ISR/IST in the scope of GREX to simulate different types of coop-

erative missions involving a variable number of heterogeneous marine craft, each

with its own dynamics, see [35]. The high level of detail with which the environment

can be modeled affords end-users the tools that are necessary to take into account

both the effect of water currents on the vehicle dynamics as well as the delays and

environmental noise that affect underwater communications. The simulation kernel

developed so far paves the way for future developments aiming at incorporating more

sophisticated acoustic communication models and communication protocols, together

with interfaces to allow seamless distributed software and hardware-in-the-loop simu-

lation.

The NetMar

SyS interface is divided into four main areas: mission environment, mis-

sion specifications, vehicles, and output interface. The mission environment area

includes three different menus: water current, coordination strategy (which defines

the inter-vehicle communication topology), and communication channel. The mission

specifications area includes a list of possible missions to be executed, e.g. Coopera-

tive Path Following and Cooperative Target Tracking. The area devoted to vehicles

contains a file with a number of different vehicle blocks (kinematics and dynamics).

Here, the user can choose the number and the type of vehicles in the formation. Final-

ly, an output interface enables the visualization of mission results and the creation of

videos from the simulations.

The simulator has been instrumental in evaluating the efficacy of selected algo-

rithms for motion control of marine vehicles. By incorporating blocks that emulate

the actual software code that is implemented on-board the different vehicles, the si-

mulator has also been a valuable tool to evaluate the software for the implementation

of MVPs, and in fact has played a key role in the preparation for the first series of

field tests in the Azores, in the Summer of 2008.

An illustrative 3D example

We now illustrate the application of the results in [3] to coordinate three AUVs

moving in three-dimensional space.

The AUVs are required to follow paths of the form p

di(

i)=[ci cos(2

d), ci

sin(2

d), d

i ], with c1 = 20m, c2 = 15m, c3 = 25m, d = 0.05m, T=400, and

= -3

/4. The initial positions are p1 = (10m, -15m, -5m), p2 = (5m, 0m, 0m), p3 =

(20m, -25m, 5m). The vehicles start at rest and the normalized reference speed was set

154

Fig. 3. Coordinated path-following of 3 AUVs, with logic-based communication.

to vr = 1m/s. The vehicles are also required to keep a formation pattern that consists

of having them aligned along a straight line in the plane. Furthermore, AUV 1 is

allowed to communicate with AUVs 2 and 3, but the latter two do not communicate

between themselves directly. To further illustrate the behaviour of the proposed co-

operative path-following control architecture, we also force the following scenario:

from t=150s to t=250s, AUV 1 is only capable of following its path with d

1/dt = 0.5.

Fig. 3 shows the trajectories of the AUVs and Fig. 4 the evolution of the overall path-

following error

| pi - pdi |, coordination error |

2| + |

3|, and the communication

signal

. The signal

∈

{0,1} indicates, by switching its value, when communica-

tions occur. Before t=150s, the vehicles adjust their speeds to meet the formation

requirements and the coordination errors converge to zero. Note the reduced number

of communications exchanged during that period. In fact, the vehicles only need to

communicate a few times during the transient phase. When AUV 1 is forced to slow

down from t

∈

[150, 250] (without transmitting explicitly to its neighborhoods its

new reference velocity), the communication rate increases in order to keep the coor-

dination error bounded.

5 Experimental Results

In July 2008 the first series of GREX field tests took place at Horta, Faial, in the

archipelago of the Azores. The tests were instrumental in bringing together the differ-

ent partners to perform hardware and software integration and paved the way for full

development of the tools that are needed for multiple vehicle cooperative control and

navigation.

155

Fig. 4. Path-following error, coordination error, and communication signal σ.

Fig. 5. a) Results of the first GREX mission at sea with the DELFIMx: Going-to and Following

a Lawn-Mowing Maneuver; b) The DELFIMx ASV (left) and the manned vessel Aguas Vivas.

It was early decided that one of the tests would involve two surface vehicles un-

dergoing joint motion: the Aguas Vivas manned vessel and the DELFIMx autonom-

ous surface vehicle equipped with a dedicated GREX computer, both shown in Fig.

5.b. The DELFIMx is an autonomous surface craft that was designed and built at the

Instituto Superior Tecnico, Lisbon, Portugal. It is a small Catamaran 4.5m long and

2.4m wide, with a mass of 380kg. Propulsion is ensured by three-bladed propellers

driven by electrical motors. The maximum speed of the vehicle with respect to the

water is 3.0m/s. The vehicle is equipped with on-board resident systems for naviga-

tion, guidance and control, and mission control. Navigation is done by integrating

motion sensor data obtained from an attitude reference unit, a Doppler Log unit, and a

DGPS (Differential Global Positioning System). Transmissions between the vehicle

and its support vessel, or between the vehicle and a control center installed on-shore

are achieved via a radio link with a range of 10km. The vehicle has a wing shaped

156

central structure that is lowered during operations at sea. Installed at the bottom of

this structure is a low drag body that can carry acoustic transducers, including those

used to communicate with submerged craft.

Two vehicles primitives were executed with success: Path Following (PF) and

Target Following (TF). Fig. 5.a shows a lawnmowing PF maneuver executed by

DELFIMx. To test the Target Following primitive, the AGUAS VIVAS manned

vessel underwent arbitrary motion at sea while transmitting its GPS position to DEL-

FIMx, see Fig. 5.b. Based on the GPS information received, DELFIMx identified on-

line, using a path fitting algorithm, the path segments traversed by Aguas Vivas (line

segments or segments of arcs identified over short receding time windows) and fol-

lowed these paths at a set speed by invoking repeatedly the PF primitive. As a conse-

quence, DELFIMx maneuvered well along the overall path "defined by" Aguas Vi-

vas, not known a priori. The results of this maneuver are shown in Fig. 6. The tests

proved extremely important in evaluating the performance of the algorithms devel-

oped for path following and target following, the aerial communication channel be-

tween Aguas Vivas and DELFIMx, and the efficacy of the software/hardware archi-

tecture adopted within the project, namely that of the GREX computer installed on-

board the DELFIMx.

Fig. 6. Experimental results of the DELFIMx performing a target following maneuver in the

Azores, PT.

6 Conclusions and Future Work

This paper gave a brief overview of some of the theoretical and practical issues that

arise in the process of developing advanced motion control systems for cooperative

multiple autonomous marine vehicles (AMVs). Many of the problems addressed were

motivated by challenging scientific mission scenarios defined in the course of the EU

GREX project, entitled Coordination and Control of Cooperating Heterogeneous

Unmanned Systems in Uncertain Environments. A general architecture for coopera-

tive autonomous marine vehicle control in the presence of time-varying communica-

tion topologies and communication losses was proposed. The architecture implemen-

tation relies on a number of Single and Multiple Vehicle Primitives, the development

of which was rooted in solid control theory. The algorithms developed were fully

157

tested in simulation using the NetMarSyS - Networked Marine Systems Simulator -

developed by ISR/IST. The same simulator was used to do hardware in the loop si-

mulations prior to tests at sea, in the Azores, in the Summer of 2008. The field tests

were instrumental in evaluating the performance of the algorithms developed for path

following and target following, the aerial communication channel between Aguas

Vivas and DELFIMx, and the efficacy of the software/hardware architecture adopted

by the project team.

Future work will address the testing of other Multiple Vehicles Primitives (includ-

ing Go-To-Formation and Cooperative Target Tracking) and the definition of a final

set of integrated tests at sea, followed by their execution in the Azores in the Fall of

2009. From a theoretical standpoint, two main lines of research are envisioned: i)

cooperative navigation exploiting non-conventional geophysical-based navigation

systems, and ii) in-depth study of the constraints imposed by the underwater channel

and underwater communication protocols.

Acknowledgements

This work was supported in part by projects GREX/CEC-IST (contract No. 035223),

FREESUBNetwork (EU under contract number MRTN-CT-2006-036186) and NAV-

Control/FCT-PT (PTDC/EEA-ACR/65996/2006) and through the FCT-ISR/IST plu-

rianual funding program. We express our sincere thanks to all members of the GREX

and FREESUBNetwork projets and our colleagues at IST/ISR for their stimulating

discussions, contributions, and the commitment to go from the laboratory to sea. This

paper is but a brief summary of the work of many.

References

1. THE GREX PROJECT (2006-2009): Coordination and Control of Cooperating Heteroge-

neous Unmanned Systems in Uncertain Environments. URL http://www.grex-project.eu.

2. Aguiar, A., Hespanha, J.P. (2007), Trajectory-tracking and path-following of underac-

tuated autonomous vehicles with parametric modeling uncertainty, IEEE Trans. on Auto-

mat. Contr., 52(8):1362–1379.

3. Aguiar, A., Pascoal, A. (2007a), Coordinated path-following control for nonlinear systems

with logic-based communication, In Proc. of the 46th Conf. on Decision and Contr., pp.

1473–1479, New Orleans, LA, USA.

4. Aguiar, A., Pascoal, A. (2007b), Dynamic positioning and way-point tracking of underac-

tuated AUVs in the presence of ocean currents. Int. J. of Control, 80(7):1092–1108.

5. Aguiar, A., Hespanha, J. P. , Kokotovic, P. (2005), Path-following for non-minimum phase

systems removes performance limitations, IEEE Trans. on Automat. Contr., 50(2):234–239,

Feb. 2005

6. Aguiar, A., Hespanha, J. P. , Pascoal, A. (2007a), Switched seesaw control for the stabiliza-

tion of underactuated vehicles, Automatica, 43 (12):1997–2008.

7. Aguiar, A., Ghabcheloo, R., Pascoal, A., Silvestre, C. (2007b), Coordinated path-following

control of multiple autonomous underwater vehicles, In Proc. ISOPE’07 - International So-

ciety of Offshore and Polar Engineers, Lisbon, Portugal.

158

8. Aguiar, A., Ghabcheloo, R., Pascoal, A., Silvestre, C., Vanni, F. (2007c), Coordinated path

following of multiple marine vehicles: Theoretical issues and practical constraints. In Proc.

IWK - 52nd Internationales Wissenschaftliches Kolloquium, Ilmenau, Germany.

9. Aguiar, A., Hespanha, J. P. , Kokotovic, P. (2008a), Performance limitations in reference-

tracking and path-following for nonlinear systems, Automatica, 44(3):598–610.

10. Aguiar, A., Kaminer, I., Ghabcheloo, R., Pascoal, A., Hovakimyan, N., cao, C., Dobrokho-

dov, V. (2008b), Coordinated path following of multiple UAVs for time-critical missions in

the presence of time-varying communication topologies, In Proc. of the 17th IFAC World

Congress, Korea, Jul. 2008.

11. Almeida, J., Silvestre, C., Pascoal, A. (2007), Coordinated control of multiple vehicles with

discrete-time periodic communications, In Proc. of the 46th Conf. on Decision and Contr.,

pages 2888–2893, New Orleans, LA, USA.

12. Almeida, J., Silvestre, C., Pascoal, A. (2008), Compliant coordination and control of mul-

tiple vehicles with discrete-time periodic communications, In Proc. of the 17th IFAC World

Congress, Korea.

13. Alonge, F., D’Iippolito, F., Raimondi, F. (2001), Trajectory tracking of underactuated

underwater vehicles, Proc. 40th IEEE CDC, Orlando, Florida, USA, Vol. 5, pp. 4421–

4426.

14. Arkak, M. (2007), Passivity as a design tool for group coordination, IEEE Trans. on Au-

tomatic Control, 52(8):1380–1390.

15. Breivik, M., Fossen, T. (2005), Principles of guidance-based path following in 2D and 3D,

Proc. 44th IEEE Conference on Decision and Control.

16. Cardigos, F., Dando, P., Vila, S., Sarradin, P., Tempera, F., Conceio, P., Pascoal, A., San-

tos, R. (2005), Characterization of the shallow water hydrothermal vent field communities

of the D. João de Castro seamount (Azores). Chemical Geology, 224:153 – 168.

17. Ghabcheloo, R., Aguiar, A., Pascoal, A., Silvestre, C., Kaminer, I., Hespanha, J. (2006),

Coordinated path-following control of multiple underactuated autonomous vehicles in the

presence of communication failures, In Proc. 45nd Conf. on Decision and Contr., San Di-

ego, CA, USA

18. Ghabcheloo, R., Aguiar, A., Pascoal, A., Silvestre, C. (2007), Synchronization in multi-

agent systems with switching topologies and nonhomogeneous communication delays, In

Proc. of the 46th Conf. On Decision and Control, pp. 2327–2332, New Orleans, LA, USA

19. Ghabcheloo, R., Aguiar, A., Pascoal, A., Silvestre, C., Kaminer, I., Hespanha, J. (2009a),

Coordinated path-following in the presence of communication losses and time delays,

SIAM Journal on Control and Optimization, 48(1):234–265.

20. Ghabcheloo, R., Kaminer, I., Aguiar, A., Pascoal, A. (2009b), A General Framework for

Multiple Vehicle Time-Coordinated Path Following Control. In American Control Confe-

rence (to appear in the Proceedings).

21. Hausler, A., Ghabcheloo, R., Kaminer, I., Pascoal, A., Aguiar, A. (2009), Path Planning

for Multiple Marine Vehicles. To appear in the Proc. IEEE Oceans, Bremen, Germany.

22. Ihle, I., Arcak, M., Fossen, T. (2007), Passivity-based designs for synchronized path-

following, Automatica, 43(9):1508–1518.

23. Jiang, Z.-P. (2002), Global tracking control of underactuated ships by Lyapunov’s direct

method, Automatica, 38(2):301 – 309.

24. Kaminer, I., Yakimenko, O., Dobrokhodov, V., Pascoal, A., Hovakimyan, N., Cao, C.,

Young, A., Patel, V. (2007), Coordinated Path Following for Time-Critical Missions of

Multiple UAVs via L1 Adaptive Output Feedback Controllers, AIAA Guidance, Navigation

and Control Conference and Exhibit.

25. Klein, D., Matlack, C., Morgansen, K. (2007), Cooperative target tracking using oscillator

models in 3D, In Proc. American Control Conference, New York City, USA.

159

26. Klein, D., Bettale, P., Triplett, B., Morgansen, K. (2008), Autonomous underwater multive-

hicle control with limited communication: Theory and experiment, In Proc. NGCUV’08 -

IFAC Workshop on Navigation, Guidance and Control of Underwater Vehicles

27. Kolda, T., Lewis, R., Torczon, V. (2003), Optimization by direct search: New perspectives

on some classical and modern methods, SIAM Review, 45(3):141–166.

28. Lefeber, E., pettersen, K., Nijmeijer, H. (2003), Tracking control of an underactuated ship,

IEEE Transactions on Control Systems Technology, 11(1):52–61.

29. Sepulchre, R., Paley, D., Leonard, N. (2007), Stabilization of planar collective motion with

all-to-all communication, IEEE Trans. On Automat. Contr., 52(5):811–824, 2007.

30. Skjetne, R., Moi, S., Fossen, T. (2002), Nonlinear formation control of marine craft, In

Proc. 41st Conf. on Decision and Contr., Las Vegas, NV.

31. Skjetne, R., Fossen, T., Kokotovic, P. (2004), Robust output maneuvering for a class of

nonlinear systems, Automatica, 40(3):373–383.

32. Stillwell, D., Bishop, B. (2000), Platoons of underwater vehicles, IEEE Control Systems

Magazine, 20:45–52.

33. Vanni, F., Aguiar, A., Pascoal, A. (2007), Nonlinear motion control of multiple autonom-

ous underwater vehicles, In Proc. of CAMS’07 – The IFAC Conference on Control Appli-

cations in Marine Sytems, Bol, Croatia, Sep. 2007.

34. Vanni, F., Aguiar, A., Pascoal, A. (2008a), Cooperative path-following of underactuated

autonomous marine vehicles with logic-based communication, In Proc. of NGCUV’08 -

IFAC Workshop on Navigation, Guidance and Control of Underwater Vehicles, Killaloe,

Ireland, Apr. 2008.

35. Vanni, F., Aguiar, A., Pascoal, A. (2008b), NetMar

SyS

– Networked Marine Systems Simu-

lator, Technical Report WP6-0108, IST.

36. Zhang, F., Fratantoni, D., Paley, Lund, J, Leonard, N. (2007), Control of coordinated pat-

terns for ocean sampling, Int. J. of Control, 80(7):1186–1199.

160