Course organization

• Subjects
Swarm intelligence basics
Swarm intelligence •
• Data clustering
http://cs.ulb.ac.be/public/teaching/infoh414 • Particle swarm optimization
• Ant colony optimization
Marco Dorigo
FNRS Research Director • Swarm robotics
IRIDIA
• Division of labour in swarms
Université Libre de Bruxelles
• Projects:
• Ant colony optimization
• Swarm robotics
2
© Marco Dorigo © Marco Dorigo

Teaching material
Exam
• PDF of slides used in lessons:
• 7 minutes for powerpoint presentation of
http://cs.ulb.ac.be/public/teaching/infoh414
results of your project (50%)
• Books:
• Questions on any subject covered during • E. Bonabeau, M. Dorigo and G. Theraulaz, Swarm Intelligence,
the course (50%) Oxford University Press, 1999
• M. Dorigo and T. Stützle , Ant Colony Optimization, MIT Press, 2004
• Both parts must be above threshold
• Software:
• Project delivery: one week before exam. • ARGoS simulator: http://www.argos-sim.info/
1 points lost for each interval of 12 hours • ACO software:
delay. Max delay 3 days (no exam after http://www.aco-metaheuristic.org/aco-code/public-software.html

this).

3 4
© Marco Dorigo © Marco Dorigo
 What is swarm intelligence?
Swarm intelligence is the complex
global behavior shown by a
distributed system that arises from
the self-organized local interactions
between its constituent agents
5 6
© Marco Dorigo
Insects, Social Insects, and Ants
• 1018 living insects (rough estimate)
• ~2% of all insects are social
• Social insects are:
– All ants
– All termites
– Some bees
– Some wasps
• 50% of all social insects are ants
• Avg weight of one ant between 1 and 5 mg
• Tot weight ants ~ Tot weight humans
• Ants have colonized Earth for over 100 million years,
Homo sapiens sapiens for approximately 50,000
years
© Marco Dorigo

Ants Ants

• Fungus growers
Leaf cutter,
fungus growing
• Breeding ants
ants • Weaver ants
Breeding ants • Harvesting ants
• Army ants
Weaver ants
• Slavemaker ants

7 8
© Marco Dorigo © Marco Dorigo
 Ant Colony Societies How Do Ants and Social Insects
Coordinate their Activities?
• Ant colony size: from as few as 30 to
millions of workers • Self-organization:
• Work division: - Set of dynamical mechanisms whereby structure
- Reproduction --> queen appears at the global level as the result of
interactions among lower-level components
- Defense --> soldiers
- The rules specifying the interactions among the
- Food collection --> specialized workers system's constituent units are executed on the
- Brood care --> specialized workers basis of purely local information, without
- Nest brooming --> specialized workers reference to the global pattern, which is an
- Nest building --> specialized workers emergent property of the system rather than a
property imposed upon the system by an external
ordering influence

9 10
© Marco Dorigo © Marco Dorigo

Self-organization How Do Social Insects Achieve

Self-organization?
Four basic ingredients: • Communication is necessary
• Multiple interactions
• Two types of communication:
• Randomness
- Direct: antennation, trophallaxis (food or
• Positive feedback liquid exchange), mandibular contact,
- E.g., recruitment and reinforcement visual contact, chemical contact, etc.
• Negative feedback - Indirect: two individuals interact indirectly
when one of them modifies the
- E.g., limited number of available foragers environment and the other responds to
the new environment at a later time
This is called stigmergy

11 12
© Marco Dorigo © Marco Dorigo
 Stigmergy
 Stigmergy
Grassé P. P., 1959

Stimulation of workers
• “La coordination des taches, la regulation des constructions ne
dependent pas directement des oeuvriers, mais des by the performance
constructions elles-memes. L’ouvrier ne dirige pas son they have achieved
travail, il est guidé par lui. C’est à cette stimulation d’un type
particulier que nous donnons le nom du STIGMERGIE Grassé P. P., 1959
(stigma, piqure; ergon, travail, oeuvre = oeuvre stimulante).”
Grassé P. P., 1959

• [“The coordination of tasks and the regulation of constructions

does not depend directly on the workers, but on the
constructions themselves. The worker does not direct his
work, but is guided by it. It is to this special form of
stimulation that we give the name STIGMERGY (stigma, sting;
ergon, work, product of labour = stimulating product of
labour).”]

13 14
© Marco Dorigo © Marco Dorigo

Stigmergy

Example: Termites Building their Termites’ Nest
Nest

1 h 18’ 3 h 23’ 5 h 15’ 8 h 13’

after start after start after start after start

15 16
© Marco Dorigo © Marco Dorigo
 Sign-based Stigmergy
 Pheromone Trail Following
Example: Trail Following in Ants
Ants and termites follow pheromone trails

While walking, ants and termites:

• May deposit a pheromone on the ground
• Follow with high probability pheromone
trails they sense on the ground

17 18
© Marco Dorigo © Marco Dorigo

Quantitative vs. Qualitative Wasps 


Stigmergy Building a Nest

Termites Wasps Video recordings by Guy Theraulaz

19 20
© Marco Dorigo © Marco Dorigo
 Artificial Nest Building
Wasps’ Nests Theraulaz & Bonabeau, 1995
Photos by Guy Theraulaz

Sensory field

21 22
© Marco Dorigo © Marco Dorigo

Some Simulation
Results More Simulation Results (1)
Theraulaz & Bonabeau, 1995

23 24
© Marco Dorigo © Marco Dorigo
 More Simulation Results (2) Types of Stigmergy
Theraulaz & Bonabeau, 1995

• Sematectonics
E.g., termites nest building
• Sign-based
E.g., ants trail following behavior
• Quantitative
E.g., ants trail following behavior and
termites nest building
• Qualitative
E.g., social wasps nest building

25 26
© Marco Dorigo © Marco Dorigo

“Artificial” Stigmergy What is swarm intelligence?

Indirect communication mediated by
modifications of environmental states Swarm intelligence: “Any attempt to
which are only locally accessible by design algorithms or distributed
the communicating agents problem-solving devices inspired by
Dorigo & Di Caro, 1999 the collective behavior of
social insect colonies
and other animal societies”
• Characteristics of artificial
stigmergy: From “Bonabeau E., M. Dorigo & G. Theraulaz, Swarm
Intelligence: From Natural to Artificial Systems, Oxford
- Indirect communication University Press, Oxford University Press, 1999, page 7”.

- Local accessibility
27 28
© Marco Dorigo © Marco Dorigo
 Swarm intelligence
What is swarm intelligence? Swarm intelligence
• Swarm intelligence is an artificial intelligence technique
based around the study of collective behavior in Distinguish between
decentralized, self-organized systems
• Scientific swarm intelligence
• Swarm intelligence systems are typically made up of a
population of simple agents interacting locally with one • Engineering swarm intelligence
another and with their environment
• Although there is normally no centralized control structure
dictating how individual agents should behave,
local interactions between such agents often
lead to the emergence of global behavior
• Examples of systems like this can be found in nature,
including ant colonies, bird flocking, animal herding, and fish From “Scholarpedia, Swarm Intelligence”
schooling

29 30
© Marco Dorigo © Marco Dorigo

Swarm intelligence Swarm intelligence

Swarm intelligence Swarm intelligence

Distinguish between Distinguish between

• Scientific swarm intelligence • Engineering swarm intelligence
is concerned with the understanding of is concerned with the design and
natural swarm systems implementation of artificial swarm
systems

31 32
© Marco Dorigo © Marco Dorigo
 Swarm intelligence Swarm intelligence

Swarm intelligence Some examples of

swarm intelligence systems
Engineering swarm intelligence
• Characteristics of swarm intelligence systems:
takes inspiration from
- Multi-agent
scientific swarm intelligence studies
- Individuals are modeled as having
to design problem-solving devices stochastic behavior
- Individuals use only local information
- Self-organized and distributed control

33 34
© Marco Dorigo © Marco Dorigo

Scientific swarm intelligence Scientific swarm intelligence

Example: Example:
Finding the shortest path Cooperative transport

• Multi-agent • Multi-agent
• Individuals use only local • Individuals use only local
information information
• Stochastic individuals • Stochastic individuals
• Distributed control • Distributed control

Video by J.-L. Deneubourg

35 36
© Marco Dorigo © Marco Dorigo
 Scientific swarm intelligence Scientific swarm intelligence

Example: Example:
Building a “bridge” Building a “bridge”

• Multi-agent • Multi-agent
• Individuals use only local • Individuals use only local
information information
• Stochastic individuals • Stochastic individuals
• Distributed control • Distributed control

Video by A. Lioni

37 38
© Marco Dorigo © Marco Dorigo

Scientific swarm intelligence Scientific swarm intelligence

Example: Example:
Building a “bridge” Building a “bridge”

• Multi-agent n Multi-agent
• Individuals use only local n Individuals use only
information local information
• Stochastic individuals n Stochastic
individuals
• Distributed control
n Distributed control

39 40
© Marco Dorigo © Marco Dorigo
 Scientific swarm intelligence

Example: Flocking birds

• Multi-agent n Multi-agent
• Individuals use only local n Individuals use only
information local information
• Stochastic individuals n Stochastic
individuals
• Distributed control
n Distributed control

41 42
© Marco Dorigo © Marco Dorigo

Scientific swarm intelligence

Example: Fish school

• Multi-agent n Multi-agent
• Individuals use only local n Individuals use only
information local information
• Stochastic individuals n Stochastic
individuals
• Distributed control
n Distributed control

43 44
© Marco Dorigo © Marco Dorigo
 How to design How to design
swarm intelligence systems swarm intelligence systems
• Essentially two ways: • Essentially two ways:

- researcher ingenuity - researcher ingenuity

- examples:
- machine learning techniques - ant colony optimization, ant-based
clustering
- machine learning techniques
- (+ researcher ingenuity)
- example: swarm robotics

45 46
© Marco Dorigo © Marco Dorigo

Engineering swarm intelligence Engineering swarm intelligence

Research method Research method

•
Observe a social behavior
Build a simple model to explain it
•

•
Observe a social behavior
Build a simple model to explain it
] biologists

• Use the model of the social behavior as a source of • Use the model of the social behavior as a source of
inspiration for solving a practical problem that has inspiration for solving a practical problem that has
some similarities with the observed social behavior some similarities with the observed social behavior

47 48
© Marco Dorigo © Marco Dorigo
 Engineering swarm intelligence From scientific to engineering swarm intelligence

Research method Examples

• Observe a social behavior • Cemetery organization ➠ data clustering

and brood sorting
• Build a simple model to explain it
Use the model of the social behavior as a source of
]
•
• Birds flocking ➠ particle swarm optimization
inspiration for solving a practical problem that has
some similarities with the observed social behavior
• Foraging ➠ ant colony optimization

• Self-assembly and ➠ robotic implementations

Computer scientists, engineers, cooperative transport
operation researchers, roboticists
• Division of labor ➠ adaptive task allocation
49 50
© Marco Dorigo © Marco Dorigo

51 52
© Marco Dorigo © Marco Dorigo