28
Engineering Self- Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Embed Size (px)

Citation preview

Page 1: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Engineering Self-Organizing SystemsCognition &

Emergence of Control

Salima HassasUniversity of Lyon

Page 2: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Summary• From Organization to Self-Organization

– Organization dynamics & self-organization– Relation entre Organization/Control

• Engineering (Self)–organization in MAS– Organization oriented approaches– Dynamic Organization et re-Organization– Self-organization

• Engineering Self-Organization: a complex system perspective for cognition– Organization as emergent control system– Cognition, representation and evolution

• Conclusions

Page 3: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

(Self) Organization

• What is an organization?

A non-random arrangement of components or parts interconnected in a manner as to constitute a system identifiable as a unit.

Page 4: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

(Self) Organization

• What is an organization?

A non-random arrangement of components or parts interconnected in a manner as to constitute a system identifiable as a unit.

Existence of a process that produces the organization

Page 5: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Different situations of (Self) Organization

a priori

Static Organization (a priori)

Process

Organization

Inte

rnal

ex

tern

al

DynamicStatic Dynamic Static a posteriori

Dynamic

Dynamic

Static

Page 6: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

• System’s organization is static, defined a priori– predefined roles, relations

• Process producing it, external to the organization, known and defined a priori

The organization is static all along the system’s life : no change/no adaptation

Page 7: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

• In Multi-Agents SystemsOrganization based

methodologies/tools:

Ex: AGR, TAEMS,MOISE+, ..etc

System dynamics and Organization explicitly specified at Design time.

No adaptation (even programmed)

Ferber &al.

Page 8: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

(Self) Organization

a priori

Static Organization

Process

Organization

Inte

rnal

ex

tern

al

DynamicStatic Dynamic Static a posteriori

Dynamic

Dynamic

Static

Dynamic Organization

Page 9: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

• System’s organization is dynamic, but known a priori– Characteristics provided

• The process producing it, is external to the

organization, but is conditioned by the environment

– Ex: Agents dynamically organize themselves to form a circle

The organization is dynamic Programmed re-organization in order to adapt to the environment constraints

circlecircle

Page 10: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

The organization is dynamic Programmed re-organization/adaption environment

Example of rule: Heterarchy if #agents<= n, Hierarchy if #agents > n

Page 11: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

• Use of Meta-Organizations + transformation rules, Observer/controler based architectures, rule based organization process, etc.

• At design-time Specify interactions rules, transformation rules, observation/control

architectureExample : Specify transformation rules : Heterarchy Hierarchy– If (condition) then select n agents, elect a leader, etc.

• At run-time Generate organization according to the specified conditions changes

Page 12: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

(Self) Organization

a priori

Static Organization

Process

Organization

Inte

rnal

ex

tern

al

DynamicStatic Dynamic Static a posteriori

Dynamic

Dynamic

Static

Dynamic Organization

Emergent Organization& static control

Page 13: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

• System’s organization is dynamic, and unknown a priori (emergent)– Environment constraints provided

• The process producing it, internal to the organization, and is conditioned by the environment

– System behavior/coupled with the environment change/constraints

The organization is emergent : System/environment coupling is a complex program

- Program the system-environment coupling (ex: bio-inspired techniques, game theory, evolutionary control)

Page 14: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

AMAS theory (Game theory inspiration) – (P. Glize, MP. Gleize, & al.)

• Basic Principle (Axelrod’s work on iterated games) => Long term perspective : Altruist strategy always wins (Cooperative attitude)

- Design time – Define cooperative/non cooperative situations– Define rules to pass from coop/non coop

- Run time The system finds by itself the adequate organization to solve

the problem (the organization is not explicit)A kind of « Situation-based » programmed adaptation

(http://www.irit.fr/ADELFE)

Page 15: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Social insects Inspiration (ants foraging, collective sorting, ..)

• Case 1: Transposing metaphors (mimic biological systems)– Routing Algorithms in Networks , ACO meta heuristic, ..etc. – Routing (Rare) Information in P2P networks (illustration)

ObservationsObservations

Biological Model

Biological Model

Induction

Biology

Page 16: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Social insects Inspiration (ants foraging, collective sorting, ..)

• Case 1: Transposing metaphors (mimic biological systems)– Routing Algorithms in Networks , ACO meta heuristic, ..etc. – Routing (Rare) Information in P2P networks (illustration)

ObservationsObservations

Biological Model

Biological Model

Induction

Biology

Computation Model

Computation Model

Computing

Page 17: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Social insects Inspiration (ants foraging, collective sorting, ..)

• Case 1: Transposing metaphors (mimic biological systems)– Routing Algorithms in Networks , ACO meta heuristic, ..etc. – Routing (Rare) Information in P2P networks (Illustration)

ObservationsObservations

Biological Model

Biological Model

Induction

Biology

ApplicationsApplications

Computation Model

Computation Model

Transposing

Computing

Page 18: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Social insects Inspiration (ants foraging, collective sorting, ..)

• Case 2: more deepened understanding– Stigmergy in Negotiation : CESNA - Exchange between

Stigmergic Negotiating Agents- (Armetta & Hassas 2006) – Environment Pressure selection +structural coupling

• Work of L. Steels – Emergence of language • Coupling structure/behaviors (retro-active co-evolution of social and

spatial organizations in MAS) (Illustration on: ants foraging) • Application on the web: Social Tagging, social networks (MySurf, UTTU)

– Case of neuronal computing + evolutionary algorithm• Selection: evolutionary algorithm• Structural coupling: change in neural networks

Page 19: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

(Self) Organization

a priori

Static Organization

Process

Organization

Inte

rnal

ex

tern

al

DynamicStatic Dynamic Static a posteriori

Dynamic

Dynamic

Static

Dynamic Organization & static programmed control

Emergent Organization& dynamic (programmed) control

Emergent Organization & emergent control

Page 20: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

• System’s organization is dynamic, and unknown a priori (emergent)– Environment constraints provided

• The process producing it, internal to the organization, and produced by system/environment dynamics coupling– System organization and behavior/environment change

constraints strongly coupled in a retro-active loop(ex: natural ants foraging)

The organization and the process are emergent : strong coupling of system/environment dynamics

Page 21: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

The organization and the process are emergent : strong coupling of system/environment

-System/environment coupling is produced by the system/environment dynamics - Need for (system) cognition and evolution

Page 22: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Multi-Agents System = Collectif of situated agents in a shared un environnement

22

Ferber, J. (1995). Les systèmes multi-agents. Vers une intelligence collective. Paris: InterEditions.

Agents are des local and autonomous units of material symbols processing

Agents inspired by human societies, but can represent : neurons, birds, fish, cells, particles, etc.

IndividualCollective

InternalExternal

Page 23: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Individual/internal Individual/external

MeSubjectivity

He,She, ThisObjectivity

<Mental states, Agents architectures>

<Agents behavior>

Interiority

The object

Collective/internal Collective/external

WeIntersubjectivity

Them, all thisInterobjectivity

<Interactions modes, Shared knowledge><Organizations, institutions,Evolution/ organizations emergence, social facts>

the noosphere The social structure

Ferber, J. (2006). Concepts et méthodologies multi-agents. In F. Amblard, & D. Phan (Ed), Modélisation et simulation multi-agents : applications pour les Sciences de l'Homme et de la Société (pp. 23-48). Paris: Lavoisier.

23

Internal to Agent External to Agent

MAS Analysis according to 4 quadrants (J. Ferber 2006)

Page 24: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

MER from a collective/Internal composition(L. Lana de Carvalho & al. ECCS’2008)

Exploring Complex System

Problem Environment

Parameters Change

Shapes

Shapes

Second Order Praxis

Agents

In evolution

Fix the solution => memory

24

AGENT*AGENTCAS*AGENT

Parameters

Indivodual/Internal

(collective action)Praxis

Individual/External

Complex System

Collective/External

emergenceBehaviors

Collective/Internal

RepresentationAdaptiveExploiting

First Order Praxis

Page 25: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

The organization and the process are emergent : strong coupling of system/environment

-System/environment coupling is produced by the system/environment dynamics - Need for (system) cognition and evolution

Self-organization= emergence of a new system that controls the initial system (to organize)

Page 26: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Conclusion

a priori

Organization=Result of

designed fixed control

Process

Organization

Inte

rnal

ex

tern

al

DynamicStatic Dynamic Static a posteriori

Dynamic

Dynamic

StaticOrganization=

Result of a programmed

control

Organization=Emergent from a

dynamics as a control

Organization=Emergent control

system

Page 27: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

27

Date d’apparition

Cognitivism• Material Representations• Logics Theorms

Dynamical Approach• Dynamic Representations• Representations part of the

cognitive development structure

Connexionism• Micro-representations• Macro-representations• Neural Networks

Enactivism• Self-organization• Natural tendency• Embodied Cognition• Emergence

Complex Systems Approach• Cognition & Representations :

Complex Systems• Representations are immerged (stables

& non-reactive)• Multi-Agents Systems

Different Approaches for CognitionDifferent Approaches for Cognition

Turing, A. (1936) Newell & Simon, H. A. (1976) Fodor, J. A. (1983)

McCulloch, W. S. & Pitts, W. (1943)Rosenblatt, F. (1962)Rumelhart, D. E. & Norman, D. A. (1981)

Maturana, H. & Varela, F. J. (1973)Varela, F. J., Thompson, E. & Rosch, E. (1991)

van Gelder, T. & Port, R. F. (1995)Thelen, E. & Smith, T. B. (1993)

Mitchell, M. (1998) Steels, L. (2003)Rocha, L. M. & Hordijk, W. (2005)

Carvalho, L. L. & Hassas, S. (2005, 2008)

Page 28: Engineering Self-Organizing Systems Cognition & Emergence of Control Salima Hassas University of Lyon

Systèmes Complexes

↑Interaction de

Fonctions Simples

Systèmes Cognitifs↑

Réactivité Brisée Spontanément

f ( c, a)

Stance Physique

Modèles équationnels, point de vue classique en sciences

C

A

C

B

AB

A’

Stance de Design

Auto-OrganisationAuto-Adaptation

Stance Intentionnelle

Représentations EmergentesAuto- Développement

Pour quoi les représentations sont importantes en psychologie ?

• Les représentations instancient l’acte intentionnel• L’auto-organisation n’assure pas à un système complexe une forme optimale.• L’auto-organisation n’arrive pas seule à guider le développement cognitif des organismes complexes.

f ( c, a)

Approche Systèmes Complexes de la Cognition

Représentations Emergentes : une Approche Multi-Agents des Systèmes Complexes Adaptifs en Psychologie Cognitive 28