Implementation and evaluation of simple artificial creatures using neural networks for perception interpretation

Abstract

This paper contains both the description and the evaluation of a simple artificial creature system. The implemented model’s perception system uses a color representation of smells which are recognized by a neural network by mapping RGB colors in drives variation vectors. The creatures behaviors are fully reactive and based on achievement of homeostasis. The drives the system seeks to balance are energy (eating), grouping(being close to other of same species), spacing (being sufficiently apart from other creatures) and fear (being far from other creatures of different species). The resulting behavior in the 3-dimensional world suggest the approach is reasonable due the seemingly natural tendency of grouping and seeking for food without the superposition of individuals with a simple criteria for action selection based on the search for homeostasis.

3 Figures and Tables

Cite this paper

@inproceedings{Batista2011ImplementationAE, title={Implementation and evaluation of simple artificial creatures using neural networks for perception interpretation}, author={Marcel Fernando Miranda Batista}, year={2011} }