Skip to Main content Skip to Navigation
New interface

Localisation collaborative visuelle-inertielle de robots hétérogènes communicants

Abstract : Localization is of crucial importance for robots navigation. This importance has allowed the emergence of several precise localization techniques. Our contribution consists of proposing a transition from an individual inertial visual localization technique to the multi-robots collaborative localization case. This work aims to achieve a collaborative localization as fast, robust and accurate as the individual starting technique. We adopt a tightly coupled MSCKF (Multi State Constraint Kalman Filter) approach to achieve the data fusion. The characteristics of this data fusion are first studied in the individual case to test the robustness and the precision under different conditions and with different observation models. The results of this study directed us towards the best structure adapted to an augmentation to the collaborative localization case. The proposed collaborative algorithm is a hierarchical process of three stages. A collaborative localization is initialized based on the relative distance measurements using Ultra-Wide Band (ULB) sensors. Then, a collaborative localization based on images overlapping using a suitable measurement model, and a data fusion structure that absorbs the computation time excess caused by the collaboration is achieved. Finally, to increase precision, an extraction of the environment constraints, followed by an integration using a truncation in the filter are proposed.
Document type :
Complete list of metadata

Cited literature [97 references]  Display  Hide  Download
Contributor : ABES STAR :  Contact
Submitted on : Thursday, October 3, 2019 - 5:10:07 PM
Last modification on : Thursday, September 8, 2022 - 3:52:26 AM


Version validated by the jury (STAR)


  • HAL Id : tel-02305101, version 1



Bilel Chenchana. Localisation collaborative visuelle-inertielle de robots hétérogènes communicants. Automatique / Robotique. Université de Limoges, 2019. Français. ⟨NNT : 2019LIMO0018⟩. ⟨tel-02305101⟩



Record views


Files downloads