Multi-Kernel Maximum Correntropy Error State Kalman Filter For LiDAR-Inertial State Estimator
作者
Tingting Chang,Liang Zhao,Qingxi Zeng
出处
期刊:Proceedings of the 2019 International Conference on Robotics, Intelligent Control and Artificial Intelligence日期:2023-12-01卷期号:: 341-346被引量:1
标识
DOI:10.1109/ricai60863.2023.10489657
摘要
The LiDAR/Inertial integrated navigation system based on the Kalman filter has gained extensive adoption in the realm of robotics and unmanned environments. However, the effectiveness of LiDAR observations is vulnerable to disruptions from external factors such as rain, snow, and dynamic obstacles. These disturbances directly contribute to a diminished level of positioning accuracy. In order to address this issue, the incorporation of a multi-kernel maximum correntropy error state Kalman filter (MMC-ESKF) is proposed. By leveraging the maximum correntropy Kalman filter (MC-KF), the MMC-ESKF method effectively mitigates the impact of external disturbances. In addition, measurement-specific suppression is achieved by customizing different kernel bandwidth for each specific channel. The proposed algorithm was validated through localization experiments, revealing its superior performance compared to conventional ESKF and fixed-kernel MC-ESKF approaches. Remarkably, the root mean square error (RMSE) of position and attitude are 0.146m and 2.766°, respectively.