Swarm Intelligence is a form of collective intelligence in swarms. It is a general term of the flexible, adaptive and collective behavior found in swarms, flocks, herds and other groups of social animals, especially for social insects such as ants, termites, locusts, wasps, and honey bees. Examples of systems with swarm intelligence can be found in all kind of animal groups: schools of fish, flocks of birds, herds of land animals.
Definition and History
Systems which exhibit swarm intelligence are composed of many individuals coordinated by decentralized control and self-organization. Swarm Intelligence is traditionally understood as "the emergent collective intelligence of groups of simple agents" (Bonabeau et al. 1999). It is the typical example of emergence and emergent phenomena. The expression "swarm intelligence" was introduced by G. Beni and U. Wang in 1989, in the context of cellular robotic systems. In these systems many simple agents occupy one- or two-dimensional environments to generate patterns and self-organize through nearest-neighbor interactions. Bonabeau, Dorigo and Theraulaz extend Beni et al.'s definition in their "Swarm Intelligence" book, and include any attempt to design algorithms or distributed problem-solving devices inspired by the collective behavior of social insect colonies and other animal societies.
Forms and Relations
|Red Queen Effect|
|Edge of Chaos|
|edit this box|
The concept of swarm intelligence is related to the two basic concepts of stigmergy and emergence, which describe the appearance of organized behavior patterns in groups of individuals. A group of individuals follows the "acts as one" or "acts as swarm" pattern: fireflies flashing in synchrony follow the rule, “I signal when you signal”, fish traveling in schools abide by the rule, “I go where you go”, birds flying in flocks follow the rule "I fly where you fly" and so forth. Coordination of collective behavior is not possible without communication. This communication is of course not always evident and visible. Sometimes the communication takes place in a direct way (for example by visual contact), sometimes the communication happens indirectly with the help of the environment (for example by invisible scents).
There are two basic forms of swarm intelligence: (1) swarm-formation and (2) stigmergy. The term swarm-formation describes the first basic form of swarm intelligence and characterizes the creation of swarms, flocks or groups by direct interaction, as it can be observed in flocks of birds, schoals of fish, etc. The rules are simple: stay close to the group, but don't come too close to individuals. The principle of swarm formation is based on a distinction between global and local, group and agent, swarm and individual: global attraction (move towards the group) combined with local repulsion (stay aways from individuals). It can be described by the boids model of Craig Reynolds.
Stigmergy is the second basic form of swarm intelligence. Swarm intelligence and stigmergy are often used synonymously and describe swarms and groups which are controlled by indirect interaction over the environment, for example in ant colonies. Ant colonies and other social insects use pheromones and scents to communicate with each other. This form of volatile communication allows the dynamic construction of trails for foraging, and enable a good trade-off between exploitation and exploration of food sources in the surrounding environment.
While the principle of stigmergy explains for example trail formation with the purpose of collective foraging, the principle of swarm formation explains the aggregation and group formation with the purpose of collective movement. Both forms and principles can be considered as a case of emergence. Emergence is the general term which describes the appearance of macroscopic phenomena out of mircoscopic interactions.
Large animal groups with coordinated movement such as swarms, herds, schools, and flocks are widespread phenomena in biology. These coherent, large-scale groups often bring negative consequences for the individual (increased competition for resources, disease transmission, and attention from predators), but also benefits (more effective foraging, reproduction, migration, diluting the chance of capture and escape from predators).
The advantage a swarm offers is often in a delicate balance by a disadvantage. The swarm can detect a predator better than an individual, but the predator can also detect a swarm better than a individual animal. A swarm can detect a food source better than an individual, but it also exploits it much faster than any individual
The coordination relies often on a kind of language, although it can be an unusual one. Honeybees perform a dance on their return to the hive, known as bee dance or waggle dance. The information contained in this "dance language" is used to select a new nest site or to exploit a new food source. Ants and other insects communicate mainly through "chemical languages": pheromones and other chemical scents and substances.
- nest building of social insects (wasps, termites: pheromones)
- path finding and food foraging (ants, locusts: pheromones)
- nest site search and nectar foraging (honey bees: waggle dance)
- collective sorting and clustering (ants: change of environment,stigmergy)
- group defense
Among locusts (grashoppers) swarming behaviour is partially a response to overcrowding. When large numbers of locusts are forced together in a small area, they being to form a swarm that can travel long distances consuming lots of vegetation. The critical tipping point is around 20 insects per square metre. Swarms of locusts can be a plague, because they consume vast amounts of crops.
A swarm of locusts allows more effective migration and travel. Each member of the swarm expends less energy than a single locust would spend in normal individual flight. Contrary to individual locusts, locust swarms can travel up to 100km or more in a single day.
Advantages and drawbacks
The fascinating thing about swarm intelligence is that complex collective behavior emerges from individuals following simple rules. Swarm-intelligence and emergence are fascinating because this is not expected or usual behavior, and the system is fault-tolerant and robust. The idea of emergence is that simple underlying rules can give rise to surprisingly complex structures. A system can evolve a large structure from repeated small-scale interactions between its smaller elements. But simple rules do not always lead to complex structures. Individuals following simple rules usually do not organize themselves at all if there is no suitable communication and coordination, they lead to simple collective behavior or chaos and confusion. True self-organization is the EXCEPTION, not the RULE. It happens rarely and not frequently.
Sometimes there are indeed forms and patterns which emerge from simple rules, but usually they are as simple as heaps, dunes or ripples. not more. An ant colony has the form of a heap and the ants are forming "streets". This is it, they are not building any other complex structures. The result of countless repeated, small-scale interactions often is not a big, global structure, rather a big mess. There are only a few interesting examples of self-organization: the flock of birds, the ant colony, etc.
Successfully identifying self-organizing systems with swarm intelligence or emergent properties is a bit like lost finding ancient cities or sunken ships with lots of gold. They are interesting, fascinating and appealing, sometimes even mysterious. Everybody would like to find or have them, but it is not easy, the obstacles are very high, and the interesting cases are very rare and mostly well-known. Therefore the search for both can be frustrating (in exceptional cases also very rewarding and pleasing). Moreover you don't now what you will see until you watch it yourself, and even if you do, in the interesting cases you don't know exactly why it has become this way.
Perhaps swarm intelligence would be less mysterious and fascinating if we could perceive the language which is used to control it directly, for example if we could perceive the odors and chemicals directly which are used by ants for communication. Unlike ants we do not communicate with chemicals.
Scientists and Labs
Software and Optimization
Software and Hardware:
Hardware and Robotics:
Many articles and resources can be found here.
- Ashley J.W. Ward et al., Fast and accurate decisions through collective vigilance in fish shoals, PNAS Vol. 108 No. 6 (2011) 2312-2315
- Iain D. Couzin, Collective cognition in animal groups, Trends Cogn Sci 13 (2009) 36–43.
- G. Beni and U. Wang, Swarm intelligence in cellular robotic systems, In NATO Advanced Workshop on Robots and Biological Systems, Il Ciocco, Tuscany, Italy, 1989.
- Herbert G. Tanner, Ali Jadbabaie and George J. Pappas, Stable Flocking of Mobile Agents, Part I: Fixed Topology, Part II: Dynamic Topology, 2003
- Designing and Understanding Adaptive Group Behavior Maja J. Mataric, 1995
- Andrea Perna et al., Individual rules for trail pattern formation in Argentine ants (Linepithema humile), http://arxiv.org/abs/1201.5827
Dance languages and nest site search of honey bees
- How Self-Organization Evolves, P. Kirk Visscher, Nature Vol. 421 (2003) 799-800
- Collective decisions and cognition in bees, P. Kirk Visscher, Scott Camazine, Nature Vol. 397 (1999) 400
- Dance Language, P. Kirk Visscher, in V.H. Resh and R. T. Cardé, eds. Encyclopedia of Insects. Academic Press (2003) pp. 284-288
- Marco Dorigo, Thomas Stützle, Ant Colony Optimization (2004) The MIT Press, ISBN 0262042193
- Eric Bonabeau, Marco Dorigo, Guy Theraulaz, Swarm Intelligence: From Natural to Artificial Systems (1999) Oxford University Press, ISBN 0195131592
- Scott Camazine et al., Self-Organization in Biological Systems (2003) Princeton University Press, ISBN 0691116245
- James Kennedy and Russell C. Eberhart, Swarm Intelligence, (2001) Morgan Kaufmann, ISBN 1558605959
- Mitchel Resnick, Turtles, Termites, and Traffic Jams: Explorations in Massively Parallel Microworlds (1997) The MIT Press, ISBN 0262181622