车载紧耦合MIMUs/GPS的神经网络辅助强跟踪滤波方法（英文）

Neural network collaborative strong tracking filtering for vehicular tightly-coupled MIMUs/GPS

微惯性测量单元由三轴正交的微机械陀螺、加速度计和微型地磁传感器组成。将上述装置与GPS接收机组合,可构成最佳导航定位模型,其中紧耦合MIMUs/GPS对全导航参数(位置、速度及姿态)的测量精度可大幅提高。由于微惯性传感器具有大漂移特性,为获得具有自适应的线性参数模型,提出了融合滤波的信息处理方法,利用强跟踪滤波实现状态预测,二阶EKF实现测量更新,并借用神经网络技术完成对状态预测的修正。由于系统组件具有非线性,该神经网络辅助的强跟踪滤波方法旨在逼近MIMUs/GPS的真实特性,并为车载用户提供更为精准的导航参数信息。动态环境下的仿真试验表明,尽管MEMS惯性传感器的精度有限,所提出的方法能够有效用于实际的导航参数解算。

The MEMS inertial measurement units（IMUs） consist of three orthogonal MEMS gyros, three orthogonal MEMS accelerometers, and three orthogonal micro magnetic sensors. Broadly speaking, an ideal positioning model of integrated navigation can be achieved by combining MIMUs（MEMS IMUs） with a miniature GPS receiver, and with this model, the performance of the tightly-coupled MIMUs/GPS mode in detecting position-velocity-acceleration（PVA） can be greatly improved. Specifically, due to the large drifts of micro inertial sensors, for the sake of a linear self-adaptive parameter model, a fusion filtering strategy is proposed, which fuses the strong tracking filter strategy in state predication, the quadratic EKF algorithm in measurement updates, and the artificial neural network（ANN） aided means in state estimation＇s correction. Since the components involved are nonlinear in nature, a collaborative strong tracking filtering design is supposed for this neural network to guarantee that the above information processing model is capable of approaching the real characteristics of MIMUs/GPS and providing higher accuracy estimates of navigation parameters for land vehicular users. Experiments under dynamic environments demonstrate that the novel filtering method can be effectively applied in practical applications of navigation parameter estimation, even if the accuracy of the MEMS inertial sensors is modest.