Complex Systems

Evolving Distributed Control for an Object Clustering Task Download PDF

Timothy D. Barfoot
Controls and Analysis,
MDA Space Missions,
9445 Airport Road,
Brampton, Ontario L6S 4J3, Canada

Gabriele M. T. D'Eleuterio
Institute for Aerospace Studies,
University of Toronto,
4925 Dufferin Street,
Toronto, Ontario M3H 5T6, Canada


Motivated by social insects, the possibility of evolving distributed control for a task requiring global coordination is investigated. The task is object clustering. A key aspect of this work is that a population of robot-like agents is allowed to select the cluster location. A detailed examination of how solutions evolved by a genetic algorithm are able to scale as key parameters are varied is presented, allowing commentary on the sensitivity of the evolved solution to changes in the environment. In most of the scaling experiments, the solution degrades gracefully about the evolutionary design point. However, in the case of constant-density scaling, the solution maintains its effectiveness as the problem is made larger.