论文部分内容阅读
针对全维无迹卡尔曼滤波(UKF)算法状态维数倍增所带来的计算负担加剧的问题,提出一种过程和量测不相关的状态切换UKF算法。该算法通过在预测和量测阶段选取不同的状态变量,降低实时滤波的状态维数及Sigma点的选取个数,减小了计算量,提高了运算速度。针对姿态确定中四元数规范化限制,给出一种参数切换算法,在滤波过程中通过四元数与修正罗德里格斯参数实时切换,解决了四元数加权均值和协方差奇异性问题。针对SINS/CCD姿态的仿真实验结果表明,与全维UKF算法相比,状态切换UKF算法估计精确度相当,估计时间缩短了约1/3。
In order to exacerbate the computational burden caused by the doubling of the state dimension of the dimensionless Unscented Kalman Filter (UKF) algorithm, a state-switched UKF algorithm with no correlation between process and measurement is proposed. By selecting different state variables in the prediction and measurement stages, the algorithm reduces the state dimension of real-time filtering and the number of selected Sigma points, reduces the computational complexity and increases the computational speed. Aiming at the limitation of quaternion normalization in attitude determination, a parameter switching algorithm is proposed. In the filtering process, quaternion and modified Rodrigues parameters are switched in real time to solve the quaternion weighted mean and covariance singularity problems. Simulation results for SINS / CCD pose show that compared with full-dimension UKF algorithm, the estimation accuracy of state-switched UKF algorithm is equivalent, and the estimation time is shortened by about 1/3.