Improved Correlation Scanning Matching Based on Grid Map
Yan Zhou, Jun Zhang, Jiazhi Yu, Zhuo Li, Yue Geng, Xiaole Bian · 2024
In recent years, the continuous in-depth research on Simultaneous Localization And Mapping (SLAM) has promoted the rapid development of autonomous driving, but also brought challenges to the research of SLAM technology. Compared to LiDAR, millimeter wave radar has significant advantages in all-weather and privacy avoidance, and has reliable stability in complex climate scenarios. However, the measurement accuracy and limited data volume of millimeter wave radar pose significant challenges in SLAM applications. In order to solve the problems of large positioning errors and incomplete mapping based on two-dimensional millimeter wave radar SLAM, a method of positioning and mapping based on grid map improved correlation scanning matching is proposed. In this method, we use the bilinear interpolation method to calculate the grid score for correlation scanning matching to obtain the optimal pose. Before obtaining the optimal pose, we use a multi-resolution map to use the coarse matching result as the initial value for fine matching to improve the positioning accuracy. Using frame and submap matching in correlation scanning to solve the problem of small and sparse data in millimeter wave radar. By comparing experimental results, it was found that the improved correlation scanning matching based on grid maps outperforms the original localization and mapping results in various scenarios.