Felix Streichert, Christian Spieth, Holger Ulmer, and Andreas Zell

Evolving the Ability of Limited Growth and Self-Repair for Artificial Embryos

Advances in Artificial Life - Proceedings of the 7th European Conference on Artificial Life (ECAL)

Abstract

In this paper we address the problem of limited growth and the difficulty of self-repair in the field of Artificial Embryology. We implemented a topological simulation of multiple cells which is continuous, structure-oriented, with a dynamically connected network of growing cells and endogenous communication between cells. The cell behavior is simulated based on models of gene regulatory networks like Random Boolean Networks and S-systems. Evolutionary Algorithms are used to evolve and optimize the parameters of the models of gene regulatory networks. We compare the performance of Random Boolean Networks and S-systems when optimized by Evolutionary Algorithms on the problem of limited growth and two implementations of cell death and signaling cell death on the problem of self-repair.

Download

[pdf], [ps]

BibTeX 

@inproceedings{streichert03evolving,
  author 		= "Felix Streichert and Chrisitian Spieth and Holger Ulmer and Andreas Zell ",
  title 		= "Evolving the Ability of Limited Growth and Self-Repair for Artificial Embryos",
  booktitle 		= "Advances in Artificial Life - Proceedings of the 7th European Conference on Artificial Life ",
  pages 		= "289-298",
  year 			= "2003",
  month 		= "14-17 September",
  editor 		= "W. Banzhaf and T. Christaller and P. Dittrich and J. T. Kim and J. Ziegler",
  series 		= {LNAI},
  volume 		= {2801},
  address 		= {Dortmund, Germany},
  publisher 		= {Springer Verlag},
  publisher_address 	= {Berlin},  
  keywords 		= {artificial life},
  ISBN 			= {3-540-20057-6},
  abstract 		= {In this paper we address the problem of limited growth and the difficulty of self-repair 
in the field of Artificial Embryology. We implemented a topological simulation of multiple 
cells which is continuous, structure-oriented, with a dynamically connected network of growing 
cells and endogenous communication between cells. The cell behavior is simulated based on 
models of gene regulatory networks like Random Boolean Networks and S-systems. Evolutionary 
Algorithms are used to evolve and optimize the parameters of the models of gene regulatory 
networks. We compare the performance of Random Boolean Networks and S-systems when optimized 
by Evolutionary Algorithms on the problem of limited growth and two implementations of cell 
death and signaling cell death on the problem of self-repair.},
  url			= "http://www.ra.cs.uni-tuebingen.de/mitarb/streiche/publications/streichert03evolving.html"
}