Swarm Intelligence

Back Home Next

 

Swarm 

Ants Carrying a Leaf

 

from Scientific American March 2000.

“Insects that live in colonies – ants, bees, wasps, termites – have long fascinated everyone from naturalists to artists.  Maurice Maeterlinek, the Belgian poet, once wrote, “What is that governs here? What is it that issues orders, foresees the future, elaborates plans and preserves equilibrium?” These indeed are puzzling questions.” [Bonabeau, “Swarm Smarts”, pg73]

 

 

What is a swarm? What is swarm intelligence? How can it be applied to solve problems?  A swarm is a large number of insects or other organisms in motion.  "Swarm Intelligence is a property of systems of unintelligent agents exhibiting collectively intelligent behavior."[White, 1997]

  Swarm Intelligence is modeled after the study of behavioral patterns of biological entities that live in colonies like insects, bees, wasps, termites and ants.  Swarm agents try to imitate in software the behavior of these insects one by one and then, when grouped together they can solve multiple problems.  In order to model correctly these insects, the modeler must analyze in detail not only the actions and patterns of a single insect, but also of a whole swarm of insects. 

 

The power of Ants

To be able to fully understand the power of Swarm Intelligence and how it can be applied to our network routing problem, we have to study what a swarm of insects can do by working together in a community.   If we look closely at one single ant, we can observe that it's a simple unsophisticated insect that performs repetitive actions, which don't show much signs of cognition nor intelligence.  The single ant seems to have its own agenda.  Now if we look at a collection of ants, we can see that these first overlooked insects can manage when working together important tasks and seem to have a special link or communication system between them.  Several studies of ants and ant colonies have shown that they can collectively solve problems and perform complicated tasks.  Several are shown below: 

Nest building and maintenance.
Forming bridges.
Emigration of a colony.
Sorting food and food items.
Carrying large items by cooperating among each other.
Finding the shortest routes from their nest to their food supplies.
Exploiting first the richest food source available while keeping track of other food options.
Regulating nest temperatures up to a single degree in variance.

Ants show two basic types of stigmergy (indirect communication through the environment):

 ·        Sematectonic: actions of one agent directly solving problems, which affect other agents.

·        Sign based: actions of agents affecting the environment.

 Swarm Intelligence is modeled after the study of behavioral patterns of biological entities that live in colonies like insects, bees, wasps, termites and ants.  Swarm agents try to imitate in software the behavior of these insects one by one and then, when grouped together they can solve multiple problems.  In order to model correctly these insects, the modeler must analyze in detail not only the actions and patterns of a single insect, but also of a whole swarm of insects. 

 The swarm has the following characteristics:

 ·        Decentralized: the system has no central control, it is highly adaptive and can adapt fast to changes in the environment.

·        Ecologist view of perception:  the entities in the system have no pre-built or defined information about the environment.  They "learn as the go".  This is very important in networking because as systems are added into the network, the routers will have to learn and adapt to the changes, in other words, be dynamic.  If routing agents are modeled after swarms they can obtain these properties and apply them as needed..

 

Examples of Swarm Intelligence

Material Transport with ant like robots:  the task of carrying loads is taken by different autonomous robots who don’t directly communicate with each other but who collectively transport a load of merchandize.  Although each individual robot is too small to carry the item by itself, the collective use of many robots is very productive.
Particle Swarm Optimization: swarm intelligence is used to optimize continuous non – linear functions.  The key advantage of particle swarm optimization is that it is in computing terms, very simple to implement and inexpensive. 
Particle Swarm for Voltage and Power Control:  it is very important to regulate power in systems and to have measurements as accurate as possible.  Using swarm, the system is able to monitor power activities locally instead of having a centralized controller, thus reducing the risk of accidents due to slow measurements. 
Statistical calculation and analysis:  swarm intelligence can be utilized to improve statistical calculations.  For example, in a system to calculate the emergence of new political parties, swarm is utilized to demonstrate the formation and interaction of coalitions that arise from each individual voter’s preference. 
Back to Top

Copyright 2001: Ivan A. Escobar Broitman
For problems or questions regarding this web contact ivan@escobarb.fsnet.co.uk
Last updated: August 19, 2001.