Previous Abstract Return to Session C5 Next Abstract

Session C5: Multisensor Integrated Systems and Sensor Fusion Technologies

A Vision-assisted GNSS-RTK Ambiguity Fixing Method Considering Visual Lane Line Ambiguity
Chuang Qian, Hongjuan Zhang, Wenzhuo Li, Hui Liu, Bijun Li, Wuhan University, China
Location: Atrium Ballroom
Alternate Number 5

High-precision navigation and positioning is one of the core technologies of emerging technologies such as automatic driving and cooperative vehicle infrastructure system. GNSS RTK is one of the most widely used high-precision positioning technologies in outdoor applications, and can achieve dynamic positioning accuracy of centimeter-level. However, in practices, GNSS RTK performance is easily affected by factors such as observation noise and environmental occlusion, which makes it difficult to fix the ambiguity correctly to achieve continuous and reliable centimeter-level high-precision positioning. Integration of GNSS and other sensors has been widely studied as the main method to solve this problem, especially visual sensors, which are widely used due to their low price. As a main visual navigation method, the map-relative localization method through visual lane line detection is commonly used to constrain or estimate the position of a vehicle. However, it is difficult to determine which lane the vehicle locates in and the position of the vehicle along the lane line. In many cases, lane line is not continuous but dash line, and we divide the lane into continuous rectangles with same sizes. Hence, positioning a vehicle along the lane line becomes finding the correct rectangle which the vehicle is in. And we call it visual lane line ambiguity. In this paper, visual lane line detection and a prior lane-level map are used to assist the ambiguity fixing of GNSS RTK, and the visual lane line ambiguity is tightly coupled with RTK. The experimental results show that in the open environment, the single-epoch ambiguity fixed rate of the lane line-assisted GNSS RTK achieved 97%, while that of the tradition GNSS RTK method was 89%. In the simulated high occlusion environment, the single-epoch ambiguity fixed rate of the lane line-assisted GNSS RTK was 75%, while fixed rate of the case without considering the lane line assistance was only 25%. And all fixed success epochs can correctly estimate the ambiguity of the visual lane line.



Previous Abstract Return to Session C5 Next Abstract