An Optimization-Based Indoor-Outdoor Seamless Positioning Method Integrating GNSS RTK, PS, and VIO

The integration of global navigation satellite system (GNSS), vision, and inertial can perform positioning in indoor-outdoor transition scenarios. However, in indoor scenes, GNSS signals fail, and the global information of visual-inertial odometry (VIO) is not observable. In this brief, we propose an optimization-based seamless indoor-outdoor positioning method. Specifically, in the factor graph framework, we fuse GNSS, vision, inertial, and pressure sensor (PS) measurements to obtain accurate global positioning information in complex indoor-outdoor transition scenarios. In addition, we construct PS factors by using inertial measurement unit (IMU) forward-backward preintegration technology to eliminate the impact of time delay of PS measurements. Moreover, we incorporate GNSS and PS factors into the covisibility graph marginalization strategy of ORB-SLAM3. Furthermore, in the initialization process, we add the alignment of GNSS-inertial and PS-inertial. Real world experiments demonstrate that the proposed algorithm has better positioning accuracy and robustness than the state-of-the-art visual-inertial algorithms in complex indoor-outdoor scenes such as underground garages and multi-story buildings. To contribute to the community, we open-source the dataset on Github: https://github.com/SYSU-CPNTLab/GVI-SYSU-Outdoor-Indoor-Dataset.