Focusing on the issue of attitude tracking for low-cost and small-size Micro-Electro-Mechanical System (MEMS) Inertial Measurement Unit (IMU) in high dynamic environment, an Adaptive Unscented Kalman Filter (AUKF) method combining sensor fusion methodology with Artificial Neural Network (ANN) is proposed. The different control strategies are adopted by fusing multi-MEMS inertial sensors under various dynamic situations. The AUKF attitude determination approach utilizing the MEMS sensor and Global Positioning System (GPS) can provide reliable estimation in these situations. In particular, the adaptive scale factor is used to adaptively weaken or enhance the effects on new measurement data according to the predicted residual vector in the estimation process. In order to solve the problem that the new measurement data is not available in case of GPS fault, an attitude algorithm based on Radial Basis Function (RBF)-ANN feedback correction is proposed for AUKF. The estimated deviation of predicted system state can be provided based on RBF-ANN in GPS-denied environment. The corrected predicted system state is used for the estimation process in AUKF. An experimental platform was setup to simulate the rotation of the spinning projectile. The experimental results show that the proposed method has better performance in terms of attitude estimation than other representative methods under various dynamic situations.

中文翻译：

---

基于融合UKF和RBF神经网络的融合方法的Micro-IMU在旋转环境中的精确姿态确定

针对高动态环境下低成本，小型微机电系统（MEMS）惯性测量单元（IMU）的姿态跟踪问题，将传感器融合方法与人工仿真相结合的自适应无味卡尔曼滤波器（AUKF）方法提出了神经网络（ANN）。通过在各种动态情况下融合多MEMS惯性传感器来采用不同的控制策略。利用MEMS传感器和全球定位系统（GPS）的AUKF姿态确定方法可以在这些情况下提供可靠的估计。特别地，自适应比例因子用于在估计过程中根据预测的残差矢量自适应地减弱或增强对新测量数据的影响。为了解决GPS故障时无法获得新的测量数据的问题，针对AUKF提出了一种基于径向基函数（RBF）-ANN反馈校正的姿态算法。在GPS受限的环境中，可以基于RBF-ANN提供估计的系统状态估计偏差。校正后的预测系统状态用于AUKF中的估计过程。建立了一个实验平台来模拟旋转弹丸的旋转。实验结果表明，该方法在各种动态情况下的姿态估计性能均优于其他代表性方法。在GPS受限的环境中，可以基于RBF-ANN提供估计的系统状态估计偏差。校正后的预测系统状态用于AUKF中的估计过程。建立了一个实验平台来模拟旋转弹丸的旋转。实验结果表明，该方法在各种动态情况下的姿态估计性能均优于其他代表性方法。在GPS受限的环境中，可以基于RBF-ANN提供估计的系统状态估计偏差。校正后的预测系统状态用于AUKF中的估计过程。建立了一个实验平台来模拟旋转弹丸的旋转。实验结果表明，该方法在各种动态情况下的姿态估计性能均优于其他代表性方法。