Swarm Intelligence Book Chapter



“the swarm seems more intelligent than the individual in it”; insects:
“the swarm exhibits behavior that individual members cannot”
Stigmergy: indirect communication (using signs or clues placed in the
environment)
SWI: Cooperative efforts of groups of agents….
•
•
•
•
•


Agents are autonomous and operate in parallel
Little or no centralized control
Stigmergy (indirect communication)
Usually use simple rules
…
Learning from foraging: http://en.wikipedia.org/wiki/Foraging
Stigmergy:
•
•
•
Ch. Eick: SWI Book Chapter
Stimulating material triggers response in the termite workers; as the
configuration changes, this change triggers a different response
Pheromone (already covered before)
Nest-building: The Structure influences the actions of the individual that
reads it
Swarm Intelligence Book Chapter II

Natural Clustering
1. Ant cemeteries; ant eggs and micro larvae are sorted in rings by age
2. Simple Agent Rule for clustering (as witnessed in the NetLogo Program):
a. Pick up single corpse
b. Move away
c. Find an area with high density corpses

Termites build highly complex structures. How do they do it
•
•
•

Micro rules
Patterns observed by termites and other sensual input act as sigmergic
trigger
I did not quite understand Figure4? Did you?
Swarm robots: Small simple robots furnished with relatively nonsophisticated ways to communicate.
Ch. Eick: SWI Book Chapter
Swarm Intelligence Book Chapter III

Foraging
•
•
•
Ch. Eick: SWI Book Chapter
“Get food without wasting too much energy”
Bacterial Foraging:
– Chemotaxis: Movement along a chemical gradient
– Bacterium has two kind of movements: Tumble and run
– When encountering an increasing gradientmore running
– Additionally, bacteria secrete chemicals that attract each other like in
particle swarms
– In summary uses some combination of exploration and exploitation
Bee Foraging:
– Waggle dance (http://www.youtube.com/watch?v=-7ijI-g4jHg )
• Straightline Movement tells where the food is
• Duration of dance tells “quality/quantity of food source”
• Bees are allowed to sample the nectar (not during the dance)
• Depending on the quality of the dance other bees are or are not
recruited in exploring the food source
• If the food source is good multiple bees will be performing the
same dance.
Swarm Intelligence Book Chapter IV

Flocking (already covered)
PSO (already covered). Simple version: New Vector (speed +
direction) is computed dependent on
• Current Vector
• Personal Best multiplied by a random number in [0,2]
• Global Best multiplied by a random number in [0,2]
and the position of the particle is updated by adding this vector to
the current position
• PSO provides a significant contribution to optimization
practice
 Current Trends:
• Hybrid algorithms
• A lot of new paper that approach optimization and multiobjective optimization using Multiple Swarms
• Some work that borrows ideas from bacterial foraging
algorithms
Ch. Eick: SWI Book Chapter
