self-organisation+in+groups

Previous: TODO Next: Generation of novel invariant features (CHIMPS)

=Self-organisation in moving groups=

An often used example of how simple individual rules can lead to complex behaviour is **self-organisation** in groups of animals. Much studied prototypes include the flocking of birds (e.g. [|BOIDS]), schooling fish (e.g. Hemelrijk [|2012]) and migrating herds (e.g. Couzin and Levin [|2005]). These models usually consider a simple inert (unchangeable) environment, in which at most the physics of the environment are included.

//[|BOIDS] (Reynolds)//
The BOIDS are a classical example of individuals with simple grouping rules in a simple (initially empty) environment. The BOIDS follow three behavioural rules: These three simple rules cause a flock in an empty environment to all move in a certain direction. Moreover, if obstacles are added to the environment the flock can change direction and break up and reform. Hence, these simple rules can cause quite complex behaviour when combined with a non-trivial environment.
 * 1) **Repulsion**: If other individuals are too close to you (within //repulsion zone//), you move away from them.
 * 2) **Alignment**: Adjust your direction to match the average of your neighbours (within //alignment zone//).
 * 3) **Attraction**: Move towards other individuals, as long as they are not too close (within //attraction zone//).

//Decision making in flocks and schools (Couzin et al [|2005])//


Couzin et al (2005) also use the rules of the BOIDS to describe flocks of animals. They then study how well these flocks can move to a certain target, e.g. a food source, if only a small amount of the inidividuals knows where to find this target. These "informed" individuals have directed movement towards t he target, while the other individuals just follow the "regular" rules. Surprisingly, they find that you only need a very low fraction of informed individuals to accurately find the target. Moreover, the larger the group, the lower the fraction that you need. Hence, especially in large groups you only need a small amount of information to find a target. Lastly, they studied the effect of two groups of informed individuals, that have different targets (e.g. half of the informed individuals preferentially moves to the left, the other half to the right). Depending on the parameter conditions, this leads to one of two outcomes: > This happens if the directional movement of the informed individuals is relatively weak, and/or if the two targets are sufficiently close to each other. > This happens if the informed individuals have strong directional movement, and/or if the two targets are far apart.
 * 1) "Averaging": The flock ends up somewhere half-way between the two targets.
 * 1) "The winner takes all": The flock ends up in either one of the targets, i.e. one group of informed individuals "wins".

//Include "Physics of environment": starling flocks (Hemelrijk and Hildebrandt [|2011])//
In the two examples above, the individuals are basically assumed to "float" in space. However, animals are of course subjected to many physical forces, that lay constraints on their movement possibilities. For instance, flying birds have to maintain a certain speed to avoid falling from the air because of gravity, and animals moving at speed in a certain direction cannot immediately turn around because of their inertia. Hemelrijk and Hildebrandt (2011) developed a model of flocking starlings that incorporates these constraints. However, other than these physical constraints they only implement BOID-like, simple grouping rules. The behaviour of the starling flocks in their model closely resembles observations of real flocks of starlings, again illustrating that only a simple set of rules and obvious physical constraints can already explain the behaviour. Furthermore, Hemelrijk and Hildebrandt have used their model to test the influence and relative importance of model parameters such as locality of the interactions and speed variability, "experiments" that could never be done in nature.

Although all these examples give us a first "WOW" over what simple behavioural rules can do, they all consider a relatively simple environment. Next, we will study the effect of interaction with a more complex environment, that can also be changed by the individuals.

Next: Generation of novel invariant features (CHIMPS)


 * References**


 * Hemelrijk CK and Hildebrandt H** (2012). Schools of fish and flocks of birds: their shape and internal structure by self-organization. Interface Focus 2(6): 726-37.
 * Couzin ID, Krause J, Franks NR, Levin SA** (2005). Effective leadership and decision-making in animal groups on the move. Nature 433(7025):513-6.
 * Hemelrijk CK and Hildebrandt H** (2011). Some causes of the variable shape of flocks of birds. PLoS ONE 6(8):e22479.


 * Couzin ID** (2009). Collective cognition in animal groups. Trends in Cognitive Science 13(1):36-43. [Interesting review]

(CHANGELOG 2014-2015) - Added page