Swarm Intelligence techniques for task allocation and sub-task merging in multi-agent systems
Ampatzis Christos

Abstract:

In this work, we address the problem of synthesizing non-reactive controllers for a swarm robotic system, called swarm-bot, using Arti cial Evolution. In particular, we evolve simple dynamical neural networks, in order to achieve autonomous decision-making agents that are able to integrate over time their perceptual experience. These agents cooperate in carrying out certain tasks by communicating their experience to the rest of the group. We show the applicability of our desision-making mechanisms in the realisation of a complex scenario.