Swarm intelligence: from natural to artificial systems
Swarm intelligence: from natural to artificial systems
Proceedings of the 29th annual conference on Computer graphics and interactive techniques
| Hi-index | 0.00 |
We investigate an artificial self-organizing multiparticle (also multi-agent or swarm) system consisting of many (up to 103) reactive, mobile agents. The agents' movements are governed by a few simple rules and interact indirectly via a pheromone field. The system generates a wide variety of complex patterns. For some parameter settings this system shows a notable property: seemingly never-ending, dynamic formation and reconfiguration of complex patterns. For other settings, however, the system degenerates and converges after a transient to patterns of low complexity. Therefore, we consider this model as an example of a class of self-organizing systems that show complex behavior mainly in the transient. In a first case study, we inspect the possibility of using a standard genetic algorithm to prolongate the transients. We present first promising results and investigate the evolved system.