ME4VIO: Manifold Encapsulation for Visual Inertial Odometry with Application to Vehicle Navigation in Urban Environments

Shaza Kaoud Abdelaziz, Sidney Givigi, Mohamed Elhabiby, Aboelmagd Noureldin

Abstract:	Visual-inertial odometry (VIO) is a technique that uses data from both cameras and inertial sensors to estimate the pose of a moving vehicle. VIO is a promising technology for autonomous navigation, as it can provide accurate and reliable pose estimates even in challenging environments, such as those with limited or no GPS coverage or in the case of modern self-driving cars with HD-maps capabilities within a map outage or update. This paper proposes a novel VIO algorithm that uses a manifold encapsulation approach to represent and mitigate uncertainty. The proposed algorithm, Manifold Encapsulation for Visual Inertial Odometry (ME4VIO), is based on the ?-manifold, a mathematical structure that can represent the uncertainty associated with pose estimates. The ?-manifold approach has several advantages over traditional VIO methods, including (i) It can more accurately represent the uncertainty associated with pose estimates. (ii) It can be more robust to noise and outliers. (iii) It can be more efficient to compute. The proposed algorithm was evaluated on a dataset collected from an actual road test in Kingston, Ontario. The results showed that ME4VIO achieved more accurate and reliable pose estimates than traditional VIO methods. In particular, ME4VIO could maintain accurate pose estimates even in challenging conditions, such as fast motion, low-textured environments, and sensor drifts. The results of this paper demonstrate that the proposed ME4VIO algorithm is a promising approach for VIO, even in challenging conditions. This qualifies ME4VIO as a potential solution for various autonomous navigation applications.
Published in:	Proceedings of the 36th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2023) September 11 - 15, 2023 Hyatt Regency Denver Denver, Colorado
Pages:	2028 - 2032
Cite this article:	Abdelaziz, Shaza Kaoud, Givigi, Sidney, Elhabiby, Mohamed, Noureldin, Aboelmagd, "ME4VIO: Manifold Encapsulation for Visual Inertial Odometry with Application to Vehicle Navigation in Urban Environments," Proceedings of the 36th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2023), Denver, Colorado, September 2023, pp. 2028-2032. https://doi.org/10.33012/2023.19388
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In