Learning-aided visual inertial odometry for mobile robots

Abstract

This research presents a novel approach to visual-inertial odometry (VIO) for challenging environments based on VINS-Fusion. The proposed method utilizes a deep learning technique to enhance the performance of the state estimation. The proposed approach employs semantic segmentation to highlight ground features such as lane markings and ground bricks. The exper- iments’ results demonstrate the proposed method’s effectiveness in improving the robustness and accuracy of the VIO system in semi-outdoor environments with dynamic objects. The re- port concludes with a summary of the main findings and recommendations for future research. This research has the potential to enhance the capabilities of autonomous systems in indoor environments, such as in factories, hospitals, and shopping centers.