基于因子图的自动驾驶融合定位研究

Research on Fusion Localization of Automated Driving Systems Based on Factor Graph

在自动驾驶系统的即时定位及地图构建问题中,当行车路况复杂多变时,如何对车辆位姿进行准确实时估计是一个关键问题。就该问题提出基于因子图的多传感器融合方法,在Matlab软件环境中实现车辆的准确实时定位。根据实际驾驶过程的特性选择试验采用的数据类型并建立相应的数学模型,分别根据GPS及IMU数据特征以及VO和IMU的数据特征建立相应的融合因子图模型,基于KITTI数据集,对所建立的融合因子图模型进行试验验证。结果表明,通过融合因子图模型可以准确地估计车辆当前时刻的位姿。

For the localization and map construction of automated driving systems,the key issue is to accurately estimate the vehicle pose for decision-making when the road conditions are complicated.This paper proposes a multi-sensor fusion method based on factor graph to achieve the accurate real-time localization of the vehicle by using Matlab.Initially,the data type used in the experiment was selected and the corresponding mathematical model was established.Next,the fusion factor graph model was established by fusing the VO and IMU data or by fusing the GPS and IMU data.Finally,the experimental verification of the model was carried out based on the KITTI dataset.The results show that the position and posture of the vehicle can be accurately estimated by the factor graph model.