Initial alignment of large misalignment angle in strapdown inertial navigation system based on Gaussian process regression
Xijing Zhao, Lixin Wang, Zhikun He, Bo Zhang, Han Zhao · 2014
Due to the large misalignment angle, the error model of a strapdown inertial navigation system (SINS) is nonlinear. To solve the problem of the model inaccurateness and the influences on alignment accuracy caused by the linearization of the nonlinear model, the nonlinear model for initial alignment of large misalignment angle in SINS is established. To improve the accuracy of the initial alignment under the nonlinear error model, Gaussian process regression (GPR) method is applied in the initial alignment of SINS and a novel initial alignment algorithm for large misalignment angle based on GPR is proposed. The algorithm uses the square-root cubature Kalman filter (SRCKF) to simulate training data for GPR. The measurements of SRCKF are set as the training input and the SRCKF estimated outputs of misalignment angles are set as the training output. To identify the misalignment angles of SINS, GPR is used to learning the nonlinear mapping relationship between inputs and outputs of training data. Simulation results verify the validity and feasibility of the novel algorithm in initial alignment.