A Review on Multi-Robot Systems Categorised by Application Domain*
Rachael N. Darmanin
1
and Marvin K. Bugeja
1
Abstract— Literature reviews on Multi-Robot Systems (MRS)
typically focus on fundamental technical aspects, like coordina-
tion and communication, that need to be considered in order to
coordinate a team of robots to perform a given task effectively
and efficiently. Other reviews only consider works that aim
to address a specific problem or one particular application
of MRS. In contrast, this paper presents a survey of recent
research works on MRS and categorises them according to
their application domain. Furthermore, this paper compiles a
number of seminal review works that have proposed specific
taxonomies in classifying fundamental concepts, such as coor-
dination, architecture and communication, in the field of MRS.
I. INTRODUCTION
Since the late 1980s researchers have been motivated to
design and build teams of robots with the ability of working
together on some given task. This motivation stems from the
fact that in many applications, Multi-Robot Systems (MRS)
bring about several advantages over Single Robot Systems. In
particular, MRS are generally more time-efficient, less prone
to single-points of failure, and typically exhibit multiple
capabilities, which in many cases yield a more effective solu-
tion to a given problem. In early works, researchers observed
natural systems, such as a swarm of bees, ants and even
humans, to study how a group of individual entities can work
together to perform a given task. The multidisciplinary nature
of these early studies, eventually led to MRS being applied
in several different application domains such as surveillance,
search and rescue, foraging, exploration, cooperative manip-
ulation, and transportation of objects, among others. This
paper reviews a number of prominent and recent research
works that aim to address various problems appertaining to
six main application domains of MRS.
In Section II we compile a number of previous surveys
on MRS and categorise them in six broad categories, while
in Section III we present a number of recent research works
on MRS and organise them according to their application
domain. We believe that this new categorisation can be
very useful to researchers who are mainly interested in one
particular domain. Furthermore, this survey can also help
promote the migration of ideas across different application
domains of MRS. Finally in Section IV, we draw a number of
conclusions about the current and most prominent challenges
in the field of MRS.
*The research work disclosed in this publication is partially funded by the
ENDEAVOUR Scholarships Scheme 2014-2020, that may be part-financed
by the European Union.
1
Rachael N. Darmanin and Marvin K. Bugeja are with
the Department of Systems and Control Engineering, Fac-
ulty of Engineering, University of Malta, MSD2080, Malta
{rachael.darmanin,marvin.bugeja}@um.edu.mt
II. RELATED WORK
The majority of literature review papers on Multi-Robot
Systems (MRS) focus on classifying the most fundamental
aspects of an MRS, such as coordination and communication.
In [1] Farinelli et al. classify these MRS features into two di-
mensions. The first, termed the coordination dimension, deals
with the different classes of cooperation schemes, such as
whether the system is centralized or decentralized, strongly
cooperative (i.e. following a strict protocol), or weakly co-
operative, among others. The system dimension classifies the
existing types of communication schemes and team decom-
position attributes. Similarly, in [2] Parker classifies MRS
according to their architecture, the heterogeneity in the team,
the type of communication scheme adopted, and the different
types of task allocation schemes. In this work, Parker also
briefly reviews some works according to their application
domain. However, the latter is not an extensive literature
review of such works. A similar reviewing approach that
categorises works based on the foundation topics of MRS,
namely coordination, task allocation and cooperation, is also
adopted in [3]–[5]. Furthermore, in contrast to these works,
in [6] Gerkey and Matari
´
c focus on one particular aspect
of MRS, namely Multi-Robot Task Allocation (MRTA), and
propose a taxonomy of task allocation schemes. The same
taxonomy is also used in [7].
Alternatively, a literature review may also focus on par-
ticular schemes forming part of the main components in an
MRS. In [8], Bernardine Dias et al. focus on classifying what
they call market-based coordination approaches, which is a
type of MRTA scheme. Such coordination schemes require
the robots to bid for the tasks that they are able to perform.
Furthermore, detailed reviews in the area of swarm robotics
are properly surveyed in [9]–[11], while those reviewing
biologically-inspired research include [12]. We encourage the
interested reader to refer to these works for more details.
In recent years, considerable attention has been given to
human-robot interaction (HRI). In this light, Goodrich et al.
[13] provide an extensive review of the history of HRI, as
well as a review of works that introduced HRI to the field
of MRS. In this work the author identifies the main topics in
the field of HRI and provides a list of challenges that could
shape the future of this interesting research area. Similarly,
Chen et al. [14], propose a review of the human factors
in systems exhibiting HRI, focussing on human supervision
of multiple robots, and maintaining the human operator’s
situation awareness while retaining authority in the decision-
making process. Both Goodrich et al. and Chen et al. provide
indicators to the current open challenges in this area.
Moreover, some review papers focus on specific problems
in MRS, such as formation control of robots in a team.
Guanghua et al. [15] review such works in light of the type
of formation architecture, and the existing formation control
strategies employed. In the latter classification the authors
classify the reviewed approaches according to the follow-
ing categories: behaviour-based approaches, leader-follower
approaches, virtual structure approaches, artificial potential
functions or graph-based approaches. In [16], the authors
review the mathematical problem of flocking, together with
existing flocking control strategies. Other problem-specific
reviews focus on patrolling algorithms [17], robotic urban
search and rescue [18], and autonomous underwater vehicles
[11], [19]. The review papers referred to in this section
have been grouped in six broad categories in Table I. In
contrast to the survey works mentioned in this section, this
paper’s contribution lies in a review of recent and notable
works in MRS categorised according to the most prominent
application domains of the field.
III. APPLICATION DOMAINS
During the years a number of general application domains
in Multi-Robot Systems (MRS) have been proposed. For
instance, both Parker [2] and Farinelli et al. [1] list the
main application domains in the field of MRS. However, the
aim of these works is not to review prominent and recent
applications of MRS, but rather to provide a review of the
main technical aspects and challenges of the field. The scope
of this section is to address this gap, and classify recent
works on MRS according to their application domains.
A. Surveillance and, Search and Rescue
Surveillance and, search and rescue applications have at-
tracted considerable attention from the MRS community over
the years. This is due to the relevance of such applications in
daily life. Surveillance applications were initially reserved to
patrolling or surveying indoor areas [21]. However, with the
introduction of unmanned aerial vehicles (UAVs), researchers
have expanded their study to include the surveillance of
outdoor areas, such as areas far out at sea [22]. One of the
main challenges in these applications is that of persistent
surveillance due to the fact that one-time coverage and ex-
ploration algorithms cannot be used directly to continuously
patrol and monitor the same area [23]. Nigam et al. [23]
propose a novel control policy for persistent surveillance that
maintains optimal performance through a formally-derived
and scalable heuristic method. In this work, the environment
TABLE I
SUMMARY OF REVIEW PAPERS
MRS Main Principles [1]–[8]
Swarm Robotics [9]–[11]
Biologically-Inspired Works [12], [20]
Human-Multi-Robot Interaction [13], [14]
Problem-Specific Works [15]–[18]
Autonomous Underwater Vehicles [11], [19]
is divided into grid cells, where each cell is attributed with
an age and the goal is to minimize the overall age of all the
cells. A control policy called the Multi-Agent Reactive Policy
is proposed to control the UAVs performing surveillance. A
similar approach is adopted in [24] where the authors solve
the task of persistent surveillance through the Vehicle Routing
Problem.
Task allocation is another challenge in such applications
because the solution to this problem must be time-efficient.
In [25], the authors make use of a market-based strategy
where the robots bid for locations that must be surveyed. In
contrast, Jeon et al. [26] calculate costs for a set of tasks,
and allocate the mission tasks to the robots according to these
costs.
Over the years disasters such as the Fukushima nuclear
accident in 2011 have enabled researchers to deploy ad-
vanced MRS in real-life applications, mostly for search and
rescue. For example, in [22] the authors propose a cooper-
ative scheme for a multi-robot team for the surveillance of
shipwreck survivors at sea. Using satellite imagery the user
can plan the mission waypoints, which are then followed by
an Unmanned Surface Vehicle (USV) carrying a UAV. When
the USV arrives at the designated way-point, the UAV takes
off and uses a grid-like search pattern and image processing
to localize survivors.
Another example is that of Gregory et al. [27], who ad-
dress a humanitarian assistance and disaster relief applicaton.
In this work the authors focus on simultaneously evaluating
the damage done to the environment, and localising the
victims according to two types of goals, namely, goals
established from prior maps, and dynamic goals established
according to the sudden detection of victims. The novelty of
this work lies in addressing unreliable autonomy and com-
munication by modelling unknown travel costs in a dynamic
variant of the Capacitated Team Orienteering Problem [28].
Moreover, in [29], the authors also address the humani-
tarian relief problem through an implementation of a het-
erogeneous robotic system. This system features land and
marine mobile stations that are responsible for coordinating
and supporting UAVs and fast-speed land or marine robots.
The authors propose solutions to two coordination technical
challenges. The first entails the localization and landing of
the UAVs —achieved through the use of visual SLAM—
and UAV battery replacement to mitigate the limited energy
constraint, which is a very common problem in MRS. In
this work the authors also propose a collaboration scheme
for the team of UAVs based on dynamic communication,
target identification and triangulation.
B. Foraging and Flocking
The task of foraging is often synonymous with swarm
robotics, which is inspired by natural colonial systems such
as those of bees and ants. This is because very often, the
decentralized team only requires implicit communication
to cooperatively collect randomly distributed objects and
transport them to a “home” location. In [30] Parker proposes
ALLIANCE, a fault-tolerant framework that assigns tasks
to robots according to their motivation and capabilities —
modelled through a behaviour set—to do these tasks. This
framework was applied to the foraging task of cleaning up
hazardous waste, where the task allocation and coordination
among the team involved assigning the robots to move the
waste or report back to the base station. In another behaviour-
based approach [31], Schneider-Font
´
an and Matari
´
c assign
segments from a territory to each robot for clean-up and
object collection, as opposed to the task assignment used in
[30]. One challenge in this domain is the optimal sharing
of navigational space. In [32] Lein and Vaughan propose
an algorithm that reduces mutual spatial interference and
exploits a non-uniform distribution of robots during foraging.
To achieve this, the authors propose a technique named
adaptive bucket brigade foraging, where the robots remain
within a variable space distribution within the environment.
Furthermore, since foraging is a task associated with natural
systems, a number of works adopt particle swarm optimiza-
tion (PSO). Particularly, Couceiro et al. [33] study the robotic
Darwinian PSO under communication constraints in the team
[33]. From these works one can conclude that in general,
foraging is not implemented using complex explicit com-
munication schemes. The preferred choice of architecture is
often decentralized, in order to allow the team members to
achieve the task with minimal interference between them.
The task of flocking, also called shepherding, is considered
to be a trait of swarm robotics. Some works make use of
behaviour-based models and the generation of safe zones,
such that the members in the MRS can follow a direction and
stay in line with the flock to maintain cohesion, but at the
same time maintain enough distance between them to avoid
collision [34]. In these cases, a decentralized architecture
is often adopted. Moreover, in [35] Sakai et al. propose a
novel flocking algorithm that treats all detected objects as
obstacles, irrespective of whether they are truly obstacles
or form part of the team. They argue that such a method
limits the amount of information handled by the team since
velocity information on neighbouring robots is not required.
On the contrary, in [36] Gu and Wang propose a leader-
follower flocking technique, which requires the followers to
communicate with their neighbours to exchange information
about the estimated position of the flocking centre. Moreover,
in [37] the authors apply reinforcement learning together
with flocking control to enable a decentralized MRS to
learn how predators should be avoided while maintaining
the connectivity required for flocking. Additionally, in [38]
the authors propose a control algorithm that allows a flock
to navigate around obstacles. Similar to foraging, in flocking
we can see a trend in the use of a decentralized architecture.
However, the complexity of the task increases in flocking
applications since the team must not only coordinate to avoid
spatial interference, but also to maintain connectivity among
the entities.
C. Formation and Exploration
In formation applications, the team of robots must main-
tain some strict arrangement while at the same time avoid
obstacles in its path. This problem becomes more complex
than flocking since an obstacle must be collectively avoided
without any of the team members leaving the formation for
a long while. A common solution is the leader-follower
approach where a trajectory-planning algorithm is imple-
mented on the leader robot and formation constraints—
distances from the leader—restrict the followers to maintain
formation around the leader [39]. Recent research is also
adopting computational intelligence methods in formation
applications. In [40], Wang et al. solve the optimal formation
problem by using a recurrent neural network. Shape theory is
used to generate a set of feasible formations and the proposed
optimal formation solution chooses the one that has the
minimum distance from the initial formation. Alternatively,
the work reported in [41] adopts fuzzy logic to achieve
formation control. Additionally, the work in [42] adopts
control schemes, such as Model Predictive Control, in order
to establish formation in the team.
In contrast to formation, in exploration the robots in a
team must distribute themselves in an unknown environment
in order to explore the area. The coordination of such
a system brings about many challenges, particularly those
related to connectivity and battery-life problems. In [43]
Banfi et al. address the problem of communication constraint
by proposing an exploration strategy under recurrent connec-
tivity. The robots only need to connect to the base station if
new observations are made, hence allowing the members to
disconnect for long periods of time until new information is
obtained. Moreover, Cesare et al. [44] address the problems
of communication and battery life. They propose a state-
based approach in which a robot explores and shares infor-
mation only if it is within the communication range and its
battery levels are above a certain threshold. If the robot does
not have enough energy to continue exploring, it waits for
another robot to meet it, and then uses its remaining energy
to relay the information to base. This rendezvous solution to
overcome communication limitations is also used in [45].
Other works in this domain focus on exploration strategies
that are tailor-made for a particular MRS. For instance, in
[46] the authors propose a circle partitioning method that
segments the environment into sections and assigns each
robot to a particular sub-region. In [47] the authors use a
flooding algorithm that aims at reducing the exploration time
and minimize the overall distance travelled by the robots
during exploration. A number of works even propose explo-
ration strategies that emerge from a graph-based approach,
such that optimal coordination can be achieved when having
a known number of robots exploring an area [48].
D. Cooperative Manipulation
The box-pushing problem has become synonymous with
MRS and it has been studied in several early works [49]–
[51]. In one of these works, namely that by Brown and
Jennings [52], the authors implement a pusher/steerer system
where the object is moved from one place to another by
small mobile robots. The steerer robot is pre-programmed
with a trajectory and the pusher robot exerts a force onto the
object, such that the steerer can follow its programmed path
by setting its heading. Hence, communication is evidently
absent from such a system. Alternatively, in [53] Sieber et al.
propose a novel linear state feedback controller to surround
an object with a formation of robots in order to transport
said object. In this work the authors propose a suboptimal
control law similar to the linear quadratic regulator (LQR)
approach, which is used to regulate the way that a group
of robots form an assembly around an object to transport it.
Other works which adopt this formation control for object
manipulation and transportation include [54] and [55].
More recently, Amanatiadis et al. [56] propose the system
AVERT which is used to extract and transport vehicles from
a specific location. The novelty of this work is focussed
on lifter mobile robots used in a system that applies the
concept of trajectory planning (using the D* Lite method),
obstacle detection, and makes use of intercommunication in
order to exchange control and trajectory information with
a command base. As opposed to the previously mentioned
works, the approach proposed in the AVERT project requires
a stable communication among the team members, which
is often missing in traditional pusher-steerer or formation-
based cooperative manipulation algorithms. In view of the
solutions presented in these works, one may also think of
foraging as another solution to cooperative object transporta-
tion and manipulation, since this involves the collection and
transportation of items to some base location.
E. Team Heterogeneity
Heterogeneity in a team of mobile robots enables the
team to handle complex tasks more efficiently and effectively
by exploiting the benefits of the diverse capabilities of its
members. In this work, we shall analyse heterogeneity from
two perspectives, namely, human-robot heterogeneity and
heterogeneous robot teams.
A human-robot interaction (HRI) may be present in a
system where the human is supervising and commanding
a team of multiple robots. In [57] Rossi et al. propose
a scheme where the human operator communicates with
a whole team of robots to assign tasks and specify the
members which must perform each task through speech
utterances. In this case this study focuses on how speech
may be segmented to simultaneously address multiple robotic
recipients. Similarly, Cacace et al. [58] apply an HRI to a
search and rescue application. In their proposed scheme they
exploit human gestures and speech to select a desired robot
for a task in a nonverbal and implicit manner. Robot selection
is estimated by a probability that a particular command
given by the user infers a set of capabilities which the
robot possesses. For instance, the command “take off” shall
probably be directed to an aerial vehicle and therefore, given
more information, the algorithm proposed by Cacace et al.
can select which aerial vehicle needs to take off. In [59],
the authors also propose a scheme that seeks to establish an
effective cooperation between a human and a team of robots
during navigation.
The element of heterogeneity in a robot team is even
evident in the earlier works of MRS. The framework AL-
LIANCE, proposed by Parker [30], supports and exploits the
inherent differences in a team of diverse robot platforms.
This is achieved by tailoring the motivational behaviour
model and behaviour sets according to the capabilities of
the robots. Similarly, Gerkey and Matari
´
c [60] propose a
market-based planner system named MURDOCH for task
allocation in a robotic team. In MURDOCH task messages
are published over a network with a subject relating to the
capabilities required from a robot to perform that task, leav-
ing only those capable robots to subscribe to these messages.
In the framework ASyMTRe [61], Tang and Parker propose
an algorithm that decomposes a general task into sub-tasks
according to pre-defined schema which reflect the different
capabilities of the robots. These schema are then connected
to assign the subtasks to the team members. Furthermore,
Jeon et al. [26] propose a scheme with leader/follower
roles for robots which are meant to survey an area and act
in the event that an intruder is detected. These roles are
defined according to the ability of the robot. A mission is
decomposed into tasks, which are then assigned to the robots
whose capabilities make them the most adequate to perform
them. Similarly, in [58] and [62] we see search and rescue
MRSs, where robots with different capabilities, such as to
provide aerial views [58], or extinguish fires and transporting
victims [62], are assigned to specific tasks. Additionally, the
element of heterogeneity is strong in ground-to-aerial vehicle
cooperation as seen in [63]–[65].
F. Adversarial Environment
In 1997 RoboCup was founded with the aim of promoting
research in robotics and artificial intelligence. This compe-
tition has shifted some of the research attention onto MRS
in adversarial environments, such as those found in soccer
competitions or battlefields. In [66] Weigel et al. propose
a novel approach that tracks the ball and the adversarial
players, and at the same time it strategically coordinates
the team. In the mission to win against the other team, a
specific role requiring a number of skills, is assigned to each
robot such that, each team member can then adopt adequate
behaviour from a behaviour set. A similar approach is also
adopted in [67], where Browning et al. propose a hierarchical
architecture of Skills, Tactics and Plays to execute low-level
actions, decide on the skills to use, and coordinate the activity
among the team. Furthermore, this problem has also been
solved by using reinforcement learning techniques, as has
been done in [68] and [69]. Alternatively, MRS have also
been involved in different adversarial environments, such
as in battlefields. For instance, in [70] Zhang et al. use a
genetic algorithm to enable robots to learn not to enter their
adversaries’ defence region, where they may be “killed”.
Similarly, this problem has also been studied and solved
in [71] and [72]. Using a somewhat different approach, in
[73] the authors look at the adversarial environment as being
inherent to an auction system, where each member bids for
a task to perform in a mission.
IV. CONCLUSIONS
In this paper, we have reviewed and classified prominent
and recent works according to six main application domains
in the field of MRS. From this survey of works it is
possible to draw up a number of conclusions about the
current challenges facing this field and its research direc-
tion. Primarily, it is clear that in most application domains,
limited communication among the team members, is one
of the major difficulties that needs to be solved in order
to design an efficient and robust system. Another pressing
issue, especially in multi-UAV applications, seems to be
energy consumption and limited battery life. Researchers are
proposing methods on how this constraint is incorporated in
the solution of a particular task. Finally, the introduction of
a human-in-the-loop is another promising research avenue to
address a number of challenging real-life scenarios. However,
this creates new challenges of its own, particularly due to the
complex interaction that exists between a human being and
a machine, which is naturally more pronounced in a human-
to-many-machines applications.
REFERENCES
[1] A. Farinelli, L. Iocchi, D. Nardi, and A. Multirobot, “Multirobot Sys-
tems : A Classification Focused on Coordination,” IEEE Transactions
on Systems, Man, and Cybernetics Part B: Cybernetics, vol. 34, no. 5,
pp. 2015–2028, 2004.
[2] L. E. Parker, “Multiple Mobile Robot Systems,” in Handbook of
Robotics, 2008, ch. 40, pp. 921 – 941.
[3] Y. Cai and S. X. Yang, “A Survey on Multi-robot Systems,” in World
Automation Congress 2012, 2012, pp. 1–6.
[4] Y. Zhi, N. Jouandeau, and A. A. Cherif, “A Survey and Analysis
of Multi-Robot Coordination,” International Journal of Advanced
Robotic Systems, INTECH, vol. 10, no. 12, p. 399, 2013.
[5] Y. Cao, W. Yu, W. Ren, and G. Chen, “An overview of recent
progress in the study of distributed multi-agent coordination,” IEEE
Transactions on Industrial Informatics, vol. 9, no. 1, pp. 427–438,
2013.
[6] B. P. Gerkey and M. J. Mataric, “A formal analysis and taxonomy
of task allocation in multi-robot systems,” International Journal of
Robotics Research, vol. 23, no. 9, pp. 939–954, 2004.
[7] A. Khamis, A. Hussein, and A. Elmogy, Multi-robot Task Allocation:
A Review of the State-of-the-Art. Springer, 2015.
[8] M. Bernardine Dias, R. Zlot, N. Kalra, and A. Stentz, “Market-based
multirobot coordination: A survey and analysis,” in Proceedings of the
IEEE, vol. 94, no. 7, 2006, pp. 1257–1270.
[9] M. Brambilla, E. Ferrante, and M. Birattari, “Swarm robotics : A re-
view from the swarm engineering perspective,” in Swarm Intelligence,
2012, vol. 7, no. 1, pp. 1–41.
[10] M. Dorigo, D. Floreano, L. M. Gambardella, F. Mondada, S. Nolfi,
T. Baaboura, M. Birattari, M. Bonani, M. Brambilla, A. Brutschy,
D. Burnier, A. Campo, A. L. Christensen, A. Decugniere, G. Di
Caro, F. Ducatelle, E. Ferrante, A. Forster, J. M. Gonzales, J. Guzzi,
V. Longchamp, S. Magnenat, N. Mathews, M. Montes De Oca,
R. O’Grady, C. Pinciroli, G. Pini, P. Retornaz, J. Roberts, V. Sperati,
T. Stirling, A. Stranieri, T. Stutzle, V. Trianni, E. Tuci, A. E. Turgut,
and F. Vaussard, “Swarmanoid: A novel concept for the study of
heterogeneous robotic swarms,” IEEE Robotics and Automation Mag-
azine, vol. 20, no. 4, pp. 60–71, 2013.
[11] B. T. Champion and M. A. Joordens, “Underwater swarm robotics
review,” 10th System of Systems Engineering Conference, SoSE 2015,
pp. 111–116, 2015.
[12] P. Agrawal, “A Review on Multi Robot Cooperation Using Bio
Inspired Neural Networks,” International Journal of Soft Computing,
Mathematics and Control (IJSCMC), vol. 2, no. 4, pp. 15–24, 2013.
[13] M. A. Goodrich and A. C. Schultz, “Human-Robot Interaction: A
Survey,” Foundations and Trends in Human-Computer Interaction,
vol. 1, no. 3, pp. 203–275, 2007.
[14] J. Y. C. Chen and M. J. Barnes, “Human Agent Teaming for
Multirobot Control : A Review of Human Factors Issues,” IEEE
Transactions on Human-Machine Systems, vol. 44, no. 1, pp. 13–29,
2014.
[15] W. Guanghua, L. Deyi, G. Wenyan, and J. Peng, “Study on Formation
Control of Multi-Robot Systems,” in 3rd International Conference on
Intelligent System Design and Engineering Applications, 2013, pp.
1335–1339.
[16] Y. Mao, G. G. Yan, and Y. T. Tian, “A review of studies in flocking
for multi-robot system,” in International Conference on Computer,
Mechatronics, Control and Electronic Engineering, CMCE 2010,
vol. 4, 2010, pp. 28–31.
[17] D. Portugal and R. Rocha, “A Survey on Multi-robot Patrolling Algo-
rithms,” in Technological Innovation for Sustainability: Second IFIP
WG 5.5/SOCOLNET Doctoral Conference on Computing, Electrical
and Industrial Systems, DoCEIS 2011, L. M. Camarinha-Matos, Ed.
Springer Berlin Heidelberg, 2011, pp. 139–146.
[18] Y. Liu and G. Nejat, “Robotic urban search and rescue: A survey from
the control perspective,” pp. 147–165, 2013.
[19] D. Cook, A. Vardy, and R. Lewis, “A survey of AUV and robot
simulators for multi-vehicle operations,” in IEEE/OES Autonomous
Underwater Vehicles, AUV 2014, 2014, pp. 1 – 8.
[20] J.-H. Son and H.-S. Ahn, “Cooperative Reinforcement Learning: Brief
Survey and Application to Bio-insect and Artificial Robot Interaction,”
in Science And Technology, 2008, pp. 71–76.
[21] A. Stancovici, M. V. Micea, and V. Cretu, “Cooperative Positioning
System for Indoor Surveillance Applications,” in International Con-
ference on Indoor Positioning and Indoor Navigation, (IPIN), no. 1,
2016, pp. 1–7.
[22] R. Mendonc¸a, M. M. Marques, F. Marques, A. Lourenc¸o, E. Pinto,
P. Santana, F. Coito, V. Lobo, and J. Barata, “A Cooperative Multi-
Robot Team for the Surveillance of Shipwreck Survivors at Sea,” in
MTS/IEEE Monterey OCEANS, 2016, pp. 2–7.
[23] N. Nigam, S. Bieniawski, I. Kroo, and J. Vian, “Control of multiple
UAVs for persistent surveillance: Algorithm and flight test results,”
IEEE Transactions on Control Systems Technology, vol. 20, no. 5, pp.
1236–1251, 2012.
[24] E. Stump and N. Michael, “Multi-robot persistent surveillance plan-
ning as a vehicle routing problem,” in IEEE International Conference
on Automation Science and Engineering, 2011, pp. 569–575.
[25] A. Pustowka and E. F. Caicedo, “Market-based task allocation in a
multi-robot surveillance system,” in Brazilian Robotics Symposium
and Latin American Robotics Symposium, SBR-LARS 2012, 2012, pp.
185–189.
[26] S. Jeon, M. Jang, S. Park, D. Lee, Y. J. Cho, and J. Kim, “Task allo-
cation strategy of heterogeneous multi-robot for indoor surveillance,”
in 8th International Conference on Ubiquitous Robots and Ambient
Intelligence, 2011, pp. 169–173.
[27] J. Gregory, J. Fink, E. Stump, J. Twigg, J. Rogers, N. Fung, and
S. Young, “Application of Multi-Robot Systems to Disaster-Relief
Scenarios with Limited Communication,” Springer Tracts in Advanced
Robotics, vol. 113, pp. 639–653, 2016.
[28] C. Archetti, D. Feillet, A. Hertz, and M. G. Speranza, “The capaci-
tated team orienteering and profitable tour problems,” Journal of the
Operational Research Society, vol. 60, no. 6, pp. 831–842, 2009.
[29] M. A. Gutierrez, S. Nair, R. E. Banchs, L. F. D. H. Enriquez, A. I.
Niculescu, and A. Vijayalingam, “Multi-robot collaborative platforms
for humanitarian relief actions,” in IEEE Region 10 Humanitarian
Technology Conference, (R10-HTC), 2016.
[30] L. E. Parker, “ALLIANCE : An Architecture for Fault Tolerant Multi-
robot Cooperation,” IEEE Transactions on Robotics and Automation,
vol. 14, no. 2, pp. 220–240, 1998.
[31] M. Schneider-Font
´
an and M. J. Matari
´
c, “Territorial multi-robot task
division,” IEEE Transactions on Robotics and Automation, vol. 14,
no. 5, pp. 815–822, 1998.
[32] A. Lein and R. T. Vaughan, “Adapting to non-uniform resource
distributions in robotic swarm foraging through work-site relocation,”
IEEE/RSJ International Conference on Intelligent Robots and Systems,
(IROS), pp. 601–606, 2009.
[33] M. S. Couceiro, R. P. Rocha, C. M. Figueiredo, J. M. A. Luz, and
N. M. Fonseca Ferreira, “Multi-Robot Foraging based on Darwin’s
Survival of the Fittest,” in IEEE/RSJ International Conference on
Intelligent Robots and Systems, 2012, pp. 801–806.
[34] F. J. Mendiburu, M. R. A. Morais, and A. M. N. Lima, “Behavior co-
