Modeling Habituation in the Cnidarian Hydra

  • Authors:

  • Malin Aktius;Mats Nordahl;Tom Ziemke

  • Affiliations:

  • University of Skövde, School of Humanities and Informatics, SE-541 28 Skövde, Sweden;Department of Applied Information Technology, Göteborg University and Chalmers University of Technology, SE-417 56 Göteborg, Sweden;University of Skövde, School of Humanities and Informatics, SE-541 28 Skövde, Sweden

  • Venue:
  • Proceedings of the 2008 conference on Tenth Scandinavian Conference on Artificial Intelligence: SCAI 2008
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

In the design of behavior-based control architectures for robots it is common to use biology as inspiration, and often the observed functionalities of insect behaviors are used as templates. While several robot behaviors have been successfully implemented using this approach, relatively little has been done when it comes to building models of animal behavior using quantitative empirical data. The work reported here uses a system identification approach to model constituent behaviors of a simple biological organism, the hydra. This paper reports on the evolutionary optimization of a behavior module that is based on the hydra's response to mechanical stimuli, which shows habituation. Two model representation schemes were investigated: a recurrent neural network, and a model based on cascaded leaky integrators. Both models were structurally and parametrically optimized by means of an evolutionary algorithm, and it was found that the leaky integrator model performed better on unseen data.