Robust visual-inertial SLAM: combination of EKF and optimization method.

This paper has presented a novel tightly-coupled monocular visual-inertial Simultaneous Localization and Mapping(SLAM) algorithm combining the advantages of both filtering methods and optimization methods. Our approach uses IMU assisted visual EKF SLAM method in the front-end to estimate the pose and velocity at frame rate. In the back-end, the standard approaches such as nonlinear optimization as well as loop closure have been used to obtain accurate trajectory of the sensor and 3D map of the environment. We also propose a novel scale computation approach that can accurately initialize the scale in short time. In addition, the optimized map points have been used to improve the accuracy of the filter. We evaluated the algorithm on a public dataset, when compared to other state-of-the-art monocular Visual-Inertial SLAM approaches, our algorithm achieves unprecedented accuracy and robustness.