基于MEMS-IMU的动态大失准角STUKF算法

STUKF algorithm for MEMS-IMU with large misalignment in dynamic

针对微机电惯性测量单元(MEMS-IMU)器件零偏误差大以及传统滤波算法在动态大失准角情况下姿态估计精度差、误差收敛慢的情况,提出了一种状态变换无迹卡尔曼滤波(STUKF)算法。首先,建立了旋转矢量非线性姿态误差模型。然后,通过误差状态变换构建了考虑姿态误差的速度误差状态,使速度误差微分方程中的比力相关项转换成了相对稳定的重力项,避免了无迹卡尔曼滤波(UKF)算法在剧烈角运动、线运动应用场景中计算的误差状态方差与实际不一致的情况。最后,以小型无人船载MEMS-IMU/卫星组合航姿系统为应用场景,经数学仿真及三轴模拟转台实验验证,以速度、位置信息为观测量,在动态大失准角情况下,STUKF姿态误差的估计精度及收敛速度优于UKF和状态变换扩展卡尔曼滤波算法(STEKF)。在三轴转台实验中,在精对准900 s后,STUKF的航向角估计误差收敛到1°(1σ)以内,优于UKF的2°(1σ)及STEKF的3°(1σ)。

Considering the significant sensor bias error of the micro-electro mechanical inertial measurement unit(MEMS-IMU)and the inaccurate estimation of vehicle attitude with slow error convergence speed using traditional filtering algorithms under the condition of large misalignment in dynamic,a state transformation unscented Kalman filter(STUKF)algorithm is proposed.Firstly,a rotation vector nonlinear attitude error model is established.Then,the velocity error state considering the attitude error is constructed through error state transformation,by which the specific force related term in the velocity error differential equation is replaced with a relatively stable gravity term.In this way,the inconsistency between the calculated error state variance and the practical value in the unscented Kalman filter(UKF)algorithm is avoided especially under severe angular and linear motion of the vehicle.Finally,taking the MEMS-IMU/GNSS integrated heading and attitude reference system for a small unmanned ship as application background,and verified by numerical simulation and triple-axis simulation turntable experiments,the estimation accuracy and error convergence speed of vehicle attitude using STUKF are better than those using either UKF or existing state transformation extended Kalman filter(STEKF)under the condition of large misalignment in dynamic with the observation of velocity and position information.In the triple-axis turntable experiments,after 900 s fine alignment,the yaw angle estimation error of STUKF converges to less than 1°(1σ),which is better than 2°(1σ)of UKF and 3°(1σ)of STEKF.