Tightly Coupled Integration of MVIO and UC-PPP based on factor graph optimization for complex urban environments

Abstract

Abstract The emergence of low-cost micro-electro-mechanical system inertial measurement units, cameras and Global Navigation Satellite System (GNSS) receivers has promoted the research of multi-sensor fusion for positioning. With the rapid development of chip technology, it has become a trend for low-cost GNSS chips to provide multi-frequency carrier phase observations while supporting multiple constellations. To enhance the positioning performance of multi-frequency and multi-system precise point positioning (PPP) in the complex urban environment, this paper proposes a tightly coupled system of monocular visual-inertial odometry (MVIO) and uncombined PPP (UC-PPP) based on factor graph optimization. The initialization of the integrated system includes two parts: one is for monocular vision and inertial navigation systems, and another is for MVIO and UC-PPP. The initialization of MVIO and UC-PPP adopts a coarse-to-fine approach to correct online the transformation of local and global frames. In addition, the sliding window and marginalization methods are adopted to retain the constraints between adjacent observations and eliminate useless observations in the window. The pedestrian and vehicle tests in the complex urban environment verify the performance of the proposed method. Compared with open-source software GVINS, the positioning accuracy of the proposed method has been further improved due to the use of carrier phase observations with higher measurement accuracy. Compared with PPP alone, the improvement of the proposed method for the low-speed and short-distance pedestrian test in the east, north, and up directions is 73.3, 54.8 and 62.7%, respectively, while the improvement for the high-speed and long-distance vehicle test is 63.0, 59.3 and 70.5%, respectively. Experiment results show that the proposed method has better positioning accuracy and performance in complex urban environments.