Evolution of CoopEration in artifiCial ants

174 Pages·2007·2.67 MB·English

Evolution of CoopEration in artifiCial ants
THÈSE NO 3943 (2007)
PRÉSENTÉE LE 9 NOVEMBRE 2007
À LA FACULTÉ DES SCIENCES ET TECHNIQUES DE L'INGÉNIEUR
Laboratoire de Systèmes Intelligents
SECTION DE MICROTECHNIQUE
ÉCOLE POLYTECHNIQUE FÉDÉRALE DE LAUSANNE
POUR L'OBTENTION DU GRADE DE DOCTEUR ÈS SCIENCES
PAR
Markus WAIBEL
Dipl.-Ing. in Technical Physics, Vienna University of Technology, Autriche
et de nationalité autrichienne
acceptée sur proposition du jury:
Prof. H. Bleuler, président du jury
Prof. D. Floreano, Prof. L. Keller, directeurs de thèse
Dr M. Chapuisat, rapporteur
Prof. O. Holland, rapporteur
Dr F. Mondada, rapporteur
Lausanne, EPFL
2007

Acknowledgements
[...] discoveriesandimprovementsinvariably
involvethecooperationofmanyminds.
AlexanderGrahamBell(1847-1922)
Many people have contributed to the work presented in this thesis. First I
would like to thank my two supervisors, Prof. Dario Floreano and Prof. Laurent
Keller, for making me part of this project and thus giving me the opportunity to
dive into this amazing research field. I thank Laurent for his radiant motivation,
endless patience and for sharing his clear mind, and Dario for his constant guid-
ance, his trust and for being an inspiration. In addition, I would like to thank
Dario for assembling the amazing group of people in the previous Autonomous
Systems Laboratory andthe current Laboratory of IntelligentSystems.
I would like to thank all my colleagues at the Laboratory of Intelligent Sys-
tems for providing the great, stimulating atmosphere that has made these years
soenjoyable. Inparticular,IwouldliketothankSaraMitriforhercalm,persistent
thinkingandourendlessdiscussions,PeterDürrforhisnever-failing,contagious
enthusiasmandhissharpwits,andDaneshTaraporeforthegreattimeweshared
in the office and for making Alice work. I would also like to thank Laurent’s re-
search group at the Department of Ecology and Evolution of the University of
Lausanne, who have helped with their advice and criticism on countless occa-
sions.
Anumberof people haveenriched my academicexperience and have helped
me feel at home in Lausanne. I am very grateful to the Pavillon-Jaune-team,
who made me feel at home from day one: Daniel Roggen, Stéphane Magnenat,
i
ii ACKNOWLEDGEMENTS
Francesco Mondada, Michael Bonani, Jean-Christophe Zufferey, Dominique Eti-
enne, Jesper Blynel, Diego Federici and Claudio Mattiussi. My special gratitude
goestoallmovie-nightparticipants,whoaretoonumeroustoname,forthemany
unforgettable memories.
InwritingthisthesisIhavebenefittedfrom commentsandhelpfuldiscussion
provided by a large number of people. I am particularly indebted to Peter Dürr
and Sara Mitri for their truly tireless efforts in offering advice. I would also like
to thank Danesh Tarapore, John Wang, Morgan Pearcy, Martijn Bosch, Claudio
Mattiussi, SimonHarding, DanielMarbach, GuyTheraulaz,RobHammon,Jean-
Louis Deneubourg and the members of the thesis committee, Hannes Bleuler,
Michel Chapuisat, Owen Holland and Francesco Mondada. In addition, I am
indebted to Julien Chassot, Gilles Caprari, Stéphane Magnenat, Antoine Beyeler,
Andres Perez-Uribe, Cyrille Dunant, Gintautas Narvydas and Danesh Tarapore
for contributing to this project.
I would like to thank Dominique Etienne and Anouk Hein - nothing would
haveworked withoutyou. IwouldalsoliketoexpressmygratitudetotheEPFL,
the University of Lausanne, and the Swiss National Science Foundation for their
support and for fundingmyresearch.
I also thank all my friends for sharing all aspects of life, my parents, Maria
andKarlfortheirguidance,trustandpatience,andmybrotherRomanandsister
Sophie for their support.
Most importantly, I would like to thank Christine for all the things that can’t
be putinto words.
Lausanne, November2007
Abstract
The evolution of cooperation is a fundamental and enduring puzzle in biology
and thesocial sciences. Hundredsoftheoretical modelshave beenproposed, but
empirical research has been hindered by the generation time of social organisms
and by the difficulties of quantifying costs and benefits of cooperation. The sig-
nificant increase in computational power in the last decade has made artificial
evolution ofsimple social robots a promising alternative.
This thesis is concerned with the artificial evolution of groups of cooperating
robots. It argues that artificial evolution of robotic agents is a powerful tool to
address open questions in evolutionary biology, and shows how insights gained
from the study of artificial and biological multi-agent systems can be mutually
beneficial for both biology and robotics. The work presented in this thesis con-
tributes to biology by showing how artificial evolution can be used to quantify
keyfactorsintheevolutionofcooperationinbiologicalsystemsandbyproviding
an empirical test of a central part of biological theory. In addition, it reveals the
importanceofthegeneticarchitecturefortheevolutionofefficientcooperationin
groups of organisms. The work also contributes to robotics by identifying three
different classes of multi-robot tasks depending on the amount of cooperation
required between team members and by suggesting guidelines for the evolution
of efficient robot teams. Furthermore it shows how simulations can be used to
successfully evolve controllers for physical robot teams.
Keywords: Artificial evolution; multi-agent systems; social insects; evolutionary
robotics; team composition; task allocation; division of labor; fitness allocation;
cooperation; altruism.
iii
iv ABSTRACT
Zusammenfassung
Die Evolution von kooperativem Verhalten ist ein grundlegendes Problem der
Biologie und der Sozialwissenschaften. Die empirische Überprüfung der zahlre-
ichen theoretische Modelle wird durch die Generationszeit sozialer Organismen
und die Schwierigkeit Kosten und Nutzen sozialen Verhaltens zu quantifizieren
erschwert. Der dramatische Anstieg der Rechnerleistung in den letzten zehn
Jahren machtdie künstliche Evolution von einfachen sozialen Robotern zur viel-
versprechenden Alternative.
Diese Dissertation beschäftigt sich mit der künstlichen Evolution von Grup-
pen kooperierender Roboter. Sie schlägt vor die künstliche Evolution von Ro-
botern als mächtiges Werkzeug zur Bearbeitung offener Fragen der Evolutions-
biologie zu verwenden und zeigt, dass Erkenntnisse aus der Erforschung natür-
licher und künstlicher Multi-Agenten-Systeme gleichzeitig einen Nutzen für die
Biologie und die Robotik bringen können. Die vorliegende Arbeit leistet einen
Beitrag zur Biologie, indem sie zeigt wie künstliche Evolution benutzt werden
kannumSchlüsselfaktorenderEvolutionsozialenVerhaltensinbiologischenSys-
temenzuquantifizierenundlieferteinenempirischenTestfüreinezentraleTheo-
rie derEvolutionsbiologie. Zusätzlich verdeutlicht sie den wesentlichen Einfluss
der genetischen Architektur auf die Evolution effizienter kooperativer Gruppen
von Organismen.
DieArbeitleistetaucheinenBeitragzurRobotik,indemsiedreiverschiedene
Klassen von Multi-Roboter-Problemen identifiziert und diese anhand des unter-
schiedlichen Masses an Zusammenarbeit, daszur Lösung desProblems notwen-
digist,unterscheidet. AusserdemschlägtsieRichtlinienzurEvolutioneffizienter
v
vi ZUSAMMENFASSUNG
RoboterteamsvorundzeigtwieComputersimulationengenutztwerdenkönnen,
um erfolgreiche Steueralgorithmen fürreale Roboterteams zuerhalten.
Schlüsselwörter: Künstliche Evolution; Multi-Agenten-Systeme; Soziale Insek-
ten; Robotik; Arbeitsverteilung; Arbeitsteilung; Kooperation; Altruismus.
Contents
Acknowledgements i
Abstract iii
Zusammenfassung v
Contents vii
1 Introduction 1
1.1 Natural Selection andthe Evolution ofCooperation . . . . . . . . . 1
1.2 Factors Influencingthe Evolution ofCooperation . . . . . . . . . . . 2
1.2.1 Direct selection . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Indirectselection . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 Group selection . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Empirical Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 ModelingCooperation . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4.1 Evolutionary Robotics . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 AQuantitative Testof Hamilton’sRule 13
2.1 State ofthe Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Materials andMethods . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.2 Costs andbenefits ofaltruism . . . . . . . . . . . . . . . . . . 17
2.2.3 Relatednessand artificial evolution . . . . . . . . . . . . . . 18
2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
vii
viii CONTENTS
2.3.1 Groups with high relatedness . . . . . . . . . . . . . . . . . . 21
2.3.2 Groups with intermediate relatedness . . . . . . . . . . . . . 23
2.3.3 Groups with lowrelatedness . . . . . . . . . . . . . . . . . . 23
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3 Division of Labor andColony Efficiencyin Social Insects 27
3.1 Division ofLabor andColony Efficiency in Social Insects . . . . . . 28
3.2 Materialsand Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2.1 Colonytasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2.2 Geneticarchitecture . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.3 Environmental and internal perturbations . . . . . . . . . . 32
3.2.4 Coloniesandselection algorithm . . . . . . . . . . . . . . . . 33
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.1 Coloniesofunrelated individuals . . . . . . . . . . . . . . . 34
3.3.2 Coloniesofhighly related individuals . . . . . . . . . . . . . 37
3.3.3 Colonieswith intermediate relatedness . . . . . . . . . . . . 39
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4 Factors inthe Evolution of Multi-AgentSystems 47
4.1 State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2 Evolutionary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Experimental Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.1 Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.2 Control andgenetic architecture . . . . . . . . . . . . . . . . 58
4.3.3 Collective tasks . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3.4 Evolutionary experiments . . . . . . . . . . . . . . . . . . . . 59
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.4.1 Task1 -Individualforaging . . . . . . . . . . . . . . . . . . . 60
4.4.2 Task2 -Cooperative foraging . . . . . . . . . . . . . . . . . . 60
4.4.3 Task3 -Altruistic cooperative foraging . . . . . . . . . . . . 60
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.5.1 Task1 -Individualforaging . . . . . . . . . . . . . . . . . . . 63
4.5.2 Task2 -Cooperative foraging . . . . . . . . . . . . . . . . . . 65
4.5.3 Task3 -Altruistic cooperative foraging . . . . . . . . . . . . 69