Analysis of swarm behavior using compound eye and neural network control

Abstract

The emergent collective intelligence of groups of simple agents known as swarm intelligence is a new exiting way of achieving a form of artificial intelligence. This paper studies a formal model for swarm intelligence inspired by biological swarms found in nature. Software agents are used to model the individuals of a swarm. Each agent is controlled by a neural network that processes position data from the others in its visible zone given by a compound eye and in this way navigates in 3D space. An additional input parameter is used to represent the agent's motivation to form a swarm. Simulations with different motivation parameters exhibit remarkable agent formations that can be considered as biologically plausable. Several ways to improve the model are discussed.