Computational Science Technical Note CSTN-067


Complex Emergent Behaviour from Evolutionary Spatial Animat Agents

K. A. Hawick and C. J. Scogings

Archived January 2009


Spatial multi-agent systems provide a powerful model framework for investigating evolutionary behaviour amongst animat agents. We have developed a microscopic animat-based model in which autonomous agents are microscopically controlled by a rule-set that can be evolved using suitable operators. Our system can support over a million animats co-existing over many generations and has already been used to explore several collective phenomena including clustering; segregation; tribal warfare and battlefront formation. We incorporate simple microscopic behaviour rules based on local view information, that determine animat feeding; breeding; movement; seeking; and avoidance. We use a simple agent state model consisting of position, current health and age and a genetic code for each animat. We find a number of spatially-rich and complex, emergent patterns from the microscopic model and discuss how the model's convergence to stable macroscopic behaviour cycles is related to the localised rule parameters. We illustrate how an animat agent population of predators and prey can evolve more effective individuals by applying genetic algorithms to the species rule-sets and how our model framework and approach can be applied to sociological and predatory phenomena.

Keywords: evolution; agent architecture; convergence; complexity analysis.

Full Document Text: PDF version.

Citation Information: BiBTeX database for CSTN Notes.

BiBTeX reference:

  chapter = {Complex Emergent Behaviour from Evolutionary Spatial Animat Agents},
  pages = {139-160},
  title = {Agent-Based Evolutionary Search},
  publisher = {Springer},
  year = {2010},
  editor = {R. Barker and T. Ray},
  author = {K. A. Hawick and C. J. Scogings},
  number = {ISBN 978-3-642-13424-1},
  month = {January},
  note = {CSTN-067},
  doi = {10.1007/978-3-642-13425-8},
  institution = {Massey University},
  timestamp = {2009.02.28}

[ CSTN Index | CSTN BiBTeX ]