Xiong Yufeng, Zhang Yu, Guo Xiaoting, Wang Chenguang, Shen Chong, Li Jie, Tang Jun, Liu Jun
Key Laboratory of Instrumentation and Science and Dynamic Measurement, Ministry of Education, North University of China, Taiyuan 030051, People's Republic of China.
Rev Sci Instrum. 2019 Jan;90(1):015101. doi: 10.1063/1.5079889.
In this paper, a seamless navigation dual-model based on Square-Root Cubature Kalman Filter (SRCKF) and Random Forest Regression (RFR) is developed to enhance the performance of the Global Positioning System (GPS)/Inertial Navigation System (INS) integrated navigation system. By using the proposed method, the system can ensure seamless navigation ability even during GPS signal outages. In the proposed dual-model, sub-model 1 that directly relates the specific force of INS to the measurement of filter and sub-model 2 that directly relates the cubature points and innovation of SRCKF to the error caused by filter are established. Combined with SRCKF and RFR algorithms, the dual-model system can predict and estimate the velocity and position of the vehicle seamlessly when GPS signals are blocked. Field test data are collected to evaluate the proposed solution, and the experimental results show that the model proposed has obvious improvement in navigation accuracy by comparison. The prominent advantages of the proposed seamless navigation method include the following: (i) the proposed dual-model can effectively provide corrections to standalone INS during GPS outages, which outperforms traditional widely used single model; (ii) the proposed combination of SRCKF and RFR achieves better performance in the prediction of INS errors than other combination algorithms.
本文提出了一种基于平方根容积卡尔曼滤波器（SRCKF）和随机森林回归（RFR）的无缝导航双模型，以提高全球定位系统（GPS）/惯性导航系统（INS）组合导航系统的性能。采用该方法，即使在GPS信号中断期间，系统也能确保无缝导航能力。在所提出的双模型中，建立了直接将INS比力与滤波器测量值相关联的子模型1，以及直接将容积点和SRCKF的新息与滤波器引起的误差相关联的子模型2。结合SRCKF和RFR算法，双模型系统在GPS信号受阻时能够无缝地预测和估计车辆的速度和位置。收集了现场测试数据来评估所提出的解决方案，实验结果表明，相比之下，所提出的模型在导航精度方面有明显提高。所提出的无缝导航方法的突出优点包括：（i）所提出的双模型能够在GPS中断期间有效地为独立INS提供校正，其性能优于传统的广泛使用的单模型；（ii）所提出的SRCKF和RFR的组合在INS误差预测方面比其他组合算法具有更好的性能。
