Skip to Main content Skip to Navigation
Habilitation à diriger des recherches

Techniques d'automatique et de traitement du signal pour l'asservissement visuel et la perception auditive en robotique

Patrick Danès 1
1 LAAS-RAP - Équipe Robotique, Action et Perception
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : The work developed in this manuscript essentially deals with a set of functional primitives in Robotics, whose scientific foundations are anchored in Automatic Control and Signal Processing. Far more than an application domain of theories developed elsewhere, Robotics, through its specificities or its constraints, often questions the state of the art in Control or Signal Processing, or even requires nontrivial methodological extensions. The richness of the dialog between these three scientific areas guided our research. A first part of our work has been geared towards the development of a generic framework to the ''multicriteria'' analysis and synthesis of visual-based control schemes, i.e.\ which take into account constraints (visibility, actuators' saturations, 3D constraints during the motion, etc.). These problems have been turned into the stability analysis / stabilization of uncertain nonlinear rational systems subject to rational constraints. The underlying theoretical issues are Lyapunov stability and LMI optimization. By duality, visual-based localization has been tackled by robust set-membership filtering techniques for rational systems. A second contribution takes place within the relatively new area of Robot Audition. We have proposed low-level auditory functions for detection, localization and extraction of sound sources. An original integrated auditory sensor developed at LAAS-CNRS has enabled their implementation. The underlying theoretical issues are array processing (beamforming, high resolution methods) and LMI optimization. More recently, the foundations of a new approach to voice activity detection have been seet, on the basis of stochastic matched filtering. In parallel to these targeted developments, a groundwork in stochastic filtering and change detection got materialized into scientific collaborations within targeted application domains: sequential Monte Carlo and Quasi Monte Carlo methods for visual tracking of people / gestures and for visual human motion capture; change detection and IMM filtering with heterogeneous state space models for dynamic scene monitoring and ARGOS localization. Some of the above themes ''merge'' into some structuring axes of our future developments. A particular effort will concern active perception (viz.\ which exploits proprioception and motion) declined in the context of binaural audition, as well as the fusion of embedded and deported auditory functions in intelligent environments.
Document type :
Habilitation à diriger des recherches
Complete list of metadata

Cited literature [146 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00604522
Contributor : Arlette Evrard <>
Submitted on : Wednesday, June 29, 2011 - 10:58:10 AM
Last modification on : Thursday, June 10, 2021 - 3:06:57 AM
Long-term archiving on: : Sunday, December 4, 2016 - 8:52:21 AM

Identifiers

  • HAL Id : tel-00604522, version 1

Citation

Patrick Danès. Techniques d'automatique et de traitement du signal pour l'asservissement visuel et la perception auditive en robotique. Automatique / Robotique. Université Paul Sabatier - Toulouse III, 2010. ⟨tel-00604522⟩

Share

Metrics

Record views

1007

Files downloads

12740