摘要
在SINS/GPS组合导航实际应用中GPS短时失效的情况难以避免,这会导致组合导航的效果下降。针对该问题,本文提出了基于偏最小二乘PLSR辅助高斯过程回归GPR的SINS/GPS组合导航的无味四元数估计器USQUE,以解决组合导航中的GPS短时失效问题。该方法以PLSR估计的位置误差作为输入,以GPS提供的位置信息作为输出对GPR进行训练。在组合导航系统出现GPS短时失效后,使用通过PLSR辅助粒子群算法优化超参数的GPR直接对辅助导航设备的位置进行预测,作为USQUE算法的量测量,从而使得USQUE算法可以正常进行量测更新。在实验中,使用车载MEMS/GPS数据,将PLSR-GPR-USQUE,PLSR-USQUE与隔离量测量的USQUE算法组合导航的效果进行比较。实验结果表明,在GPS短时失效的情况下,PLSR-GPR-USQUE具有良好的估计精度。
In the practical application of SINS/GPS integrated navigation, it is difficult to avoid the GPS short-term failure, which will lead to the decline of the integrated navigation effect. In order to solve this problem, a quad unscented Kalman filter(USQUE) based on partial least square(PLSR) assisted Gaussian process regression(GPR) for SINS/GPS integrated navigation is proposed to solve the problem of GPS short-term failure in integrated navigation in this paper. The position error estimated by PLSR is took as the input and the position information provided by GPS is took as the output to train the GPR. After the GPS fails for a short time in the integrated navigation system, the GPR optimized by PLSR auxiliary particle swarm optimization algorithm is used to directly predict the position of the auxiliary navigation equipment. As the measurement of the USQUE algorithm, the USQUE algorithm can update the measurement normally. In the experiment, PLSR-GPR-USQUE, PLSR-USQUE and isolation measure use the MEMS/GPS data to compare the effect of integrated navigation. The experimental results show that PLSR-GPR-USQUE has good estimation accuracy in the case of GPS short-term failure.
作者
张梦得
胡柏青
田佳玉
李开龙
ZHANG Mengde;HU Baiqing;TIAN Jiayu;LI Kailong(Naval University of Engineering,Wuhan 430033,China)
出处
《测绘科学技术学报》
北大核心
2019年第5期447-451,457,共6页
Journal of Geomatics Science and Technology
基金
国家自然科学基金项目(61703419)。
