Skip to Main content Skip to Navigation
Journal articles

Coopération de systèmes d’apprentissage par renforcement multiples.

Benoît Girard 1, 2 Mehdi Khamassi 1, 2
2 AMAC
ISIR - Institut des Systèmes Intelligents et de Robotique
Abstract : Initially developed in the field of artificial intelligence, reinforcement learning methods are an essential component of adaptive robotic control architectures. Two main classes of algorithms have been proposed: with and without internal models of the world. The first one has heavy computational costs but is very adaptive, while the second one is cheap but slow to converge. The combination of these algorithms within a single robotic architecture could possibly benefit from the advantages of each one. We present here these two families of algorithms, as well as the combination methods that have been proposed and tested in the neuroscience and robotics field.
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-01524743
Contributor : Benoît Girard <>
Submitted on : Thursday, May 18, 2017 - 5:28:52 PM
Last modification on : Friday, March 19, 2021 - 3:32:54 AM

Identifiers

  • HAL Id : hal-01524743, version 1

Citation

Benoît Girard, Mehdi Khamassi. Coopération de systèmes d’apprentissage par renforcement multiples.. Techniques de l'Ingenieur, Techniques de l'ingénieur, 2016, pp.S7793. ⟨hal-01524743⟩

Share

Metrics

Record views

204