基于SW-XGBoost的组合导航定位方法  

Integrated navigation and positioning method based on SW-XGBoost

作　　者：刘丁柯 胡晓丽 古天龙[1] 宾辰忠[1] LIU Dingke;HU Xiaoli;GU Tianlong;BIN Chenzhong(Guangxi Key Lab of Trusted Software,Guilin University of Electronic Technology,Guilin 541004,China;Practice and Experiment Station,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学广西可信软件重点实验室,广西桂林541004 [2]桂林电子科技大学教学实践部,广西桂林541004

出　　处：《桂林电子科技大学学报》2020年第5期377-382,共6页Journal of Guilin University of Electronic Technology

基　　金：国家自然科学基金(61862016);广西自然科学基金(2017GXNSFAA198283);广西高校中青年教师科研基础能力提升项目(2019KY0226)。

摘　　要：针对传统GPS/BD与惯性导航系统(INS)组合的导航技术容易出现过拟合及急转弯道路下定位精度不足的问题,提出一种基于滑动窗口的极限梯度提升决策树(SW-XGBoost)的组合导航定位方法。根据车辆当前与历史状态的相关性,采用滑动窗口来提升对惯性器件噪声的抑制能力,使用极限梯度提升决策树(XGBoost)方法建立车辆状态和定位误差的预测模型,并使用粒子群算法对XGBoost预测模型的参数进行优化。实际路段测试表明,与仅使用XBGoost的组合导航定位方法相比,该方法在急转弯路况下定位精度提高了29.2%。The navigation technology combined with traditional GPS/BD and inertial navigation system(INS)is prone to the problem of overfitting and insufficient positioning accuracy under sharp turning roads.Therefore,a combined navigation positioning method based on SW-XGBoost is proposed.According to the correlation between the vehicle's current and historical state,the sliding window is used to improve the noise suppression ability of inertial devices,the eXtreme Gradient Boosting(XGBoost)method is used to establish the vehicle state and positioning error prediction model,and the particle swarm optimization algorithm is used to optimize the parameters of the XGBoost prediction model.The actual section test shows that compared with the combined navigation and positioning method using only XBGoost,the method presented in this paper improves by 29.2%under the condition of sharp turning.

关 键 词：组合导航 XGBoost 滑动窗口 GPS中断 

分 类 号：TN911.72[电子电信—通信与信息系统]