An Efficient Schmidt-EKF for 3D Visual-Inertial SLAM

Abstract

It holds great implications for practical applications to enable centimeter-accuracy positioning for mobile and wearable sensor systems. In this paper, we propose a novel, high-precision, efficient visual-inertial (VI)-SLAM algorithm, termed Schmidt-EKF VI-SLAM (SEVIS), which optimally fuses IMU measurements and monocular images in a tightly-coupled manner to provide 3D motion tracking with bounded error. In particular, we adapt the Schmidt Kalman filter formulation to selectively include informative features in the state vector while treating them as nuisance parameters (or Schmidt states) once they become matured. This change in modeling allows for significant computational savings by no longer needing to constantly update the Schmidt states (or their covariance), while still allowing the EKF to correctly account for their cross-correlations with the active states. As a result, we achieve linear computational complexity in terms of map size, instead of quadratic as in the standard SLAM systems. In order to fully exploit the map information to bound navigation drifts, we advocate efficient keyframe-aided 2D-to-2D feature matching to find reliable correspondences between current 2D visual measurements and 3D map features. The proposed SEVIS is extensively validated in both simulations and experiments.

Distributed Visual-Inertial Cooperative Localization

“…Significant research efforts have recently been devoted to visual-inertial navigation system (VINS) [2], while primarily focusing on improving single-robot VINS accuracy, efficiency, and robustness [3]- [5]. The extension to the multirobot case is not sufficiently explored as a naive approach would be prohibitively costly and non-realtime.…”

“…2) Keyframe-aided 2D-to-2D Matching: To find the feature correspondences between robots, as in [5], [29], we leverage DBoW2 [30], and each robot has a database for all other robots. When a robot receives feature tracks and descriptors from other robots, these are then directly appended…”

VGM-Bench: FPU Benchmark Suite for Computer Vision, Computer Graphics and Machine Learning Applications

“…Kalman Filtering -It is a widely used estimation tool that aims at predicting a series of the state of a process taking as input a series of noisy measurements and the initial state. In Computer vision and Robotics, it plays an important role, especially when dealing with Simultaneous Localization and Mapping (SLAM) [12] and Parallel Tracking and Mapping (PTAM) [16] problems. To this extent, the benchmark mimics the localization case of study in which the state consists of the 3D position and the 3D velocity of a robot.…”

Schmidt or Compressed filtering for Visual-Inertial SLAM?

“…However, this system only yields the accurate pose estimation within the prior map. [Geneva et al, 2019b] implemented the idea from Schmidt Kalman filter [Schmidt, 1966] which selectively including the feature information into the state vector and considers them as a nuisance parameter. There is no updating process for the keyframe state, but cross-correlation with the active state is still maintained.…”