计算智能辅助模式下的车载MIMUs/GPS导航系统信息融合策略研究

The development and maturation of the MEMS technology opened the door to realize the promotion of low cost and low power dissipations to the navigation systems. The proposed research aims to constitute the vehicular MIMUs/GPS combination with miniature MEMS inertial units, and attempts to explore the strategies of multi-sensor information fusion in the tightly-coupled integration mode by using the fast Kalman filter algorithm together with the computational intelligent theories. The proposed research consists of four stages. Firstly, with fully considering the passive information on gravity filed and geomagnetic filed, the angle drifts of the MEMS gyroscopes’ integral will be corrected in the vehicular attitude measurements. Secondly, Our research program focuses on the practical applications of vehicular carriers, and the studies have been dedicated to the specific fields with respect to this type of system, especially on the issues of data processing and navigation calculating under the tightly-coupled framework. Thirdly, when it comes to the establishments of state model for the filter estimates, we always take into account the accuracy of the MEMS units & measured input related problems. Specifically, the ‘pseudo -pseudo _ range rate-heading’ combined observation model will be proposed and established, and the general design theory of Hessian Matrix based upon QEKF (Quadratic Extended Kalman Filter) will be deduced. Finally, the filter strategies which fuse the non-linear Gaussian filter and computational intelligent means will be performed to decrease the non-linear error, correct the estimates of the state equations, and improve the real-time performances and accuracy of the navigation system. Overall, our analyses will provide valuable technical supports and guidelines for the design of tightly-coupled navigation systems.

MEMS技术的发展成熟大大促进了导航系统的低成本、低功耗。本课题运用微小型的MEMS器件构成车载MIMUs/GPS，借助Kalman滤波快速算法，并辅以计算智能理论，探讨紧耦合模式下的多传感器信息融合策略。研究内容包括：在车载姿态测量中，充分利用重力场和地磁场两类无源信息，修正MEMS陀螺积分后的角度漂移；结合陆地载体应用实际，对组合系统工作在紧耦合状态下的数据处理与导航解算作出专门研究；滤波估计状态模型的建立始终考虑MEMS器件精度与量测输入相关问题，提出并搭建基于“伪距-伪距率-航向角”的组合观测模型，推导QEKF下的通用Hessian（海森）矩阵计算理论；进一步开展非线性高斯滤波与计算智能方法的融合策略分析，减小系统非线性误差、修正滤波器状态方程预报值、提高系统实时性与滤波精度，为紧耦合导航系统设计提供技术支持。

MEMS技术的发展成熟使导航系统的低成本、低功耗成为可能。本课题运用微小型的MEMS器件构成车载MIMUs/GPS，借助Kalman滤波快速算法，并辅以计算智能理论，探讨紧耦合模式下的多传感器信息融合策略。（1）以锥运动作为检验MIMUs姿态解算算法优劣的机体工作环境，经分析、推演典型等效旋转矢量算法，提出分别从陀螺采样方式、补偿项的扩展以及补偿系数求解等三方面对其进行优化设计，确保车载MIMUs长时姿态更新精度；（2）针对嵌入在导航系统中的磁罗盘易受到外部环境干扰问题，提出采用带有椭球约束的最小二乘法分段法，及牛顿-拉夫逊迭代对磁场干扰补偿系数进行分步求取，减小外部磁场干扰对相关参数估计的影响；（3）针对微惯性传感器具有大漂移特性，提出紧耦合MIMUs/GPS系统融合滤波策略，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，为车载用户提供更为精准的导航参数信息；（4）针对GPS失锁等信号不可用的情况，提出Adaboost优化的神经网络辅助常规Kalman滤波，修正MIMUs单机工作下逐渐积累的导航参数误差；（5）针对MIMUs/GPS观测噪声统计特性随时间及周围环境的变化而变化的特点，提出一种基于正态云模型的模糊自适应滤波方法(NCMFAF)。实时监测系统理论残差与实际残差的协方差差异程度，对观测噪声方差阵系数进行自适应调整。搭建了车载MIMUs/GPS组合导航仿真平台，并通过数字仿真及转台实验测试等对上述方法进行验证。本课题的完成为紧耦合导航系统设计提供了技术支持，具有重要理论意义和实际应用价值。