Modelling Flocks of Birds from the Bottom Up

Abstract

We argue that compositional specification based on formal languages can facilitate the modelling of, and reasoning about, sophisticated collective behaviour in many natural systems. One defines a system in terms of individual components and local rules, so that collective behaviours emerge naturally from the combined effect of the different actions of the individual components. With appropriate linguistic constructs, this can yield compact and intuitive models that are easy to refine and extend in small incremental steps. In addition, automated workflows implemented on top of this methodology can provide quick feedback, thereby allowing rapid design iterations. To support our argument, we consider flocking, a well-known example of emergent behaviour in collective adaptive systems. We build a minimalistic bottom-up model of a flock of birds incrementally, discussing specific language constructs as we go along. We then describe a prototype simulator, and use it to validate our model in a controlled experiment, where a flock is attacked by a bird of prey. The flock effectively reacts to the attack by splitting into smaller groups and regathering once the threat subsides, consistently with both natural observations and previous models from the literature.