The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge


Research Paper #865

Title:Artificial Lampreys: Comparing Naturally and Artificially Evolved Swimming Controllers
Authors:Ijspeert,AJ; Hallam,JC; Willshaw,D
Date:Aug 1997
Presented:To appear in the Proceedings of the 4th European Conference on Artificial Life, ECAL97, Brighton
Keywords:
Abstract:This paper studies and compares naturally and artificially evolved neural networks controlling the motion of an animal: the anguilliform swimming of a lamprey. The swimming controller of the lamprey extensively studied and modeled by Grillner and Ekeberg, is reproduced. A real number genetic algorithm is used to develop alternative artificial controllers composed of neurons similar to those of the biological model. We examine the feasibility of using Genetic Algorithms to evolve neural controllers and evaluate the quality of the evolved networks by comparing them with the biological controller. Results show that artificial controllers can be obtained with architectures other than the biological one, and which, in some ways, are more efficient.
Download:POSTSCRIPT COPY


[Search These Pages] [DAI Home Page] [Comment]