Skip to Main content Skip to Navigation
Theses

Le contrôle non linéaire par réseaux de neurones formels: les perceptrons affines par morceaux

Abstract : The purpose of this work is to expose new results about using a specific class of artificial neural networks (the Piecewise Affine Perceptrons: PAP) in a closed loop optimal control design. The main obtained results are: some properties of PAP about the function they emulate,, a constructive representation theorem of Continuous Piecewise Affine Functions by PAPs, which allow to formally build a PAP from a collection of affine functions, some heuristics about perceptrons learning in a closed loop in an optimal control framework, theoretical stability results about PAP controllers. The last part is dedicated to the application of those results to automatic building of engine combustion controllers, which conducted two two international patents by Renault.
Document type :
Theses
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00009592
Contributor : Charles-Albert Lehalle <>
Submitted on : Thursday, July 14, 2005 - 10:10:13 AM
Last modification on : Thursday, December 10, 2020 - 12:30:35 PM
Long-term archiving on: : Friday, April 2, 2010 - 10:40:53 PM

Identifiers

  • HAL Id : tel-00009592, version 1

Citation

Charles-Albert Lehalle. Le contrôle non linéaire par réseaux de neurones formels: les perceptrons affines par morceaux. Mathématiques [math]. Université Pierre et Marie Curie - Paris VI, 2005. Français. ⟨tel-00009592⟩

Share

Metrics

Record views

330

Files downloads

3010