Accurate Initial State Estimation in a Monocular Visual–Inertial SLAM System

Abstract

The fusion of monocular visual and inertial cues has become popular in robotics, unmanned vehicles and augmented reality fields. Recent results have shown that optimization-based fusion strategies outperform filtering strategies. Robust state estimation is the core capability for optimization-based visual–inertial Simultaneous Localization and Mapping (SLAM) systems. As a result of the nonlinearity of visual–inertial systems, the performance heavily relies on the accuracy of initial values (visual scale, gravity, velocity and Inertial Measurement Unit (IMU) biases). Therefore, this paper aims to propose a more accurate initial state estimation method. On the basis of the known gravity magnitude, we propose an approach to refine the estimated gravity vector by optimizing the two-dimensional (2D) error state on its tangent space, then estimate the accelerometer bias separately, which is difficult to be distinguished under small rotation. Additionally, we propose an automatic termination criterion to determine when the initialization is successful. Once the initial state estimation converges, the initial estimated values are used to launch the nonlinear tightly coupled visual–inertial SLAM system. We have tested our approaches with the public EuRoC dataset. Experimental results show that the proposed methods can achieve good initial state estimation, the gravity refinement approach is able to efficiently speed up the convergence process of the estimated gravity vector, and the termination criterion performs well.

Online IMU Self-Calibration for Visual-Inertial Systems

“…We performed dataset experiments to evaluate the proposed online IMU self-calibration method. At present there are several publicly available datasets for VINS evaluation, such as UMich NCLT [45], EuRoc [46], PennCOSYVIO [47], Zurich Urban MAV [48], TUM VI [49], etc., in which, EuRoc is a very popular dataset and used by many papers [8,20,38,39,50,51,52]. For a completely comparison of these datasets, please refer to [49].…”

Simultaneous Localization and Mapping (SLAM) for Autonomous Driving: Concept and Analysis

“…This method is still for the pose graph. Other studies seek to obtain a good initial guess by introducing inertial measurements to support initialization [221,222] or conducting parameter calibration [223][224][225].…”

ICP Algorithm: Theory, Practice and Its SLAM-oriented Taxonomy

“…For SLAM, the state estimation task [37] is equivalent to estimating where the robot is in the environment, given all information about where it was, and what command it was given [38,39]. The mathematical illustration of state estimation is shown by Formula 2, where 𝜃 represents the state estimation task, 𝑥 𝑡 stands for the state or position of the robot at time 𝑡, 𝑧 1:𝑡 means all the state information before time 𝑡, and 𝑢 1:𝑡 represents the sequence of commands we give to the robot before time 𝑡. 𝜃 = 𝑝(𝑥 𝑡 |𝑧 1:𝑡 , 𝑢 1:𝑡 )…”