swarm intelligence systems

36
Swarm Swarm Intelligence Intelligence Systems Systems —————————————— —————————————— Christian Jacob Christian Jacob [email protected] Department of Computer Science University of Calgary

Upload: others

Post on 12-Sep-2021

8 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Swarm Intelligence Systems

SwarmSwarmIntelligenceIntelligence

SystemsSystems————————————————————————————

Christian JacobChristian [email protected]

Department of Computer ScienceUniversity of Calgary

Page 2: Swarm Intelligence Systems

CellularCellularAutomataAutomata

GlobalGlobal Effects from Local Effects from LocalRulesRules

Page 3: Swarm Intelligence Systems

Cellular AutomataCellular Automata

✦✦ The CA space is a lattice of cells with aThe CA space is a lattice of cells with aparticular geometry.particular geometry.

✦✦ Each cell contains a variable from aEach cell contains a variable from alimited range (e.g., 0 and 1).limited range (e.g., 0 and 1).

✦✦ All cells update synchronously.All cells update synchronously.

✦✦ All cells use the same updating rule,All cells use the same updating rule,depending only on local relations.depending only on local relations.

✦✦ Time advances in discrete steps.Time advances in discrete steps.3

Page 4: Swarm Intelligence Systems

One-dimensional finite CA architectureOne-dimensional finite CA architecture

time

✦✦ K = 5 localK = 5 localconnectionsconnectionsper cellper cell

✦✦ SynchronousSynchronousupdate in discreteupdate in discretetime stepstime steps

A. Wuensche: The Ghost in the Machine, Artificial Life III, 1994. 4

Page 5: Swarm Intelligence Systems

Cellular Automata:Cellular Automata:Local Rules — Global EffectsLocal Rules — Global Effects

5

--> 1D-CA Demos--> 1D-CA Demos

Page 6: Swarm Intelligence Systems

Time Evolution of the Time Evolution of the iithth Cell Cell

C f C C C C Cit

i Kt

it

it

it

i Kt( )

[ / ]( ) ( ) ( ) ( )

[ / ]( )( ,..., , , ,..., )+

− − + +=12 1 1 2

With periodic boundary conditions:

x C Cx N x< = +1: x N C C Nx x> = −:

6

Page 7: Swarm Intelligence Systems

Value Range and Update RulesValue Range and Update Rules

✦✦ For For VV different states (= values) per cell different states (= values) per cellthere are there are VVKK permuations permuations of values in aof values in aneighbourhood neighbourhood of size K.of size K.

✦✦ The update function The update function ff can be can beimplemented as a lookup table with Vimplemented as a lookup table with VKK

entries, giving Ventries, giving VVVKK possible rules. possible rules.

7

Page 8: Swarm Intelligence Systems

Example Update RuleExample Update Rule

✦✦ V = 2, K = 3V = 2, K = 3

✦✦ The rule table for rule The rule table for rule 3030::

111 110 101 100 011 010 001 000111 110 101 100 011 010 001 000 0 0 0 1 1 1 1 00 0 0 1 1 1 1 0

8

See examples ...

Page 9: Swarm Intelligence Systems

2-D CA:2-D CA:Emergent Pattern FormationEmergent Pattern Formationin Excitable Mediain Excitable Media

Neuron excitationNeuron excitation

Neuron excitation (relaxed)Neuron excitation (relaxed)

HodgepodgeHodgepodge

9

Page 10: Swarm Intelligence Systems

Cellular Automata

Random BooleanNetworks

Classifier Systems

SwarmSystems

10

Page 11: Swarm Intelligence Systems

Hölldobler & Wilson, 1990

AntsAnts

Page 12: Swarm Intelligence Systems

Hölldobler & Wilson, 1990

Self-organizationTeam workCompetition...andHeavy Loads

Page 13: Swarm Intelligence Systems

Hölldobler & Wilson, 1990Teamwork Living Architecture

Page 14: Swarm Intelligence Systems

Hölldobler & Wilson, 1990

Living BridgeLiving Network

Page 15: Swarm Intelligence Systems

Hölldobler & Wilson, 1990Together we are strong ...

Page 16: Swarm Intelligence Systems

Ant ForagingAnt ForagingBehaviourBehaviour

Learning about EmergentLearning about EmergentSystemSystem Behaviours Behaviours

Page 17: Swarm Intelligence Systems

Experimental setupfor studying ant foreaging behaviour

Ant Ant ForeagingForeaging and Shortest Paths and Shortest Paths

Bonabeau et al., 1999

Page 18: Swarm Intelligence Systems

Shortest Path DiscoveryShortest Path Discovery

(a) Ants walking between nest(a) Ants walking between nestand food sitesand food sites

(b) An obstacle is placed in the(b) An obstacle is placed in themiddle.middle.

(c) Ants turn left or right, while(c) Ants turn left or right, whiledropingdroping pheromone ... pheromone ...

(d) … and finally the shortest(d) … and finally the shortestpath emerges.path emerges.

Page 19: Swarm Intelligence Systems

Adaptation to EnvironmentalAdaptation to EnvironmentalChangesChanges

(a) The newly found shortest path(a) The newly found shortest path

(b) Moving the obstacle(b) Moving the obstacle (c) Discovery of new shortest path(c) Discovery of new shortest path

Page 20: Swarm Intelligence Systems

MassivelyMassively Parallel ParallelMicro WorldsMicro Worlds

StarLogoStarLogo

Mitchel ResnickMitchel Resnick (MIT, 1997) (MIT, 1997)

Page 21: Swarm Intelligence Systems

Agent-Based EvolutionAgent-Based Evolution

✦✦ Massive ParallelismMassive Parallelism

✦✦ Interacting AgentsInteracting Agents

✦✦ CooperationCooperation

✦✦ CompetitionCompetition

✦✦ Emergent SystemEmergent System Behaviour Behaviour

StarLogo Demo

Page 22: Swarm Intelligence Systems

Emergent SystemEmergent System Behaviour BehaviourSimulatedSimulated

Ant ForagingAnt Foraging

CollectiveCollectiveForagingForaging

EquidistantEquidistantFood SitesFood Sites

Randomly DistributedRandomly DistributedFood SitesFood Sites

Page 23: Swarm Intelligence Systems

Emergent SystemEmergent System Behaviour BehaviourSimulatedSimulated

Ant ForagingAnt Foraging

to look-for-foodif not carrying-food? [ifelse (ask patch-here [pheromone]) < 0.2 [right random 40 left random 40] [set-heading uphill pheromone] forward 1]end

to find-foodif (not carrying-food?) and ask patch-here [food > 0] [set-carrying-food? True ask patch-here [set-food food - 1] set-drop-size 35 right 180 forward 1]end

to return-to-nestif carrying-food? [ask patch-here [add-pheromone-drop] set-drop-size drop-size - 0.6 set-heading uphill nest-scent forward 1]end

to find-nestif carrying-food? and ask patch-here [nest?] [set-carrying-food? False right 180 forward 1]end

Page 24: Swarm Intelligence Systems

DemoDemo

FollowingFollowing

BehaviourBehaviour

Page 25: Swarm Intelligence Systems

InteractionsInteractionsamongamong

Social InsectsSocial Insects

Page 26: Swarm Intelligence Systems

Interactions among Social InsectsInteractions among Social Insects

✦✦ Direct InteractionsDirect Interactions–– Food or liquid exchangeFood or liquid exchange

–– Visual or tactile, or Visual or tactile, or scentuousscentuous contactcontact

✦✦ Indirect InteractionsIndirect Interactions: : StigmergyStigmergy–– PheromonesPheromones

–– IndividualIndividual behaviour behaviour modifies themodifies theenvironment (e.g., by putting up environment (e.g., by putting up signs =signs =stigmastigma),), which in turn modifies the which in turn modifies thebehaviour behaviour of other individuals.of other individuals.

Page 27: Swarm Intelligence Systems

DemoDemo

ShepherdsShepherds

andand

SheepSheep

Page 28: Swarm Intelligence Systems

DemoDemo

StigmergyStigmergyinin

ActionAction

Bonabeau et al., 1999

Page 29: Swarm Intelligence Systems

ComplexComplex Systems Systems

EmergentEmergentBehaviours andBehaviours and Patterns Patterns

fromfromLocal InteractionsLocal Interactions

Page 30: Swarm Intelligence Systems

Stevens et al., 1988

Page 31: Swarm Intelligence Systems

Nuridsany & Pérennou, 1996

Page 32: Swarm Intelligence Systems

Ernst, 1998

Nuridsany & Pérennou, 1996

Page 33: Swarm Intelligence Systems

What to Learn from AntWhat to Learn from AntColonies as Complex SystemsColonies as Complex Systems

✦✦ Fairly simple units generateFairly simple units generatecomplicated global complicated global behaviourbehaviour..

✦✦ “If we knew how an ant colony works,“If we knew how an ant colony works,we might understand more about howwe might understand more about howall such systems work, from brains toall such systems work, from brains toecosystems.”ecosystems.”(Gordon, 1999)(Gordon, 1999)

Page 34: Swarm Intelligence Systems

Emergence in Complex SystemsEmergence in Complex Systems

✦✦ How do How do neuronsneurons respond to each other respond to each otherin a way that produces thoughts?in a way that produces thoughts?

✦✦ How do How do cellscells respond to each other in a respond to each other in away that produces the distinct tissues ofway that produces the distinct tissues ofa growing embryo?a growing embryo?

✦✦ How do How do speciesspecies interact to produce interact to producepredictable predictable changeschanges, over time, in, over time, inecological communities?ecological communities?

✦✦ ......

Page 35: Swarm Intelligence Systems

Swarm SystemsSwarm SystemsProvidingProviding NewNew Insights ... Insights ...

Page 36: Swarm Intelligence Systems

ReferencesReferences

✦ Bonabeau, E., Dorigo, M., and Theraulaz, G. (1999). SwarmIntelligence: From Natural to Artificial Systems. New York, OxfordUniversity Press.

✦ Ernst, A. M., ed. (1998). Digest: Kooperation und Konkurrenz,Heidelberg, Spektrum Akademischer Verlag.

✦ Gordon, D. (1999). Ants at Work. New York, The Free Press.✦ Hölldobler, B., and Wilson, E. O. (1990). The Ants. Cambridge,

MA, Harvard University Press.

✦ Nuridsany, C., and Pérennou, M. (1996). Microcosmos: TheInvisible World of Insects. New York, Stewart, Tabori & Chang.

✦ Resnik, M. (1997). Turtles, Termites, and Traffic Jams.Cambridge, MA, MIT Press.

✦ Stevens, C. F., et al. (1988). Gehirn und Nervensystem.Heidelberg, Spektrum Akademischer Verlag.