【作者】 夏全喜；

【导师】 钱华明；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导与控制， 2010， 博士

【摘要】 军用车辆使用环境比较复杂,为了满足武器系统快速机动发射的需求,要求导航系统具有高精度、高可靠性、自主性和抗干扰性。任何单一的导航系统都难以满足这一要求,多传感器信息融合技术成为导航系统研究的主要技术途径。本文以陆战平台为应用背景,以提高导航系统定位精度、适应能力和可靠性为目标,设计了微捷联惯导系统(Micro Strapdown Inertial Navigation System, MSINS)、全球定位系统(Global Positioning System, GPS)和电子罗盘(Electrical Compass, EC)组合的导航系统,并对其相关的关键技术进行了研究,主要研究工作如下:基于FPGA和ARM搭建了MSINS/GPS/EC组合导航系统硬件平台。为了提高组合导航子系统的精度,完成了对微惯性组件(Micro Inertial Measurement Unit, MIMU)的整体标定和性能测试,运用Allan方差分析方法对陀螺仪随机噪声进行分析和辨识,完整全面地掌握了噪声特性。针对MEMS陀螺仪随机噪声较大,会严重影响系统精度的问题,首先对随机漂移建立AR模型,然后提出基于状态扩增法设计Kalman滤波器滤除随机噪声,试验验证了该方法的有效性。深入研究了捷联惯导系统状态变量的可观测度问题。针对广泛应用的奇异值分解方法只能分析状态的估计精度,不能分析收敛速度的问题,引入主元分析方法,提出了基于主元分析的可观测度分析算法。该方法利用状态变量在主元子空间的投影来确定状态的可观测度信息,定义了估计精度和估计速度两项指标,分别求取两项指标的估计值。从奇异值分解的物理意义入手,简化了可观测度求解方法,全面地获取状态的可观测度信息。仿真和试验结果验证了该方法的有效性。对组合导航系统的数据融合算法进行了研究。为了降低GPS和EC的量测噪声,提出一种量测值平均化处理法和强跟踪Kalman滤波器相结合的算法。根据滤波器收敛性判据动态确定平均化时间,在保证滤波器收敛的前提下,使平均化时间最长,解决了传统量测值平均化方法难以确定平均化时间的问题,从而最大幅度降低了量测噪声,提高了系统精度。为了提高组合系统在GPS和EC无效时的定位精度,研究了利用车辆运动学约束辅助MSINS工作的导航算法。提出一种将运动学约束辅助法、方位陀螺漂移消减法以及交互多模型法相结合的导航算法。利用运动学约束辅助法(Motion Constraints Aided,MCA),引入约束条件,构造虚拟观测量,设计滤波器进行状态估计,提高了位置、速度的估计精度;采用方位陀螺随机漂移消减算法(Heuristic Drift Reduction, HDR),设计闭环网络,对方位陀螺仪的随机漂移进行实时估计和补偿,提高了航向角的估计精度。将两种算法并列执行,其结果作为交互多模型算法(Interactive Multiple Model, IMM)中各子滤波器的观测量,并以载体运动方式构建系统方程,设计滤波器进行滤波,进一步提高了系统精度。更多还原

【Abstract】 The application environment of military vehicle is very complex,in order to meet the needs of rapid mobile launch of weapon system,navigation system should have good performance such as high accuracy, high reliability, autonomy and anti-jamming. Any single navigation system is hard to meet this requirement, so multi-sensor information fusion technology has been the main technical means of navigation systems. In this paper, considering military platform as application background, aiming as to improve the navigation system accuracy and the ability to adapt complex environment and high reliability, SINS system, GPS system and EC integrated navigation system are designed. Moreover, this paper did some research on their key technologies. The main research works of this paper are demonstrated as follows:Based on FPGA and ARM, an integrated navigation system hardware platform of MSINS/GPS/EC is designed. The performance of the system has been tested by experiments. Experimental verification is done respectively and overall calibration of MIMU. Random noise of gyroscope has been analyzed and identified with Allan variance, which contributes to grasp the noise characteristics completely and comprehensively. In order to solve the problem of the system accuracy which is influenced enormously by MEMS random drift error, a gyro random noise AR model has been established. Also, state augmenting method is proposed to design Kalman filter, which is used to filter the random noise. After testing and verification this approach is able to greatly reduce the random noise of gyro, moreover, it lays the foundation for improving the system accuracy.The observable degree of SINS has been studied. The Singular Value Decomposition(SVD) method of observability matrix is widely used in the field. However, this method could only analysis the estimate accuracy of state vector, but not the convergence rate. In order to solve this problem, Relative Principal Component Analysis (RPCA) algorithm is introduced, and a new observable degree analysis algorithm based on RPCA is proposed. The estimate accuracy and estimate velocity of observable degree is distinguished. Obtain the two estimates mentioned respectively, so the observable degree information of state could become more completely and comprehensively. Analysis from the physical meaning of singular value, simplified the solution of observable degree. Simulations and experimental results demonstrate the effectiveness of this method. Moreover, it is simpler and easier to implement compared with the traditional SVD method.The data fusion method of integrated navigation system has been studied. A Federal Kalman filter has been applied in SINS/GPS/EC integrated navigation system. Considering the low accuracy of short time output of GPS and EC, measurement averaging method is introduced. However, the average time is hard to determine in conventional method. In order to solve this problem, a new algorithm combined strong tracking kalman filter with convergence criterion of tilter is proposed. This method can reduce observable noise and improve the accuracy of system without any loss of the filter divergence.Due to GPS and EC is vulnerable to interference; this paper studied Navigation algorithm for low-precision inertial when it works alone. Vehicle model and movement characteristics are introduced as constraint information to improve system accuracy. At first, considering the lateral and vertical velocity of vehicle is usually zero, the constraint information was introduced as virtual measurement to design filters, which can estimate state vector and largely improve the system accuracy. Secondly, in order to reduce the rapid growth error of azimuth, HDR algorithm is introduced. A closed loop network is designed to estimate and compensate the random drift of azimuth gyroscope real-time. At last, Interactive Multiple Model (IMM) is proposed to apply in SINS, with the equations of vehicle motion as system equation and output of SINS as measurement. Simulation and experimental results demonstrate the algorithm can further improve the system accuracy.更多还原