Skip to Main content Skip to Navigation
Theses

Méthodes de partage d'informations visuelles et inertielles pour la localisation et la cartographie simultanées décentralisées multi-robots

Abstract : In mobile robotics, simultaneous mapping and localization (SLAM) methods are an essential algorithmic brick in order to perceive the environment and autonomously navigate within it. In a visual context, single-robot SLAM methods have now reached high maturity, which has allowed the development of collaborative SLAM methods. However, enhancing the agents’ autonomy while facing informational, network and resource constraints raises the question about the nature of the data to be transmitted between the robots. The objective of this thesis is to design methods for sharing visual and inertial information for collaborative SLAM, which enforce the autonomy of agents and allow them to refine their navigation when they visit common areas. To this end, we have investigated multiple exchange paradigms for decentralized visual-inertial and stereo-visual SLAM architectures. First, we proposed three ways of summarizing visual-inertial data, from which ones we built collaborative SLAM methods. Those are respectively based on the exchange of rigid visual-inertial submaps, condensed packets computed through marginalization and consistent sparsification techniques, and pruned packets built through the selection of raw visual-inertial factors. We evaluated those methods on standard datasets, as well as on our custom AirMuseum dataset, which we purposively designed. Finally, from a mapping perspective, we applied the developped architecture to multi-robot dense SLAM, by extending a dense collaborative mapping method based on the exchange, the registration and the fusion of locally consistent submaps.
Document type :
Theses
Complete list of metadata

https://tel.archives-ouvertes.fr/tel-03273943
Contributor : Abes Star :  Contact Connect in order to contact the contributor
Submitted on : Tuesday, June 29, 2021 - 3:54:10 PM
Last modification on : Wednesday, October 13, 2021 - 3:52:05 PM
Long-term archiving on: : Thursday, September 30, 2021 - 7:10:13 PM

File

R_DUBOIS.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03273943, version 1
`

Citation

Rodolphe Dubois. Méthodes de partage d'informations visuelles et inertielles pour la localisation et la cartographie simultanées décentralisées multi-robots. Automatique / Robotique. École centrale de Nantes, 2021. Français. ⟨NNT : 2021ECDN0001⟩. ⟨tel-03273943⟩

Share

Metrics

Record views

137

Files downloads

142