Incremental learning of Bayesian sensorimotor models: from low-level behaviours to large-scale structure of the environment

Abstract : This paper concerns the incremental learning of hierarchies of representations of space in artificial or natural cognitive systems. We propose a mathematical formalism for defining space representations (Bayesian Maps) and modelling their interaction in hierarchies of representations (sensorimotor interaction operator). We illustrate our formalism with a robotic experiment. Starting from a model based on the proximity to obstacles, we learn a new one related to the direction of the light source. It provides new behaviours, like phototaxis and photophobia. We then combine these two maps so as to identify parts of the environment where the way the two modalities interact is recognisable. This classification is a basis for learning a higher level of abstraction map that describes the large-scale structure of the environment. In the final model, the perception–action cycle is modelled by a hierarchy of sensorimotor models of increasing time and space scales, which provide navigation strategies of increasing complexities.
Liste complète des métadonnées

Cited literature [32 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-00537809
Contributor : Julien Diard <>
Submitted on : Friday, November 19, 2010 - 1:57:07 PM
Last modification on : Thursday, February 7, 2019 - 4:31:24 PM
Document(s) archivé(s) le : Friday, December 2, 2016 - 11:08:44 PM

File

diard10_author.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00537809, version 1

Collections

Citation

Julien Diard, Estelle Gilet, Eva Simonin, Pierre Bessiere. Incremental learning of Bayesian sensorimotor models: from low-level behaviours to large-scale structure of the environment. Connection Science, Taylor & Francis, 2010, 22 (4), pp.291-312. 〈hal-00537809〉

Share

Metrics

Record views

639

Files downloads

267