Skip to Main content Skip to Navigation

Modélisation et commande de processus par réseaux de neurones ; application au pilotage d'un véhicule autonome

Abstract : Artificial neural networks allow the construction of a wide family of nonlinear models and controllers using statistical learning. The purpose of this thesis is to establish how to implement neural networks for the modeling and control of nonlinear dynamic processes, and to evaluate their contribution for these tasks. At the theoretical level, we present the neural modeling and control in a general framework, from the viewpoint of classic automatic control. Concerning modeling, the results about linear systems help us to derive the theoretical nonlinear optimal predictors associated to various hypotheses made about the type of noise that affects the process; a training methodology is proposed that provides the optimal neural predictors implementing the theoretical predictors. Then, we propose a family of neural control systems, the properties and the links to classic linear and nonlinear control systems of which are investigated; in particular, we concentrate on robustness issues, which leads us to develop neural internal model control. At the practical level, we illustrate our approach and our results with an industrial application, the real-time control of an autonomous four-wheel drive vehicle, the steering wheel, throttle and breaks of which are controlled by neural networks.
Complete list of metadatas

Cited literature [86 references]  Display  Hide  Download
Contributor : Isabelle Rivals <>
Submitted on : Tuesday, March 5, 2013 - 3:52:20 PM
Last modification on : Wednesday, December 9, 2020 - 3:12:37 PM
Long-term archiving on: : Sunday, April 2, 2017 - 9:26:19 AM


  • HAL Id : pastel-00797072, version 1


Isabelle Rivals. Modélisation et commande de processus par réseaux de neurones ; application au pilotage d'un véhicule autonome. Automatique / Robotique. Université Pierre et Marie Curie - Paris VI, 1995. Français. ⟨pastel-00797072⟩



Record views


Files downloads