Abstract

The global behaviour of foraging ants, swarming bees, flocking birds, or schooling fish is a compelling nature's delight. While the collective behaviour of the swarm is so coherent, individual members of the group are very simple, autonomous, and limited in capabilities, memory and communication radii. Research in swarm intelligence indicates that this aggregated emergent behaviour can be simulated using virtual agents modelled on swarming agents in nature.

We study collections of virtual agents modelled on flocking birds, schooling fish, herds and foraging ants, investigating causes of emergent behaviour. The virtual agents we study can only communicate locally and indirectly via the environment (stigmergy communication). The key focus is on evaluating relationships between emergent behaviour and the activities of individual agents with the aim of prescribing specific routines (methods) as causes of specific emergent behaviour. Consequently, generalized routines are deduced and used in predicting other emergent formations (simply by altering  routine parameters). This research is of direct relevance in further applications of swarm intelligence metaphors, especially in complex problem solving, nanotechnology, material science and general agent control designs.