.png){kind=icon}
.png){kind=icon}
Return
Lane-Level Positioning Method Based on Vector Map Matching for Autonomous Driving: A Polynomial Fitting Method
.png){kind=icon}
PRE
AI
DOI：10.1109/JSEN.2024.3382343.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In this article, a high-accuracy real-time vector map matching method based on polynomial fitting is proposed for autonomous driving. The real-time kinematic (RTK) positioning accuracy decreases due to the influence of multipath effects. Simultaneously, prolonged usage of the inertial navigation system results in a degradation of positioning accuracy, failing to meet the requirements of a level 4 (L4) autonomous driving system. Consequently, the most effective method for carrier position is to integrate vector map matching created from high-precision positioning points, despite the fact that this method faces two distinct difficulties. First, a sparse description of vector maps in straight segment points causes reduced matching accuracy. Second, curve segment points are computationally intensive, and these data make it difficult to directly match vector maps. This article uses a polynomial fitting map compression method that provides a new description of vector maps without increasing internal storage and computation to address the vector map matching problem. Using this method, the storage space of the map is compressed, and the positioning ability at the curve segment is enhanced. The results show that the positioning accuracy of missing fixed solutions is less than 20 cm, meeting the requirements of L4 autonomous driving vehicles without expensive light detection and ranging (LIDAR) equipment.
Keywords:
Vectors
Roads
Autonomous vehicles
Satellite navigation systems
Point cloud compression
Laser radar
Fitting
Autonomous vehicle navigation
integrated navigation
lane-level positioning
vector map