Predicción de Situaciones no Deseadas Basada en Representaciones Multimodales (Prediction of Undesired Situations based on Multi-Modal Representations)

Bruno Lara (bruno.lara@uaem.mx)1, Juan Manuel Rendón-Mancha (rendon@uaem.mx)1, Marcos A. Capistran (marcos@cimat.mx)2


1Universidad Autónoma del Estado de Morelos
2Centro de Investigación en Matemáticas

This paper appears in: Revista IEEE América Latina

Publication Date: May 2007
Volume: 5,   Issue: 2 
ISSN: 1548-0992


Abstract:
Using forward models as a basic cognitive tool, the cornerstone of the research presented in this paper is the importance of prediction and action as part of the perceptual process of a cognitive system. An artificial agent equipped with a forward model is let to interact with its environment in order to learn the prediction of undesired situations. The forward models are implemented as an artificial neural network trained with data coming form a simulated agent. The network is tested and then implemented on-line on the simulated agent to solve an obstacle avoiding task while seeking a light source. The trained system learns to successfully predict a multimodal sensory representation formed by visual and tactile stimuli. The results presented here are very encouraging and represent the starting point for more research on the use and advantages that cognitive models can provide on artificial autonomous agents.

Index Terms:
Direct Models, Action, Perception, Neural Networks   


Documents that cite this document
This function is not implemented yet.


[PDF Full-Text (1897)]