High-precision, consistent EKF-based visual-inertial odometry

By Mingyang Li, Anastasios I. Mourikis. Int. J. Robotics Res., 2013

840 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/0be0c13803cd08e81b7adaada537e91222eb1491

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

In Electrical and Electronic Engineering

Citations in the corpus

A General Optimization-based Framework for Global Pose Estimation with Multiple Sensors (2019, methodology): “Among these work, [6, 7, 9] are filter-based methods while [8, 10, 11] are optimization-based methods.”
Using Inertial Sensors for Position and Orientation Estimation (2017, neutral): “For pose estimation, inertial sensors are often combined with measurements from for instance a global navigation satellite system (GNSS) [71, 147, 59], an ultrawideband (UWB) system [74, 132, 110, 27, 30] or cameras [26, 61, 79, 94].”
Large-scale direct SLAM for omnidirectional cameras (2015, neutral): “This can be done in a filtering framework [3], [4], [5], or in a keyframe-based non-linear optimization framework [6], [7], [8].”
Robust Stereo Visual Inertial Odometry for Fast Autonomous Flight (2017, background): “As has been shown in [2], [23], the EKF-based VIO for 6-DOF motion estimation has four unobservable directions corresponding to global position and rotation along the gravity axis, i.”
LSD-SLAM: Large-Scale Direct Monocular SLAM (2014, background): “The fundamental idea behind feature-based approaches (both filtering-based [15,19] and keyframe-based [15]) is to split the overall problem – estimating geometric information from images – into two sequential steps: First, a set of feature observations is extracted from the image.”
Visual-Inertial Navigation: A Concise Review (2019, methodology): “, unobservable subspace) that ideally spans four directions corresponding to the global translation and global rotation about the gravity vector [19, 38, 40, 51].”
A Commute in Data: The comma2k19 Dataset (2018, methodology): “They were computed with a tightly coupled GNSS/INS/Vision optimizer, where raw GNSS measurements and ORB [17] features were fed into a Multi-State Constraint Kalman Filter (MSCKF) [18], [19].”
Aerodynamics and Control of Quadrotors (2017, methodology): “Currently, the majority of sensor fusion algorithms such as SLAM use extended Kalman Filtering (EKF) Li and Mourikis [2013] which require higher computational cost than a non-linear complementary filter approach which can be more suited for computationally constrained small aerial vehicles.”
VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator (2017, background): “Direct approaches [2], [3], [22] minimize photometric error, while indirect approaches [12], [15], [17] minimize the geometric displacement.”
A robust and modular multi-sensor fusion approach applied to MAV navigation (2013, background): “While showing unmatched accuracy and consistency [5], [9] they are commonly designed for a particular sensor setup with limited modularity.”
Autonomous aerial navigation using monocular visual‐inertial fusion (2018, neutral): “= 1 kλmin H [ u 1 ] + J (19) = 1 kλmin ⎢⎢⎢⎣ −h1− −h2− −h3− ⎥⎥⎥⎦ [ u 1 ] + ⎢⎢⎢⎣ J1 J2 J3 ⎥⎥⎥⎦ (20)”
Robust visual inertial odometry using a direct EKF-based approach (2015, background): “As known from previous work the inclination angles and the robocentric velocities of visual-inertial setups are fully observable [17], and we can nicely observe the initial decrease of the corresponding covariance (especially when the system gets excited).”
ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual–Inertial, and Multimap SLAM (2020, background): “The initial system was perfected in [34] and extended to stereo in [35], [36].”
Adaptive VIO: Deep Visual-Inertial Odometry with Online Continual Learning (2024, methodology): “A classic VIO system is mainly composed of visual as sociation, IMU preintegration, and back-end nonlinear opti mization [16, 24, 30, 37, 42] (or filtering [16, 17]), as shown in Fig. l (a).”
Advanced closed loop visual navigation for micro aerial vehicles (2014, background): “(Kelly and Sukhatme, 2011)) involve including the 3D positions of landmarks in the state vector, making real-time processing infeasible soon due to the growing state vector, if these landmarks are not marginalized out and / or sliding window techniques are used (Li and Mourikis, 2013).”
A General Optimization-based Framework for Local Odometry Estimation with Multiple Sensors (2019, methodology): “MSCKF [6, 7] was a popular EKF-based VIO (Visual Inertial Odometry), which maintained several camera poses and leveraged multiple camera views to form the multi-constraint update.”
IONet: Learning to Cure the Curse of Drift in Inertial Odometry (2018, background): “Current approaches to track wheeled objects are mainly based on visual odometry or visual-inertial odometry (VIO) (Li and Mourikis 2013; Bloesch et al. 2015).”
GVINS: Tightly Coupled GNSS–Visual–Inertial Fusion for Smooth and Consistent State Estimation (2021, background): “Based on MSCKF, [3] makes improvements on its accuracy and consistency, and [4] aims to overcome its numerical stability issue especially on mobile devices.”
Iterated extended Kalman filter based visual-inertial odometry using direct photometric feedback (2017, methodology): “While the above handling of rotations has been used similarly in previous ﬁltering frameworks (e.g. Li and Mourikis (2013); Bloesch et al. (2013)), we extend the methodology to 3D unit vectors on the 2-sphere S 2 .”