Surveillance temps-réel des systèmes Homme-Machine. Application à l'assistance à la conduite automobile

Miguel Gonzalez-Mendoza 1
1 LAAS-DISCO - Équipe DIagnostic, Supervision et COnduite
LAAS - Laboratoire d'analyse et d'architecture des systèmes [Toulouse]
Abstract : This work is developed in the framework of human-machine systems monitoring, in which the human operator is a decision element in the control loop. This type of systems requires a global automated real-time monitoring, including the detection of technical hitches and human faults. Assuming the technological part works "correctly" and/or that it is supervised by a diagnosis system, we are centered on the human operator part, the most critical to supervise. We present a diagnosis system and its application to driver's monitoring. The system is supported by information provided by onboard sensors. This work has been developed in the framework of the European project AWAKE and the French national project "factors of vigilance degradation and safety in transports" supported by the PREDIT program.
This work proposes a real-time general strategy for monitoring the vigilance level of the driver (slow dynamics) and the monitoring of the risk level related to the current situation (instantaneous dynamics), through:
* Temporal and frequential analysis of vehicle's mechanical signals (performance measures) processed by wavelets and filters, to extract dynamic, statistical and frequential characteristics of driver's performance degradation,
* Characteristics learning by Support Vector Machines for which we developed adapted implementation strategies for an on-line learning and large-scale problems,
* FIS fusion, in order to exploit the human expertise and to produce a diagnosis of the level of risk related to lane departure.
* Validation of such an approach through EEG and EOG (physiological measurements) and self evaluations (subjective measurements).
We apply this methodology to various experiments under the framework of AWAKE and PREDIT carried out in simulators or demonstrators.
Document type :
Theses
Liste complète des métadonnées

Cited literature [29 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00336732
Contributor : Miguel Gonzalez-Mendoza <>
Submitted on : Wednesday, November 5, 2008 - 2:37:10 AM
Last modification on : Friday, April 12, 2019 - 4:23:50 PM
Document(s) archivé(s) le : Monday, June 7, 2010 - 8:24:54 PM

Identifiers

  • HAL Id : tel-00336732, version 1

Citation

Miguel Gonzalez-Mendoza. Surveillance temps-réel des systèmes Homme-Machine. Application à l'assistance à la conduite automobile. Informatique [cs]. INSA de Toulouse, 2004. Français. ⟨tel-00336732⟩

Share

Metrics

Record views

752

Files downloads

3491