Skip to Main content Skip to Navigation

Modélisation incrémentale et localisation par amers pour la navigation d'un robot mobile autonome en environnement naturel

Abstract : This thesis deals with natural environment modelling and autonomous mobile robot localisation. The environment, unknown or partially known, is incrementally perceived by the robot with a 3D sensor (laser range finder or stereovision). Application of this work are intervention mobile robotics and planetary exploration. A first part specifies the properties that the environment model should possess and provides a critical analysis of the different representations studied in the litterature. The proposed approach is then introduced. It is based on a decomposition of the environment in two entities: the ground and the objects. The second part deals with the modelling of a perception. It explains how the robot segments the ground and the objects from 3D images and illustrates this process with numerous examples using data coming from different sensors and scenes. A geometrical model is computed independently for each object and topological relations (characterizing their layout) are established between them. Specific objects, called landmarks, are selected on the basis of precision and distinction criteria. The robot computes a feature for each one (the top) and its precision according the landmark shape and resolution. The third and last part deals with incremental modelling of the whole environment. After each perception, the robot recognizes the landmarks or a specific configuration of landmarks in the model. Using the matchings, it estimates both its own position and precision and the landmark's. The localisation and map building processes are achieved with an Extended Kalman Filter. Correlations between robot and objects are kept in order to maintain the consistency of the global map. Several experiments of incremental modelling are presented and analysed methodically. Robot and landmark positions as estimated by the system are compared with reference values, and the validity of the localisation and the built map are shown.
Document type :
Complete list of metadata
Contributor : Emilie Marchand <>
Submitted on : Tuesday, April 3, 2007 - 3:25:42 PM
Last modification on : Friday, January 10, 2020 - 9:08:08 PM
Long-term archiving on: : Friday, May 13, 2011 - 9:26:06 PM


  • HAL Id : tel-00139789, version 1


Stéphane Betge-Brezetz. Modélisation incrémentale et localisation par amers pour la navigation d'un robot mobile autonome en environnement naturel. Automatique / Robotique. Université Paul Sabatier - Toulouse III, 1996. Français. ⟨tel-00139789⟩



Record views


Files downloads