OATAO - Open Archive Toulouse Archive Ouverte Open Access Week

Development of GNSS/INS/SLAM Algorithms for Navigation in Constrained Environments

Ben afia, Amani. Development of GNSS/INS/SLAM Algorithms for Navigation in Constrained Environments. PhD, Signal, Image, Acoustique et Optimisation, Institut National Polytechnique de Toulouse, 2017

[img]
Preview
(Document in English)

PDF (Author's version) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
7MB

Abstract

For land vehicles, the requirements of the navigation solution in terms of accuracy, integrity, continuity and availability are more and more stringent, especially with the development of autonomous vehicles. This type of application requires a navigation system not only capable of providing an accurate and reliable position, velocity and attitude solution continuously but also having a reasonable cost. In the last decades, GNSS has been the most widely used navigation system especially with the receivers decreasing cost over the years. However, despite of its capability to provide absolute navigation information with long time accuracy, this system suffers from problems related to signal propagation especially in urban environments where buildings, trees and other structures hinder the reception of GNSS signals and degrade their quality. This can result in significant positioning error exceeding in some cases a kilometer. Many techniques are proposed in the literature to mitigate these problems and improve the GNSS accuracy. Unfortunately, all these techniques have limitations. A possible way to overcome these problems is to fuse “good” GNSS measurements with other sensors having complementary characteristics. In fact, by exploiting the complementarity of sensors, hybridization algorithms can improve the navigation solution compared to solutions provided by each stand-alone sensor. Generally, the most widely implemented hybridization algorithms for land vehicles fuse GNSS measurements with inertial and/or odometric data. Thereby, these Dead-Reckoning (DR) sensors ensure the system continuity when GNSS information is unavailable and improve the system performance when GNSS signals are degraded, and, in return the GNSS limits the drift of the DR solution if it is available. However the performance achieved by this hybridization depends thoroughly on the quality of the DR sensor used especially when GNSS signals are degraded or unavailable. Therefore, this Ph.D. thesis, which is part of a common French research project involving two laboratories and three companies, aims at extending the classical hybridization architecture by including other sensors capable of improving the navigation performances while having a low cost and being easily embeddable. For this reason, the use of vision-based navigation techniques to provide additional information is proposed in this thesis. In fact, cameras have become an attractive positioning sensor recently with the development of Visual Odometry and Simultaneous Localization and Mapping (SLAM) techniques, capable of providing accurate navigation solution while having reasonable cost. In addition, visual navigation solutions have a good quality in textured environments where GNSS is likely to encounter bad performance. Therefore, this work focuses on developing a multi-sensor fusion architecture integrating visual information with the previously mentioned sensors. In particular, the contribution of this information to improve the vision-free navigation system performance is highlighted. The proposed architecture respects the project constraints consisting of developing a versatile and modular low-cost system capable of providing continuously a good navigation solution, where each sensor may be easily discarded when its information should not be used in the navigation solution

Item Type:PhD Thesis
Uncontrolled Keywords:
Institution:Université de Toulouse > Institut National Polytechnique de Toulouse - Toulouse INP (FRANCE)
Laboratory name:
Research Director:
Macabiau, Christophe and Escher, Anne-Christine
Statistics:download
Deposited On:02 Nov 2021 14:54

Repository Staff Only: item control page