Professional Documents
Culture Documents
Textbook Swarm Intelligence 11Th International Conference Ants 2018 Rome Italy October 29 31 2018 Proceedings Marco Dorigo Ebook All Chapter PDF
Textbook Swarm Intelligence 11Th International Conference Ants 2018 Rome Italy October 29 31 2018 Proceedings Marco Dorigo Ebook All Chapter PDF
Textbook Swarm Intelligence 11Th International Conference Ants 2018 Rome Italy October 29 31 2018 Proceedings Marco Dorigo Ebook All Chapter PDF
https://textbookfull.com/product/swarm-intelligence-12th-
international-conference-ants-2020-barcelona-spain-
october-26-28-2020-proceedings-marco-dorigo/
https://textbookfull.com/product/swarm-intelligence-10th-
international-conference-ants-2016-brussels-belgium-
september-7-9-2016-proceedings-1st-edition-marco-dorigo/
https://textbookfull.com/product/cyberspace-safety-and-
security-10th-international-symposium-css-2018-amalfi-italy-
october-29-31-2018-proceedings-arcangelo-castiglione/
https://textbookfull.com/product/discovery-science-21st-
international-conference-ds-2018-limassol-cyprus-
october-29-31-2018-proceedings-larisa-soldatova/
Decision and Game Theory for Security 9th International
Conference GameSec 2018 Seattle WA USA October 29 31
2018 Proceedings Linda Bushnell
https://textbookfull.com/product/decision-and-game-theory-for-
security-9th-international-conference-gamesec-2018-seattle-wa-
usa-october-29-31-2018-proceedings-linda-bushnell/
https://textbookfull.com/product/foundations-of-intelligent-
systems-24th-international-symposium-ismis-2018-limassol-cyprus-
october-29-31-2018-proceedings-michelangelo-ceci/
https://textbookfull.com/product/static-analysis-25th-
international-symposium-sas-2018-freiburg-germany-
august-29-31-2018-proceedings-andreas-podelski/
https://textbookfull.com/product/computational-logistics-9th-
international-conference-iccl-2018-vietri-sul-mare-italy-
october-1-3-2018-proceedings-raffaele-cerulli/
https://textbookfull.com/product/similarity-search-and-
applications-11th-international-conference-sisap-2018-lima-peru-
october-7-9-2018-proceedings-stephane-marchand-maillet/
Marco Dorigo · Mauro Birattari
Christian Blum · Anders L. Christensen
Andreagiovanni Reina · Vito Trianni (Eds.)
LNCS 11172
Swarm Intelligence
11th International Conference, ANTS 2018
Rome, Italy, October 29–31, 2018
Proceedings
123
Lecture Notes in Computer Science 11172
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology Madras, Chennai, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7407
Marco Dorigo Mauro Birattari
•
Swarm Intelligence
11th International Conference, ANTS 2018
Rome, Italy, October 29–31, 2018
Proceedings
123
Editors
Marco Dorigo Anders L. Christensen
Université Libre de Bruxelles University of Southern Denmark
Brussels Odense
Belgium Denmark
Mauro Birattari Andreagiovanni Reina
Université Libre de Bruxelles University of Sheffield
Brussels Sheffield
Belgium UK
Christian Blum Vito Trianni
Artificial Intelligence Research Institute National Research Council
Bellaterra Rome
Spain Italy
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
These proceedings contain the papers presented at ANTS 2018, the 11th International
Conference on Swarm Intelligence, held at the National Research Council (CNR) in
Rome, Italy, during October 29–31, 2018. The ANTS series started in 1998 with the
First International Workshop on Ant Colony Optimization (ANTS 1998). Since then,
ANTS, which is held bi-annually, has gradually become an international forum for
researchers in the wider field of swarm intelligence. In 2004, this development was
acknowledged by the inclusion of the term “Swarm Intelligence” (next to “Ant Colony
Optimization”) in the conference title. Since 2010, the ANTS conference is officially
devoted to the field of swarm intelligence as a whole, without any bias towards specific
research directions. This is reflected in the title of the conference: “International
Conference on Swarm Intelligence.”
This volume contains the best papers selected out of 69 submissions. Of these, 24
were accepted as full-length papers, while 12 were accepted as short papers. This
corresponds to an overall acceptance rate of 52%. Also included in this volume are 7
extended abstracts.
All the contributions were presented as posters. The full-length papers were also
presented orally in a plenary session. Extended versions of the best papers presented at
the conference will be published in a special issue of the Swarm Intelligence journal.
We would like to take this opportunity to thank the large number of people that were
involved in making this conference a success. We would like to express our gratitude to
the authors who contributed their work, to the members of the International Program
Committee, to the additional referees for their qualified and detailed reviews, and to the
staff at the Institute of Cognitive Sciences and Technologies (ISTC) of the CNR for
helping with organizational matters.
We hope the reader will find this volume useful both as a reference to current
research in swarm intelligence and as a starting point for future work.
Organizing Committee
General Chair
Marco Dorigo Université Libre de Bruxelles, Belgium
Vice-General Chair
Mauro Birattari Université Libre de Bruxelles, Belgium
Publication Chair
Andreagiovanni Reina University of Sheffield, UK
Program Committee
Michael Allwright Université Libre de Bruxelles, Belgium
Prasanna Balaprakash Argonne National Laboratory, USA
Jacob Beal BBN Technologies, USA
Giovanni Beltrame Polytechnique Montréal, Canada
Tim Blackwell Goldsmiths, University of London, UK
Mohammad Reza Bonyadi The University of Adelaide, Australia
Darko Bozhinoski Université Libre de Bruxelles, Belgium
Alexandre Campo Université Libre de Bruxelles, Belgium
Marco Chiarandini University of Southern Denmark, Denmark
Maurice Clerc Independent Consultant on Optimisation, France
Carlos Coello Coello CINVESTAV-IPN, Mexico
Oscar Cordon University of Granada, Spain
Nikolaus Correll University of Colorado Boulder, USA
Guido De Croon Delft University of Technology, The Netherlands
VIII Organization
Additional Reviewers
Nicolas Cambier Université de Technologie de Compiègne, France
Yue Gu University of Sheffield, UK
Bahar Haghighat Ecole Polytechnique Fédérale de Lausanne,
Switzerland
Matthew Hall University of Sheffield, UK
Marcos Oliveira Leibniz Institute for the Social Sciences, Germany
Anil Ozdemir University of Sheffield, UK
Diego Pinheiro Florida Institute of Technology, USA
Judhi Prasetyo Middlesex University Dubai, UAE
Leonardo Stella University of Sheffield, UK
Contents
Full Papers
Short Papers
Does the ACOR Algorithm Benefit from the Use of Crossover? . . . . . . . . . . 342
Ashraf M. Abdelbar and Khalid M. Salama
Individual Activity Level and Mobility Patterns of Ants Within Nest Site . . . 378
Kazutaka Shoji
Extended Abstracts
1 Introduction
Cooperative transportation without explicit communication, but based only on
on the indirect exchange of information through the physical interaction with
the environment is a very important feature for social animals. From a scientific
point of view, the problem has been addressed as Stigmery theory [8]; later, it
has been regarded as the main coordinating mechanism in groups of ants for
object transportation [9] (see Fig. 1 on the left) and has been indeed observed
and studied in [3,5,6,13].
Such skills observed in nature have inspired researchers to transfer them to
swarms of simple robotic agents. In fact, avoiding explicit communication would
reduce hardware and software complexity, and overcome possible communica-
tion failure issues. So far, the interest has been mainly focused on terrestrial
systems [3,9,17,22], where the possibility of decentralized transportation based
on physical interactions has been proved. Instead, in this work, we are interested
in the communication-less aerial transportation of objects by swarms. This is
particularly interesting not only from a scientific point of view, thanks to the
c Springer Nature Switzerland AG 2018
M. Dorigo et al. (Eds.): ANTS 2018, LNCS 11172, pp. 3–15, 2018.
https://doi.org/10.1007/978-3-030-00533-7_1
4 C. Gabellieri et al.
Fig. 1. Left: red ants in cooperative transportation. Right: the tension in spider webs
influences their natural frequency [2].
higher complexity of the problem, but also because a simple, robust and scalable
object aerial manipulation technique could meet the requirements demanded by
many real applications. Aerial transportation can benefit from a larger workspace
and the independence from uneven terrains. However, aerial robots, e.g., multi-
rotors, though agile and low cost, are typically characterized by a limited pay-
load. Hence, a cooperative approach is a very suitable solution. Some examples
of applications can be found in industrial contexts, in agriculture, and in search
and rescue missions to carry necessary equipment or first aid.
Considering the lack of results for aerial systems compared to grounded ones,
our purpose is to start filling this gap. In particular, we investigate the possibil-
ity of a communication-less approach for cooperative aerial manipulation with a
swarm of flying agents. Furthermore, we investigate if and how the load and the
cable physically connecting the robots may play the role of an implicit commu-
nication channel exploiting the forces exchanged. To the best of our knowledge,
this is the first work proposing a bio-inspired algorithm for communication-less
aerial manipulation by flying swarms, going beyond the two-robot-‘only’ sce-
nario, recently considered in [7,20,21].
The algorithm that we propose exploits a leader-follower paradigm where the
leader agent knows where to go and hence steers the object toward the desired
position. On the other hand, the follower agents follow the leader and help to
sustain the weight and manipulate the load, exploiting only the implicit infor-
mation contained in the force received from the load itself. It has been observed
that also groups of army ants Eciton burchelli [5] and Dorylus wilverthi [6] adopt
a distinct caste distribution in transportation groups, in the sense that groups
have a significant tendency to contain only one submajor, i.e., a particular type
of ant. These species of army ants have proved to be very efficient in transporting
objects together. Additionally, in the same works, it has been noticed that it is
usually a single ant, the submajor, that starts the motion of the object, and then
the rest of the group moves accordingly. Such a behavior is actually replicated
in the leader-follower paradigm proposed in this paper.
While animals usually deploy items by directly touching them (direct manip-
ulation), in our framework we have chosen an indirect manipulation technique
A Study on Force-Based Collaboration in Flying Swarms 5
of the load through cables. This choice is motivated by different reasons and
it has been proved to be a very effective solution for cooperative aerial trans-
portation [10–12,19,23]. First of all, a cable attached to the agent center of mass
allows to minimize the coupling between the rotational dynamics of the agent
itself from the rest of the system dynamics. This is particularly useful for aerial
agents that are underactuated – the most common case – since they need to
change their attitude to be able to apply forces in any direction. Furthermore,
compared to other possible decoupling gripping mechanisms, as the ones in [14],
cables are simpler, low cost, and in general lighter. In this paper we demonstrate
through several numerical simulations that a swarm of N flying agents is capable
of collaborative manipulation skills based only on implicit communication. We
show that a twofold major role is played by the internal force applied to the
transported object. Internal forces are forces applied at the contact points on
the object that stretch or compress it without producing any movement, since
they counterbalance each other. The condition of zero internal force corresponds
to the case in which the agents transport the object while keeping the cables
vertical and applying only a force that compensates for the gravity. Firstly, we
have found that nonzero internal forces allow the swarm to univocally set the
attitude of the commonly transported object, and secondly, that larger internal
forces reduce the convergence time of the overall system to such unique equilib-
rium. This creates an interesting analogy with the role of tension in spider webs,
see Fig. 1-right. The breadth of analysis covers also the thorough investigation
of the leader forces depending on the swarm parameters and the analysis of the
benefits of a saturated nonlinear law for the leader force in order to tradeoff
compliance/safety and transportation accuracy.
The paper structure is the following: Sect. 2 illustrates the dynamic model
of the system. Then, we formulate the hypotheses regarding the properties of
the swarm, supported by the numerical results presented in Sect. 3. A thorough
discussion follows in Sect. 4. Final conclusions and future developments are pre-
sented in Sect. 5.
2 Model
is the ‘control’ force of each agent. To better understand how to control a multi-
rotor aerial robot in such a way, the reader can refer to [21]. Such control
force models three simple actions. First, the agent implements a spring like
action to move towards a desired position or follow a desired path (the apex d
indicates the ‘desired’ quantities). Secondly, the agent implements a dissipative
derivative term proportional to the velocity error. This action damps the oscil-
lations induced by the spring action. Finally, there is a force bias indicated with
πi = πi,x πi,y πi,z . This bias is essential to make the flying agents sustain
the weight of the load. We shall show that it plays an important role also for
shaping the system equilibria. Through πi it is possible to set reference internal
forces (forces that do not result in a motion of the object).
The static equilibrium equation of the object subject to the gravity vector
g ∈ R3 is given by [mL g 01×3 ] = W f, ¯ where W ∈ R6×3N is the grasp matrix,
that maps the forces at the contact points to a wrench applied at the object
center of mass and f¯ ∈ R3N ×1 collects the equilibrium cable forces. Resolving
the equation for f¯ we obtain:
3 Numerical Study
For a system of only two aerial agents and a beam/like load, in [21] we formally
proved, using a Lyapunov-based approach, the stability or instability of all the
possible equilibria and the passivity of the overall controlled system. However,
it is not trivial to extend those theoretical results to N > 2 and to a more
generally shaped object. In particular, it is not straightforward to solve the so
called equilibria inverse problem in [21], namely to find all the possible positions
of the agents and forces in the cables for each stable pose of the load. The authors
in [1] showed that such problem is very complex to solve even if just less than
six cables of assigned length are used. Analytical answers to the problem are
therefore difficult to reach. However, a numerical study of whether some of the
properties discovered in [21] for the two-agent system apply also to the swarm
case, where the number of follower agents may be arbitrary large and the object
not only a bar, is equally interesting. Thus, in this section we extrapolate some
conjectures on the expected behavior of the swarm system and validate these
hypotheses through a wide numerical study.
Goal of the study Scenario Internal forces CoM Points Li Unknown parameters
Internal force role Sa t = 0 [N] G ∈ I On a circle None
Sb t = 0 [N] G∈
/ I On a circle None
Sc t = 0 [N] G∈
/ I On a circle None
Parametric uncertainty Sd t = 0 [N] G∈
/ I Random mL , N (only bounds)
Table 1 contains the description of the simulated conditions. The load has
mass mL = 5 kg and inertial matrix equal to JL = I3 kgm2 . The leader agent,
set as the agent 1, is fed with a sufficiently smooth reference trajectory consti-
tuted by a 5th order polynomial in 3D (rest to rest trajectory with zero acceler-
ation at start and end points), which lasts 10 s and covers 2 m. M1 = 0.5I3 kg,
Mi = 0.01I3(N −1) kg, B1 = 100I3 Nsm−1 Bi = 1.5I3(N −1) Nsm−1 , K1 =
1000I3 Nsm−1 , and Ki = 0I3(N −1) Nm−1 for i = 1. When N > 30, we changed
the apparent mass Mi , and the damping Bi of the followers, i.e., for i = 2, ..., N
reducing both by 90%. This allows the leader agent to drag the system without
8 C. Gabellieri et al.
1 1
0.5
0.5
0
0
-0.5
-0.5
-1
5 -1
6 5 6
4 4
0 2 2
0 0 0
-2 -2
-5 -4 -5 -4
-6 -6
(a) t = 0 (b) t = 0
1.5
1
0.5 1
0
0.5
-0.5
-1 0
6 5
4
2
0 0 -0.5
-2
5
-4 -5 0
-6 2 4 6
-5 -4 -2 0
-6
(c) cables randomly attached on the (d) Simulated system during trans-
surface of the object portation
Fig. 2. Simulated system for the case N = 6 ((a), (b), (c)) and N = 60 (d)). The
object is depicted as a grey surface, and the light spot on it coincide with its center of
mass G. In case G ∈ / I, the light spot is instead the projection of G on I. The cables
are the black lines, while the circles represent the robots (the darker one is the leader).
5 0.5
40
0 0
30 -5 -0.5
20 -10 -1
-15 -1.5
10
-20 -2
0 5 10 15 0 10 20 30 40 50 0 10 20 30 40 50
45 5
40 0
35 0 -0.2
30 -0.4
25 -5 -0.6
20 -0.8
15 -10 -1
10 -1.2
5 -15 -1.4
0 10 20 30 40 50 0 50 100 150 200 250 300 350 0 50 100 150
0.3
40
35 0 0.2
30 -1
0.1
25
-2
0
20
-3
15 -0.1
-4
10
-0.2
5 -5
0 -0.3
2 4 6 8 10 12 14 16 18 0 10 20 30 40 50 60 0 5 10 15 20
Fig. 3. Object attitude during transportation. First row: scenario Sa : fint = 0 N and
G ∈ I. Second row: scenario Sb : fint = 0 N and G below I. Third row: scenario Sc :
fint = 0.8N and G placed below I. The dotted line is the desired value. The first,
second, and third columns refer to the trajectories of the yaw (ψ), pitch (θ), and roll
(φ) angles of the object, respectively.
final orientation of the load for the scenario Sa , and the second line of Fig. 3 an
arbitrary final yaw for the scenario Sb .
In the scenario Sc we set the force bias πi in (2) so that the system reaches
an equilibrium with the cables forces applying non zero horizontal forces. More
in detail, indicating with ¯li the vector that connects the projection of G on I to
the position of point Li at the final attitude, we set [πi,x πi,y ] = fint ¯li /||¯li ||,
where fint is thus the intensity of each agent’s planar force bias. In other words,
the desired horizontal forces in the cables are oriented radially and outward the
object in the final configuration, similarly to what is depicted in Fig. 2(b). Since
the object reaches an equilibrium, and the external wrench on the object at
the equilibrium is only the vertical force due to the gravity, as in (3), such non
vertical components of the cable forces, which do not cause any motion of the
object and do not compensate any external wrench, generate an internal force
t = 0. In this way, the object reaches always the same attitude: zero pitch,
zero roll, and the same yaw. The third line of Fig. 3 shows the results of such
simulations. Notice that the same results have been obtained even when G ∈ I.
10 C. Gabellieri et al.
2 2
Fig. 4. Intensities of the vertical force fi,z and of the horizontal force fi,x + fi,y that
the agents apply to the object during the transportation for four different parameters
of the followers in scenario Sc with N = 70. The dotted line refers to the leader agent,
while the solid lines refer to the followers. In the subcaptions i = 2, . . . , N .
Figure 4 shows the evolution of the forces that the agents apply to the load in
four different cases belonging to scenario Sc .
We also decided to see whether the value of fint influences the speed of
convergence with which the swarm stabilizes the object in scenario Sc . Figure 5
shows the evolution of the load attitude when transported by a group of five
agents for different values of fint .
We present now the results concerning scenario Sd , where neither the exact
real mass of the load, mL , nor the exact total number of agents, N , are known by
the agents. Only some upper and lower bounds are given, indicated with mmax ,
mmin , Nmax and Nmin respectively. Therefore, we introduced a particular choice
of πi .
⎡ ⎤ ⎧
0 ⎪
⎨−KZ if |fi,z | > fmax
πi = ⎣ 0 ⎦ where Ui =
z
KZ if |fi,z | < fmin . (4)
⎪
⎩
fi,z + Ui
z
0 otherwise
A Study on Force-Based Collaboration in Flying Swarms 11
0.4 25 2
0
0.2 20
-2
0
15 fint = 2 N -4
fint = 2 N fint = 2 N
-0.2 fint = 4 N
fint = 4 N 10 -6 fint = 4 N
-0.4 fint = 6 N
fint = 6 N -8 fint = 6 N
fint = 0.5 N
fint = 0.5 N 5
-0.6
-10 fint = 0.5 N
-0.8 0 -12
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100
Fig. 5. Evolution of the object attitude (yaw, pitch and roll, respectively) during trans-
portation for different values of the intensity of the internal force in scenario Sc with
N = 5.
25 10 20
20 0
10
15 -10
0
10 -20
-10
5 -30
-20
0 -40
-5 -50 -30
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
Fig. 6. Attitude of the load (yaw, pitch and roll, respectively) during the transportation
task in scenario Sd with random cables attachment points and uncertain load mass and
agents number.
Namely, we are defining a dead-zone in the sensed vertical force such that if
the robots perceive a vertical force in the cable that is inside a certain range
defined by fmax and fmin , they ignore it; otherwise, they apply an upward or
downward force trying to restore a vertical force inside the predefined range.
KZ is a constant that determines the responsiveness of the robots in trying to
maintain the vertical force inside [−fmax , −fmin ]. We choose fmax = mNmax min
g
and fmin = Nmax . In this way, we guarantee that the overall force exerted
mmin g
10 10
5 5
0 0
0 20 40 60 80 100 0 20 40 60 80 100
||f|| ||f||
20 30
master
20 master
10
10
0 0
0 20 40 60 80 100 0 20 40 60 80 100
(a) (b)
Fig. 7. The three components of pR1 and pdR1 (dotted lines of the same color) and the
intensity of the cable forces. By modeling fC1 as in (5) it is possible, with the same
parameters, to have the leader follow its desired trajectory under normal conditions
(a), and to limit the forces exerted in case the swarm does not follow it (b).
(i.e., energy consumption and safety) and tracking error (i.e., performance). A
behavior that is too compliant can compromise the reach of the final position,
a behavior that is too stiff can require a large amount of force from the leader
side and thus a lot of energy consumption and an increased risk of ruptures.
Due to space limitations we omit the performed numerical results showing such
intuitive trade-off. A possible solution to better deal with such trade-off is to
introduce a nonlinear saturation model like, e.g., the following one:
4 Discussion
but it changes its behavior consistently if the followers stay still and not follow
as expected (see Fig. 7).
Finally, the results in Fig. 6, which refer to scenario Sd , show the potential
of the algorithm also in conditions were the swarm is not completely aware of
the parameters of the system. Even with a limited and imperfect knowledge the
agents are capable of commonly carrying an object in a stable way.
5 Conclusions
This work is a simulative study on communication-less cooperative object trans-
portation by swarms of aerial agents. The main focus concerns the important
role of the internal force to make asymptotically stable certain equilibria of the
system and to enhance the manipulation capabilities of the swarm. An imperfect
knowledge of the system parameters has been also treated. Several parallelisms
with biological examples are discussed as well. Validation of the proposed algo-
rithm on real platforms and theoretical proofs is left as an important future work.
Of course, in order to realize a practical implementation, additional aspects, such
as a collision avoidance technique, are required, especially for very large number
of agents. Furthermore, the relaxation of some hypothesis, e.g., on the cables
attachment points or on the rigidity of the load, may represent an interesting
enlargement of the application domain.
References
1. Abbasnejad, G., Carricato, M.: Direct geometrico-static problem of undercon-
strained cable-driven parallel robots with n cables. IEEE Trans. Robot. 31(2),
468–478 (2015)
2. Alam, M.S., Wahab, M.A., Jenkins, C.H.: Mechanics in naturally compliant struc-
tures. Mech. Mater. 39(2), 145–160 (2007)
3. Berman, S., Lindsey, Q., Sakar, M.S., Kumar, V., Pratt, S.C.: Experimental study
and modeling of group retrieval in ants as an approach to collective transport in
swarm robotic systems. Proc. IEEE 99(9), 1470–1481 (2011)
4. Brescianini, D., D’Andrea, R.: Design, modeling and control of an omni-directional
aerial vehicle. In: 2016 IEEE International Conference on Robotics and Automa-
tion, Stockholm, Sweden, pp. 3261–3266, May 2016
5. Franks, N.R.: Teams in social insects: group retrieval of prey by army ants (eciton
burchelli, hymenoptera: Formicidae). Behav. Ecol. Sociobiol. 18(6), 425–429 (1986)
6. Franks, N.R., Sendova-Franks, A.B., Anderson, C.: Division of labour within teams
of new world and old world army ants. Anim. Behav. 62(4), 635–642 (2001)
7. Gassner, M., Cieslewski, T., Scaramuzza, D.: Dynamic collaboration without com-
munication: vision-based cable-suspended load transport with two quadrotors. In:
2017 IEEE International Conference on Robotics and Automation, Singapore, pp.
5196–5202, May 2017
A Study on Force-Based Collaboration in Flying Swarms 15
1 Introduction
In swarm robotics, communication plays a central role and can significantly
enhance collective performance [3]. Designing effective communication mech-
anisms is challenging and design choices can have an important impact on
the effectiveness, complexity, and cost of a swarm [2]. Notwithstanding the
advancements achieved in the last decade [4,7,24,29,34,43,51], the design of
robot swarms is still at dawn and no generally applicable methodology has been
proposed so far [8,11,21]. Automatic design methods are a promising way of
approaching the issue [6,15]. In automatic methods, the design problem is cast
into an optimization problem: a space of solutions is searched via an optimiza-
tion algorithm, with the goal of maximizing a performance measure. Most of the
The proposed method was implemented and tested by KH. The experiments were
designed by the three authors. This paper was drafted by KH, refined by MB, and
revised by the three authors. The research was conceived and directed by MB.
c Springer Nature Switzerland AG 2018
M. Dorigo et al. (Eds.): ANTS 2018, LNCS 11172, pp. 16–29, 2018.
https://doi.org/10.1007/978-3-030-00533-7_2
Automatic Design of Communication-Based Behaviors for Robot Swarms 17
research on the automatic design of robot swarms has been inspired by neuro-
evolution [37,47]. In this approach, robots are controlled by a neural network,
whose parameters are obtained via artificial evolution [12,27,31,38,44,45,47,50].
Other methods have been proposed that are based on different control archi-
tectures and/or different optimization algorithms [16,18,22,30]. Among them,
Chocolate [16] produces probabilistic finite state machines by using the irace
optimization algorithm [35] to assemble preexisting low-level behaviors and con-
ditions, and to fine-tune their parameters. The low-level behaviors, which define
the actions that individual robots can perform, are: exploration, stop, phototaxis,
anti-phototaxis, attraction to neighbors, repulsion from neighbors. The condi-
tions, which define events that cause a transition between low-level behaviors,
are: black-floor, white-floor, gray-floor, neighbor-count, inverted-neighbor-count,
fixed-probability.
In this paper, we study the automatic design of collective behaviors that rely
on communication. In particular, we are interested in exploring the case in which
messages exchanged by the robots do not have an a priori defined semantics. We
wish to develop an automatic design process that, on a per-mission basis, defines
(i) the conditions under which a robot broadcasts a message and (ii) the effects
that this message has on the behavior of the receiving peers.
We introduce Gianduja, a new instance of AutoMoDe [18]. Gianduja extends
Chocolate by adding the capability of locally broadcasting a single message and
reacting to it. We test Gianduja on three missions that we shall call aggrega-
tion, stop, and decision. We present results of experiments performed with a
swarm of 20 e-puck robots [36].
Within the evolutionary robotics approach, it has already been shown that an
automatic design process can (implicitly) give a semantics to an a priori mean-
ingless message. Nonetheless, this has been demonstrated only on teams of two
robots [1,49]. The novel elements that we propose in this paper are that: (1) we
study the emergence of a message semantics in swarm robotics and we demon-
strate it with a swarm of 20 robots; (2) we show that a message semantics can
emerge also when robots are controlled by a finite state machine; and (3) we
consider three different missions in which the emerging semantics is different.
2 Related Work
Communication—be it direct or indirect, explicit or implicit—is an integral part
of most robot swarms demonstrated so far. As a result, the literature on com-
munication in swarm robotics is extremely large and covering it goes beyond the
scope of this paper. In particular, we will not cover studies in which communi-
cation has been a priori defined by the designer—e.g., [2,3,9,14,28]. Instead, we
will focus on studies in which communication has been automatically designed.
The vast majority of studies in which communication emerged from an auto-
matic design process belong within evolutionary robotics [37,47,48]. Quinn et al.
[41,42] were the first to study the emergence of communication between agents.
In their studies, robots move in an arbitrary direction while staying close to each
Another random document with
no related content on Scribd:
machine. Fortunately I had presence of mind enough to keep my
watch going, as well as the captain’s chronometer, for otherwise I
would have had no knowledge of the passage of time. Once or twice
the scarlet women visited the ship, but seemed nervous and wary,
and made no effort to approach or molest me, merely gazed about
as if searching for something—perhaps for me—and then retiring.
Several times, too, I ventured on deck, and peered over the ship’s
side, but saw none of the giantesses, although with the glasses I
could see crowds of the beings about the city in the distance.
“Also, I noticed among them, several individuals who were much
smaller than the rest, and who appeared to be men, although I could
not be sure. I also discovered, and almost lost my life in the
discovery, that the atmosphere of this place is unfit for human beings
to breathe, and is thick with sulphurous fumes. Close to the ground
these fumes are so dense that a person would succumb in a few
moments, but at the height of the Chiriqui’s decks, nearly seventy
feet above the rocky bed on which she rests, the air is breathable,
although it causes one to choke and cough after a few minutes. And
I am sure that the houses of these giant beings have been built on
the summits of the basalt columns in order to avoid the suffocating
fumes of the lower levels. Later, too, I learned that the membrane-
like frills upon these creatures are a sort of gills, or as I might say,
natural gas-masks, which by some means enable the beings to
breathe the sulphur-laden air. But even with these, they avoid the
lower areas where the fumes are the worst, and only visit them when
necessity arises, which accounts for my being left in peace, with
none of the horrible women near the ship, for days at a time. I
discovered the presence of the sulphur gas on the first day when,
attempting to eat, I removed my gas-mask. Suffocating as I found
the fumes, I was compelled to endure them, and gradually I became
slightly accustomed to them, so that now I have little trouble in
breathing during the short time it takes me to eat my meals. At all
other times I must wear the apparatus, and I thank God that this is
so, for I know now that it is the gas-mask which so far has preserved
my life.
“On the tenth day after my arrival I noticed a number of the
giantesses gathering about the huge, spherical airship which still
rested on its cradle near the Chiriqui, but which, I have forgotten to
state, ceased to emit its green or red lights after it had landed. Lying
there it resembled nothing so much as a gigantic can-buoy or a
floating mine, if one can imagine a buoy two hundred yards in
diameter.
“On the day I mentioned, all interests seemed to be centered on
the thing, and cautiously peering from the shelter of the deck-house,
I watched the proceedings. Presently several of the women entered
the sphere through an opening in its middle band; the aperture
closed behind them, and immediately there was a low, humming
sound as of machinery. As the sounds issued from the sphere, the
cables to which were attached the smaller spheres (which glowed
red when carrying the Chiriqui through the air) were drawn in until
the two smaller spheres were resting in recesses at the axes of the
large sphere, and where they appeared merely as hemi-spherical
projections. Then, slowly at first, but with ever increasing speed, the
slender rods about the large sphere began to move back and forth,
or rather in an oscillating manner, until they were vibrating with such
rapidity that they appeared merely rays of light. Slowly, majestically,
the immense globe rose from its cradle, and gathering headway,
leaped upward to an immense height. Then, tilting at an angle, it
passed over the city and headed for an immense pinnacle of rock,
which, fully seven miles from where I stood, reminded me of a
gigantic chimney or funnel.
“Although it was barely visible to the naked eye, I could see it
distinctly through the glasses, and I watched it with the most intense
and concentrated interest. For a few moments it remained, poised a
hundred feet or so above the pinnacle. Then, from the towering,
tapering rock, a terrific jet of steam roared forth, and striking the
great spherical machine above it, hurled it upward and beyond my
vision. Give close heed to these words, whoever may, by God’s
grace, be listening to what I say, for upon them may hinge the fate of
the human race. Only by this means, by being shot upward by this
titanic jet of steam, can the airship leave this subterranean land and
emerge through the crater by which it entered bearing the Chiriqui.
Within this place it can sail at will; once above the crater opening it
can travel anywhere, although it cannot land; but by some unknown
force or magnetic attraction or freak of gravitation the machine
cannot ascend through the crater, although, when over it, it will drop
like a plummet through the opening. And herein—for the sake of
humanity, listen to this and remember my words—lies a means of
destroying the machine, for by surrounding the crater with powerful
guns the sphere can be shelled as it emerges and utterly destroyed.
To attempt to do so as it returns to the crater would be suicidal, for
once in the outer air, it emanates vast quantities of most poisonous
gas, and all living things within a radius of several miles would be
struck down unconscious, as were my companions on the Chiriqui.
Even if gas-masks were worn, it would be most difficult to destroy
the machine as it descended, for it travels with incredible speed in its
descent and, moreover, the terrible creatures who man the thing
would see that enemies lurked near and would find some means of
destroying them, or by the mysterious magnet force they control,
would draw even the heaviest cannon to the machine as an ordinary
magnet draws needles or iron filings. So if the thing is to be
destroyed, it must be done as the machine emerges from the crater.
Would to God that I could tell where the crater is, but beyond feeling
sure it is at the summit of an Andean peak, I have no means of
locating it.
“But I was telling of what occurred on that tenth day when the
spherical airship was projected from my sight by the blast of steam.
As the machine vanished, the women who had watched its
departure, returned to their city, and I swept the landscape with my
glasses, wondering at the bleak, terrible scenery and bizarre colors.
Here the message broke off abruptly, and Frank and I sat staring at
each other, fearing to speak lest we might interrupt or miss the words
which might come, and listening with straining ears at the head-sets.
For an hour we sat there and then, once more the voice spoke.
“The doom that I feared is approaching. I have been here for three
months and this will, I know, be my final message. Oh that I could
only be sure that someone has heard my words, that my fate has not
been in vain but has served to warn my fellow beings. But I must
hurry on. I have learned everything of importance. I have watched,
studied and have even learned to understand much of the language
of these beings. I found that there were men. They are puny beings
compared to the women, though ten-foot giants compared to normal
men, and they are cowed, abject, mere slaves of the females. Only
enough male children are permitted to survive to propagate the race.
All others are killed.
“As they reach manhood only those males of super-intelligence,
strength and virility are permitted to live. The others are destroyed
and—yes, horrible as it sounds, their bodies, like those of the
murdered infants and of the aged, sick or infirm, are devoured. And
as fast as the males attain middle age their lives are forfeited. Long
ago these beings subsisted upon the few wild creatures which
roamed their land; but long ago all these were exhausted and human
flesh became the only meat. There is no vegetable food, and for a
time the sacrificed surplus males, and the aged, provided food for
the race. But gradually the male births decreased, female children
preponderated, and with the increased population resulting, the
males were too few to nourish the others. Then, through what
damnable accident or design I do not know, the creatures went forth
in their airship and discovered the teeming millions of human beings
on earth.
“But the bulk of humanity was and still is safe from them, at least
until new means of attacking mankind are devised, for the globular
airship cannot approach the land. The very power it uses to lift the
greatest steamships and carry them off, draws the machine to the
earth and holds it fast. But above water, which acts as an insulator
apparently, the apparatus can operate at will. And they have a two-
fold purpose in capturing ships. All the available metal in this land
was exhausted in constructing two of the spherical machines. One of
these never returned from its first trip, and only the one remains. To
construct more, these giant women plan to use the metal salvaged
from captured ships, until a vast fleet of the infernal things is ready to
go forth and wipe the seas clean of ships and human beings. And
the bodies of the men and women, struck down by the gas, are to
serve as food for these demons in human form.
“This is the most horrible, blood-curdling thing of all. Rendered
unconscious by the gas, the victims remain in a state of suspended
animation indefinitely, exactly as do grubs, spiders and insects when
stung by certain species of wasps and placed in their nests to
provide food for their young. Stacked in great storage vaults these
breathing, living, but paralyzed human beings are kept, and as
needed, are taken out.
“Already they have a supply on hand sufficient to last them for
over a year. Some of the Cyclops company are still preserved; there
are over three hundred from the Chiriqui, hundreds from other ships,
and the entire crew of the McCracken.
“All these things I learned little by little, and mainly through a
friend, for marvelous as it may seem, I have a friend—if friend he
can be called, a miserable, trembling, terrified male, who, doomed to
death, sought to escape his fate and sought refuge with me,
dreading my presence less than his doom, and hoping that such a
feared and almost reverenced being as myself might protect him. For
two months he has been my companion, but he cannot eat anything
but meat and the supply of meat upon the ship is getting low, and
sooner or later he must succumb. And the women, maddened at his
escape from their clutches, though not yet daring to approach too
closely to me, are getting bolder. Some time, at some unguarded
moment, they will find the poor fellow alone and will fall upon him.
And in his terror, in an effort to buy his life, he will, I know, reveal to
them that I am but an ordinary mortal, a man who eats and drinks
and who survived the gas by mechanical and not supernatural
means. But I will not be taken alive by these fearful female
cannibals. When the time comes, as I know it will, I will blow my
brains out, and though they may devour my body they will not rend
me alive. No more ships have been brought in here since the
McCracken was captured. But this I know is due to the fact that all
the energies of these creatures are being devoted to building
additional air machines. This work goes on in a vast cavern beyond
the city where tremendous forces, furnaces with heat beyond human
conception and machines of which we know nothing, are controlled
by the internal steam, the radiant energy and the magnetic powers of
the earth’s core.
“And now, again let me implore any and all who may hear my
words to give close attention to what I say, for here again is a means
by which humanity may combat and destroy these ghastly, gigantic
cannibals. The spherical air-machines are helpless from above.
Their magnetic or electrical forces extend only downwards. The
gasses they throw out are heavier than air and descend but cannot
ascend, and by means of swift planes, huge bombs and machine
guns, the things can be easily destroyed. And they cannot travel
without throwing off the dazzling green light. Only when motionless
are they dark. And so they will offer easy marks and can be readily
detected. So, I beseech you who may hear, that the governments
are notified and warned and that a fleet or many fleets of airplanes
properly equipped patrol the seas, and at first sight of one of the
green meteors rise above it and utterly destroy it without mercy.
“Wait! I hear a terrified scream.... I am back again at the
transmitter. It was the fellow who has been with me. Poor devil! He
has met his fate, but after all it was the custom of his people, and,
moreover, he would have starved to death in a few days. For that
matter I, too, face starvation. The ship’s stock is running low; all the
food upon the McCracken was destroyed in cutting up that vessel,
and unless another ship is captured I will have no food after two
weeks more. What a strange thought! How terrible an idea! That the
awful fate of hundreds of my fellows would be my salvation! But I will
never live to die from hunger. I can hear the terrible screams of my
late companion on the deck outside. God! It is the end! The fellow
must have told the enraged females. His body has been torn to
shreds. With bloody hands and reeking lips they are rushing towards
the upper deck where I sit. They are here! This is my last word! God
grant that I have been heard! I am about to⸺”
Crashing in our ears came the report of a pistol.
The End
1 The message as it came in, was halting, and interrupted, with many unintelligible
words and repetitions, as if the sender were laboring under an intense strain or was an
amateur. For the sake of clarity and continuity, the communication has been edited and
filled in, but not altered in any detail.
2 The metropolitan papers reported the meteor on the eighteenth and stated it was
observed by those on the Chiriqui on the evening of the seventeenth, but it must be
remembered that the Chiriqui was in the western Pacific and hence had gained a day
in time.
Updated editions will replace the previous one—the old editions will
be renamed.
1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside the
United States, check the laws of your country in addition to the terms
of this agreement before downloading, copying, displaying,
performing, distributing or creating derivative works based on this
work or any other Project Gutenberg™ work. The Foundation makes
no representations concerning the copyright status of any work in
any country other than the United States.
1.E.6. You may convert to and distribute this work in any binary,
compressed, marked up, nonproprietary or proprietary form,
including any word processing or hypertext form. However, if you
provide access to or distribute copies of a Project Gutenberg™ work
in a format other than “Plain Vanilla ASCII” or other format used in
the official version posted on the official Project Gutenberg™ website
(www.gutenberg.org), you must, at no additional cost, fee or expense
to the user, provide a copy, a means of exporting a copy, or a means
of obtaining a copy upon request, of the work in its original “Plain
Vanilla ASCII” or other form. Any alternate format must include the
full Project Gutenberg™ License as specified in paragraph 1.E.1.
• You pay a royalty fee of 20% of the gross profits you derive from
the use of Project Gutenberg™ works calculated using the
method you already use to calculate your applicable taxes. The
fee is owed to the owner of the Project Gutenberg™ trademark,
but he has agreed to donate royalties under this paragraph to
the Project Gutenberg Literary Archive Foundation. Royalty
payments must be paid within 60 days following each date on
which you prepare (or are legally required to prepare) your
periodic tax returns. Royalty payments should be clearly marked
as such and sent to the Project Gutenberg Literary Archive
Foundation at the address specified in Section 4, “Information
about donations to the Project Gutenberg Literary Archive
Foundation.”
• You provide a full refund of any money paid by a user who
notifies you in writing (or by e-mail) within 30 days of receipt that
s/he does not agree to the terms of the full Project Gutenberg™
License. You must require such a user to return or destroy all
copies of the works possessed in a physical medium and
discontinue all use of and all access to other copies of Project
Gutenberg™ works.
• You comply with all other terms of this agreement for free
distribution of Project Gutenberg™ works.
1.F.
1.F.4. Except for the limited right of replacement or refund set forth in
paragraph 1.F.3, this work is provided to you ‘AS-IS’, WITH NO
OTHER WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR ANY PURPOSE.