CN102819030B - Method for monitoring integrity of navigation system based on distributed sensor network - Google Patents

Abstract

The invention discloses a navigation system integrity monitoring method of a distributed sensor network, which belongs to the technical field of navigation and positioning. The integrity monitoring of the navigation system based on the distributed sensor network is carried out by adopting the hierarchical processing method of the integrity monitoring processing of the sensor level and the integrity monitoring processing of the system level. In the sensor-level integrity monitoring stage, the GNSS receivers are monitored using the RAIM method, and k SRIMU network nodes are monitored using a comprehensive method based on the moving window-parity vector method and the discrete wavelet transform method; In the integrity monitoring stage, integrity monitoring is carried out based on the new information processing method and the moving window information processing method. The method of the invention can effectively monitor step faults and slope faults from distributed sensor network nodes to the entire distributed navigation system level, and comprehensively enhance the integrity performance of the navigation system based on the distributed sensor network.

Description

Integrity Monitoring Method of Navigation System Based on Distributed Sensor Network

technical field

The invention relates to a navigation system integrity monitoring method based on a distributed sensor network, and belongs to the technical field of navigation system integrity monitoring.

Background technique

Integrity refers to the ability of the navigation system to provide timely and effective warning information when the error caused by failure or performance deterioration exceeds the acceptable limit value (warning threshold) during the use of the navigation system. In order to ensure the reliability of the navigation system, the integrity monitoring of the navigation system is required, and its main purpose is to detect and isolate faults. Integrity monitoring performs processing such as detection statistics and threshold judgment through the analysis of redundant information such as hardware and software.

At present, there have been many studies on the integrity monitoring of navigation systems at home and abroad, which are usually divided into snapshot method (snapshot) and continuous method (sequential). The snapshot method uses the measurement information of a single epoch to detect and isolate instantaneous step faults, and is usually used for faults with large changes. Typical methods include the least squares residual method, parity vector method, etc. In addition, extensive research at home and abroad GNSS (Global Navigation Satellite System) Receiver Autonomous Integrity Monitoring (RAIM) is also a snapshot method. The snapshot method can detect step faults of navigation sensors or GNSS signals, but cannot detect slowly changing ramp faults caused by inertial sensor drift, etc. For the detection of slow-changing slope faults, continuous methods based on historical accumulated information are usually used, such as continuous probability ratio detection method (SPRT), dynamic model-based method, etc., but the current algorithm is time-consuming, even reaching tens of minutes. Brenner et al. gave the multi-solution separation method (MSS) based on the Kalman filter group, and performed integrity detection according to the Kalman filter of all measurement sets and different measurement subsets, and applied it to IN/GPS/atmospheric data of Honeywell Company Hybrid navigation system (HIGH); Diesel et al. gave an autonomous integrity extrapolation method (AIME), applied to Litton's GPS/IRS combined system. For GNSS receiver navigation, the RAIM method is a relatively effective integrity monitoring method commonly used at present; for inertial navigation integrity monitoring, GLRT (generalized relief ratio) method, odd-even vector method, etc. are usually used; for multi-sensor combinations, such as GNSS and Combination of inertial navigation systems, MSS and AIME are integrity monitoring methods that currently have reported engineering applications.

The navigation system using a distributed sensor network is a new navigation system design concept. It is a new generation of low-cost, small-volume, and light-weight navigation sensors in recent years, such as MEMS (micro-electromechanical system) inertial sensors, MSIS ( Micro solid-state inertial sensors), fiber optic gyroscopes, GNSS receivers, etc., as well as new technologies developed on the basis of high-speed and large-capacity embedded microprocessors and distributed modular electronic equipment. It configures multiple inertial sensor systems in multiple positions of the carrier (such as aircraft, ships, large spacecraft, etc.) to form a distributed inertial network topology, which can not only provide redundant distributed measurement information for carrier navigation, Moreover, it provides a local measurement system for the electronic equipment of the carrier, such as radar tracking, equipment load, etc., and can also provide inertial state information for local motion compensation of the carrier electronic equipment. The navigation structure based on the distributed sensor network can improve the level of fault tolerance and dynamically configure the sensor system functions by reconstructing and sharing limited computing resources.

Current integrity monitoring methods are usually aimed at individual inertial navigation systems, or individual GNSS navigation systems, or multi-sensor navigation systems that are suitable for traditional centralized filtering or federated filtering structures, but cannot be directly applied to distributed navigation systems middle. In addition to step faults caused by electronic devices and mechanical components, inertial sensors usually have slowly changing drifts, and the motion status of each network node is not uniform. The usual integrity monitoring methods cannot be directly applied to distributed sensors. network structure. For navigation systems based on distributed sensor networks, there is currently no effective integrity monitoring method to ensure the overall performance of the entire navigation system.

Contents of the invention

The present invention proposes a navigation system integrity monitoring method based on a distributed sensor network for a novel navigation system based on a distributed inertial sensor network, which overcomes the disadvantage that the existing integrity monitoring method cannot be directly applied to a distributed navigation system, Improve the integrity of distributed navigation systems.

The present invention adopts following technical scheme for solving its technical problem:

A method for integrity monitoring of a navigation system based on a distributed sensor network adopts a hierarchical processing method of sensor-level integrity monitoring processing and system-level integrity monitoring processing to monitor the integrity of a navigation system based on a distributed sensor network. Among them, the navigation system based on distributed sensor network includes GNSS (Global Navigation Satellite System) receiver, k SRIMU (Slanted Redundant Inertial Measurement Unit) network nodes, k is a natural number, each network node can have the same performance or different In the navigation processing, the information of other network nodes is shared for information fusion. One of the SRIMU network nodes is also fused with the information of the GNSS receiver, which has higher navigation performance and acts as the master node. The processing steps of integrity monitoring and navigation calculation are as follows:

(1) In the sensor-level integrity monitoring stage, GNSS receivers are monitored using the RAIM (Receiver Autonomous Integrity Monitoring) method, and the measurement information of k SRIMU network nodes is sent to the FDI of k SRIMU network nodes respectively. (fault detection and isolation) processing unit to perform fault detection and isolation processing;

(2) The inertial information processed by the FDI processing units of k SRIMU network nodes is respectively input into k inertial measurement fusion units, and the inertial information of SRIMUs that have been monitored by the sensor-level integrity is fused to obtain relative to the three Inertia information for the calculation of the axis-orthogonal coordinate system;

(3) Input the calculated inertial information after fusion processing of k inertial measurements into k local KF (Kalman filter) for local navigation information calculation. Among them, each local KF receives all the shared inertial measurement fusion information; in addition, the local KF of the master node will also fuse the GNSS receiver information after RAIM monitoring, which has higher performance than the navigation solution of other filters.

(4) Input the innovations of k local KFs into the system-level integrity monitoring processing unit, use the integrity monitoring method based on innovation processing to monitor the system-level integrity of the navigation system, and send the integrity information to In the k local navigation state update units.

(5) Finally, k local navigation state update units receive the same type of navigation state information of k local KFs, perform fusion processing, and obtain the final updated navigation information. In the k local navigation state updates, according to the integrity information provided by the system-level integrity monitoring process, if a local KF has a fault, the navigation state information of the local KF is eliminated in the fusion process.

The beneficial effects of the present invention are as follows:

1. Currently, the navigation system based on the distributed sensor network is a new navigation system design concept, and there is no effective integrity monitoring method for this type of navigation system. The present invention can solve this problem.

2. Using the hierarchical processing method of sensor-level integrity monitoring processing and system-level integrity monitoring processing, from the distributed sensor network nodes to the entire navigation system level, the integrity performance is comprehensively enhanced.

3. The designed integrity monitoring algorithm has good real-time performance, high reliability, and small amount of calculation. It can not only detect and isolate fast-changing step faults, but also detect and isolate slow-changing ramp faults.

4. In sensor-level integrity monitoring, the comprehensive method of MW-PV method and discrete wavelet transform method can not only detect and isolate multiple faults in turn, but also overcome the single parity vector method. For 4 sensors, it can only The disadvantage of detecting but not isolating faults (when there is only 1 redundant observation, the parity vector method cannot diagnose which sensor the fault occurs on), when there is only 1 redundant observation, it can still effectively detect and isolate faults. At the same time, using the moving window (MV) method, the slowly changing slope fault can be further detected and isolated.

5. On the main network nodes, use external auxiliary navigation sensors of GNSS receivers. Utilizing the fact that local KF innovations and GNSS actual measurement are independent, by monitoring local KF innovations of all network nodes, system-level integrity monitoring is realized.

6. In the system-level integrity monitoring, the residual inspection based on filter innovation and the innovation moving window method can be used to detect not only the fast-changing step faults at the system level, but also the slow-changing faults at the system level. ramp fault detection.

Description of drawings

Fig. 1 is a schematic diagram of the implementation of the navigation system based on the distributed sensor network of the present invention.

Fig. 2 is a flow chart of the integrity monitoring of the navigation system based on the distributed sensor network of the present invention.

FIG. 3 is a schematic diagram of the SRIMU processing flow for sensor-level integrity monitoring of the present invention.

FIG. 4 is a schematic diagram of a system-level integrity monitoring process flow in the present invention.

Detailed ways

The invention will be described in further detail below in conjunction with the accompanying drawings.

, Implementation route of navigation system based on distributed sensor network

As shown in Figure 1, taking an aircraft motion carrier as an example (other motion carriers such as ships, large spacecraft, etc. are similar), k sensor network nodes composed of SRIMU, k is a natural number, distributed configuration in the aircraft The multiple positions of the network constitute a distributed sensor network topology, which can not only provide redundant distributed measurement information for the navigation of the carrier, but also provide a local measurement system for the electronic equipment of the carrier, such as radar tracking and equipment load, and at the same time Inertial state information for local motion compensation of the carrier electronics can also be provided. The navigation structure based on the distributed sensor network can improve the level of fault tolerance and dynamically configure the sensor system functions by reconstructing and sharing limited computing resources. Usually the main navigation equipment of the aircraft is located at its center, and is also equipped with other navigation aid systems such as GNSS receivers, which are regarded as the main network nodes in the present invention. Therefore, in the local KF, in addition to making full use of each inertial measurement fusion information in the observation, GNSS observation information has also been added.

, Overall scheme of navigation system integrity monitoring based on distributed sensor network

As shown in Figure 2, for the integrity monitoring of the navigation system based on the distributed sensor network, a hierarchical processing method of sensor-level integrity monitoring processing and system-level integrity monitoring processing is adopted. Among them, the navigation system based on distributed sensor network includes GNSS (Global Navigation Satellite System) receiver, k SRIMU (Slanted Redundant Inertial Measurement Unit) network nodes, k is a natural number, each network node can have the same performance or different In the navigation processing, the information of other network nodes is shared for information fusion. One of the SRIMU network nodes is also fused with the information of the GNSS receiver, which has higher navigation performance and acts as the master node. In order to obtain the best system integrity monitoring effect, in this embodiment, the SRIMU network node 1 at the center of the carrier is selected as the master node. In the stage of sensor-level integrity monitoring, the integrity monitoring of the GNSS receiver is carried out using the RAIM (Receiver Autonomous Integrity Monitoring) method, and the measurement information of k SRIMU network nodes is sent to the FDI (fault detection and isolation) processing unit to perform fault detection and isolation processing; the inertial information processed by the FDI processing units of k SRIMU network nodes are respectively input into k inertial measurement fusion units, and the SRIMUs that have undergone sensor-level integrity monitoring The inertial information of k inertial measurements is fused to obtain the calculated inertial information relative to the three-axis orthogonal coordinate system; the calculated inertial information after k inertial measurement fusion is input into k local KF (Kalman filter) for local For navigation information calculation, each local KF receives all the shared inertial measurement fusion information, and in the local KF of the master node (local KF1 in this specific embodiment), the GNSS receiver information after RAIM monitoring will also be fused. The navigation solution of other filters has higher performance; the innovation of k local KFs is input into the system-level integrity monitoring processing unit, and the integrity monitoring method based on innovation processing is used to perform system-level integrity of the navigation system Integrity monitoring, and send integrity information to k local navigation state update units; finally, k local navigation state update units receive the same type of navigation state information (ie position, velocity, attitude information) of k local KFs , to perform fusion processing to obtain the final updated navigation information. In the update of the k local navigation states, according to the integrity information provided by the system-level integrity monitoring process, if there is a fault in a certain local KF, it will be eliminated in the fusion process Navigation state information of this local KF.

, sensor-level integrity monitoring processing

As shown in Figure 3, in the sensor-level integrity monitoring process, the GNSS receiver adopts the usual RAIM method for integrity monitoring. The present invention focuses on the integrity monitoring of SRIMU, and adopts the following steps:

(1) Overall steps

a. Create observation equation

，n为大于3的自然数(当n=3时为最小配置，此时不具备故障检测和隔离能力)，首先将n个传感器信息发送到基于MW-PV（移动窗口-奇偶向量）法的故障检测处理单元中，建立观测方程如下 Note that the number of sensors of the a-th obliquely mounted redundant inertial measurement unit is n, where,

, n is a natural number greater than 3 (when n=3, it is the minimum configuration, and it does not have the ability to detect and isolate faults at this time), first send n sensor information to the fault based on the MW-PV (moving window-parity vector) method In the detection processing unit, the observation equation is established as follows

     （1）                                                

(1)

为真实的状态向量（三轴角速度、三轴加速度等）；

为n个传感器的安装矩阵；

为n维的故障向量，当第i（

）个传感器出现故障时，

的第i个元素

为非零值，否则为零；

为传感器的测量噪声。 in, is an n-dimensional measurement vector;

is the real state vector (three-axis angular velocity, three-axis acceleration, etc.);

is the installation matrix of n sensors;

is an n-dimensional fault vector, when the i-th (

) sensor failure, the

the ith element of

non-zero value, otherwise zero;

is the measurement noise of the sensor.

． Calculate parity vector

The parity vector of formula (1) can be expressed as

                                                     （2）

(2)

为n-3维奇偶向量，它直接反映了故障的偏差信息；

为（n-3）×n维奇偶空间矩阵，具有如下性质：

，

，其中

为n-3维零矩阵，

为n-3维单位矩阵。因此，

为安装矩阵

的零空间矩阵，在本发明中通过对

转秩矩阵的奇异值分解（SVD）得到 in,

is an n-3 odd-even vector, which directly reflects the deviation information of the fault;

It is a (n-3)×n odd-even space matrix with the following properties:

,

,in

is n-3 dimensional zero matrix,

is the n-3 dimensional identity matrix. therefore,

for installation matrix

The null space matrix of , in the present invention by

Singular value decomposition (SVD) of the trans-rank matrix is obtained

                             （3）

(3)

为3×3维酉矩阵；Σ为半正定3×n维对角矩阵；是n×n维酉矩阵，

为其共轭转置；

为对角矩阵，其对角线上的元素即为

的奇异值；

为

的前3行（即的前3列）；

为

的后n-3行，即由

零空间的张成。因此，奇偶空间矩阵

为 in,

is a 3×3-dimensional unitary matrix; Σ is a positive semi-definite 3×n-dimensional diagonal matrix; is an n×n dimensional unitary matrix,

its conjugate transpose;

is a diagonal matrix, the elements on the diagonal are

singular value of

for

The first 3 lines of first 3 columns);

for

The last n-3 lines, that is, by

Zhang Cheng of Zero Space. Therefore, the parity space matrix

for

                                                      （4）

(4)

。由式（1）、（2）和（4）可得奇偶向量为 At this time, there are

. From equations (1), (2) and (4), the parity vector can be obtained as

                  （5）

(5)

c. Calculate detection statistics

为故障

与噪声

的函数，与状态量

无关。当传感器无故障时，

为零均值的n-3维正态分布白噪声序列，其方差为 From formula (5), we can see that the parity vector

for failure

with noise

function, and the state quantity

irrelevant. When the sensor is not faulty ,

is an n-3-dimensional normal distribution white noise sequence with zero mean, and its variance is

                                 （6）

(6)

为噪声标准差。当某个传感器出现故障时，

不再是零均值的白噪声，其均值为

，方差为。因此，可定义检测统计量为 in,

is the noise standard deviation. When a sensor fails,

is no longer white noise with zero mean, its mean is

, with a variance of . Therefore, the detection statistic can be defined as

                                                （7）

(7)

服从自由度为n-3的中心化

分布；当出现故障时，

服从非中心化

分布，设非中心化参数为

。 When the sensor is not faulty,

Subject to centralization with n-3 degrees of freedom

distribution; when a failure occurs,

subject to decentralization

distribution, let the non-centralization parameter be

.

Calculate detection threshold

From the parity vector and detection statistics, the following assumptions are made:

According to the assumed conditions, when there is no fault, the SRIMU is in the normal detection state, and if there is an alarm, it is a false alarm. When the false alarm rate PFA is given, there is

                       (8)

(8)

。通过比较检测统计量 

与检测门限

，如果

则表明存在故障，否则无故障。 The detection threshold can be obtained from the above formula

. By comparing detection statistics

and detection threshold

,if

It indicates that there is a fault, otherwise there is no fault.

． Moving window handling

If the above-mentioned integrity monitoring of parity vector processing is used alone, it is very effective for fast-changing step faults, but the detection effect for slow-changing slope faults is not obvious. The present invention further adopts the method of moving window processing to detect slope faults, and establishes a first-in-first-out parity vector stack structure with a length of L on the basis of parity vector detection

                      （9）

(9)

时刻的奇偶向量，为

时刻的奇偶向量，

为

时刻的奇偶向量。此时，检测统计量为 in, for

The parity vector of moments, for

The parity vector of moments,

for

Parity vector of moments. At this point, the detection statistic is

                   （10）

(10)

与检测门限

，如果则表明存在故障，否则无故障。 Then by comparing the detection statistics

and detection threshold

,if It indicates that there is a fault, otherwise there is no fault.

． Fault Diagnosis Isolation

When a fault is detected, the measurement information is further sent to the comprehensive fault diagnosis and isolation processing unit based on MW-PV method and wavelet analysis method, and step faults or slope faults are diagnosed respectively according to parity vector and moving window processing. The signal is isolated, and the faulty sensor information is eliminated in the observation equation, and the observation equation is reconstructed. Among them, for the 1st to n-4th faults, the wavelet analysis method is not required for fault diagnosis and isolation, and for the n-3th fault (that is, when diagnosing and isolating faults in the last 4 sensors), due to the parity vector method cannot diagnose , the discrete wavelet transform method is used at this time. Design the fault diagnosis function of the i-th sensor as

                                  （11） ,

(11)

为奇偶空间矩阵

的第i个列向量。如果所有传感器均无故障，则所有的故障诊断函数都为0；如果第i个传感器出现了故障

，则第i个故障诊断函数为

。因此，对应于最大故障诊断函数的第i个传感器即可认为出现了故障，需要对其进行隔离。在故障隔离时，将第i个传感器的测量量从观测方程（1）剔除。 in,

is the parity space matrix

The ith column vector of . If all sensors have no faults, all fault diagnosis functions are 0; if the i-th sensor has a fault

, then the i-th fault diagnosis function is

. Therefore, the i-th sensor corresponding to the maximum fault diagnosis function can be considered to be faulty and needs to be isolated. During fault isolation, the measurement of the i-th sensor is removed from the observation equation (1).

(2) Processing flow

a. 1st failure

When the first fault occurs, the n-3 odd-even vector is calculated by the observation equation, and the step fault can be effectively detected according to the n-3 odd-even vector; then, the slope fault can be effectively detected by using the moving window processing; when When a fault is detected, step faults or slope faults are diagnosed respectively according to the n-3 odd even vector and moving window processing, the fault signal is isolated, fault sensor information is eliminated from the observation equation, and the observation equation is reconstructed.

． 2nd failure

When the second fault occurs, on the basis of the observation equation after the isolation of the first fault, the n-4 Odditch even vector is calculated, and the aforementioned processing steps are used to detect and isolate the step fault directly from the n-4 Odditch even vector, Ramp faults are further detected and isolated by moving window processing.

． 3rd failure to n-4th failure

Fault detection and isolation in sequence using the same steps as above

d. n-3th failure

When the n-3th fault occurs, the parity vector at this time has only 1 dimension. Since the fault can only be detected according to the 1-dimensional parity vector, which sensor is faulty cannot be diagnosed. Therefore, when the fault is detected according to the 1-odd even vector, in the fault diagnosis and isolation process, the multi-scale signal decomposition based on the discrete wavelet transform method is used to diagnose and isolate the fault.

Compared with Fourier transform and fast Fourier transform, wavelet transform is a local transform in time and frequency domain, which has the characteristics of multi-resolution analysis. For stable signals, narrow windows are used for signal transients or sudden changes (high frequency), and wide windows are used for slow signal changes (low frequency), which can effectively extract signal waveform features, and are known as digital microscopes. Wavelet analysis is especially suitable for analyzing and processing non-stationary signals because of its excellent characteristics of time-frequency multi-resolution analysis, and has been widely used in signal denoising and image processing. The invention adopts the wavelet analysis method to decompose the SRIMU output signal in multiple scales, so that the fault can be effectively diagnosed and isolated even when the parity vector method cannot be used.

，其中表示信号分解的第

级尺度，N表示第N个离散时间步，它可以被分解为近似信号部分

和详细信号部分

Record the discrete signal sequence of the sensor for fault diagnosis as

,in Indicates the second part of the signal decomposition

level scale, N represents the Nth discrete time step, which can be decomposed into approximate signal parts

and detailed signal section

， 

（12）

,

(12)

和

分别为低通高通滤波器和高通滤波器系数，可以由尺度函数和小波函数

的2尺度关系得到 in,

and

are the low-pass high-pass filter and the high-pass filter coefficients respectively, which can be determined by the scaling function and wavelet function

The 2-scale relation of

，

         （13）

,

(13)

。本发明的小波函数采用Daubechies小波，具体分解算法步骤如下： in,

. Wavelet function of the present invention adopts Daubechies wavelet, and concrete decomposition algorithm step is as follows:

得到的近似信号

，另一组是作用高通滤波器

得到的细节信号

，这两个信号都是原信号在滤波器作用下以尺度2的下采样。低频部分表征信号本身特征，高频部分表征信号的细微差别。 In the first step, for a given original signal of length K , according to the formula (12), two sets of data are generated, and one set is a low-pass filter

The approximate signal obtained

, and the other set is to act as a high-pass filter

get detail signal

, both signals are the downsampling of the original signal with a scale of 2 under the action of the filter. The low-frequency part represents the characteristics of the signal itself, and the high-frequency part represents the nuances of the signal.

，利用上述的方法再次分解，直到所需要的层数。在分解过程中为对信号做下采样，则信号长度保持不变。 In the same way as in the second step, the low-frequency part signal obtained in the first step

, using the above method to decompose again until the required number of layers. In order to downsample the signal during decomposition, the signal length remains unchanged.

步内完成。对小波分解后的信号，根据诊断阀值进行诊断，如果超过诊断阀制，则认为该传感器出现故障。其中诊断阀值为 For a signal of length K , the entire algorithm is at most

completed in one step. The signal decomposed by wavelet is diagnosed according to the diagnostic threshold, if it exceeds the diagnostic threshold, it is considered that the sensor is faulty. where the diagnostic threshold is

                                               （14）

(14)

为该传感器的测量噪声标准差；K为离散信号序列的长度；为安全系数，其选择可以根据系统特性和导航系统运行环境确定。 in,

is the measurement noise standard deviation of the sensor; K is the length of the discrete signal sequence; It is a safety factor, and its selection can be determined according to the system characteristics and the operating environment of the navigation system.

, inertial measurement fusion and local KF

，采用加权最小二乘法进行求解，得到各个网络节点的计算惯性测量估计信息

。

为真实状态

的估计。 In the inertial measurement fusion stage, each SRIMU network node rebuilds the observation equation for the sensor measurement information after integrity monitoring and processing: , that is, the fault has been eliminated in equation (1)

, using the weighted least squares method to solve, and obtain the computational inertial measurement estimation information of each network node

.

for the real state

estimate.

In the local KF stage, the inertial measurement and estimation information of each network node is combined with the inertial measurement and estimation information of other nodes to construct a local Kalman filter to solve the local navigation state estimation of the network node.

，其中

表示该节点的局部导航状态，即3维的位置、速度、姿态误差共9维状态向量；

为传感器误差，即陀螺和加速度计误差，对于主节点的局部KF（本具体实施方式中主节点的局部KF为KF1）中还包含GNSS的钟漂和频漂误差。观测量为

，为其它节点惯性测量估计信息与网络节点的差分残差向量，对于局部KF1还包含伪距差向量。则局部KF模型为 For the local KF of the kth node, let its state quantity be

,in

Represents the local navigation state of the node, that is, the 3-dimensional position, velocity, and attitude error, a total of 9-dimensional state vectors;

is the sensor error, that is, the error of the gyroscope and the accelerometer, and the local KF of the master node (the local KF of the master node in this specific embodiment is KF1 ) also includes GNSS clock drift and frequency drift errors. Observed as

, is the difference residual vector between the inertial measurement estimation information of other nodes and the network node, and also includes the pseudorange difference vector for the local KF1. Then the local KF model is

                （15）

(15)

为

到

时刻的状态转移矩阵；

和

分别为系统噪声和测量噪声向量。基于卡尔曼滤波递推方程组,进行局部导航状态估计。 in,

for

arrive

The state transition matrix at each moment;

and

are the system noise and measurement noise vectors, respectively. Based on the Kalman filter recursive equations, the local navigation state estimation is carried out.

, System-level integrity monitoring and processing

As shown in Figure 4, system-level integrity monitoring is performed by monitoring the new information of the distributed local KF. The local KF receives the inertial measurement fusion information of all network nodes for Kalman filter calculation. In addition to the inertial measurement fusion information, the local KF1 also receives the GNSS pseudorange measurement information.

、

、……、，以及新息的方差

、

、……、

发送到基于新息处理的完好性监测单元，进行系统级完好性监测。 In the process of solving local KF1, local KF2, ..., local KFk, their filter innovations

,

,..., , and the variance of the innovation

,

,...,

Send it to the integrity monitoring unit based on the information processing for system-level integrity monitoring.

For the innovation process of any local KF, the filter innovation of epoch t is

                                               (16)

(16)

为t历元的量测；

为量测矩阵；

为一步预测值。

 类似于方程（5）中的奇偶向量。当局部KF系统无故障时，

为零均值的n维正态分布白噪声序列（n为观测向量的维数），其方差为 in,

is the measurement of epoch t;

is the measurement matrix;

is a one-step predictive value.

Similar to the parity vector in equation (5). When the local KF system is fault-free,

is an n-dimensional normal distribution white noise sequence with zero mean (n is the dimension of the observation vector), and its variance is

                                       (17)

(17)

将不再是零均值的白噪声。定义检测统计量为 in, is the one-step forecast mean square error; is the measurement noise variance matrix. When the local KF system fails,

will no longer be white noise with zero mean. Define the detection statistic as

(18)

 服从自由度为n的中心化

分布，当出现故障时

服从非中心化分布，设非中心化参数为

。检测门限

的计算与前述的传感器级的奇偶向量法完好性监测类似，如方程（8）所示，所不同的是自由度由n-3修改为n。通过比较检测统计量 

与检测门限

，如果

则表明存在故障，否则无故障。 When the local KF system is fault-free,

Subject to centralization with n degrees of freedom

distribution, when a failure occurs

subject to decentralization distribution, let the non-centralization parameter be

. detection threshold

The calculation of is similar to the integrity monitoring of the aforementioned sensor-level parity vector method, as shown in equation (8), the difference is that the degrees of freedom are changed from n-3 to n. By comparing detection statistics

and detection threshold

,if

It indicates that there is a fault, otherwise there is no fault.

一直很小，因此检测不灵敏。因此，本发明在系统级完好性监测中，也采用了移动窗口处理，对斜坡故障进行检测，在新息处理的基础上，建立一个长度为L的先进先出的新息向量堆栈结构 Similar to sensor-level integrity monitoring, based on the above-mentioned innovation-based processing, it is essentially a snapshot method, that is, it is processed according to the innovation of the current epoch, so it is very effective for fast-changing step faults, but for slow-changing slopes Fault, since the local KF is a recursive filter equation group, it will track the fault and cause

is always small, so the detection is not sensitive. Therefore, in the system-level integrity monitoring, the present invention also adopts moving window processing to detect slope faults, and establishes a first-in-first-out innovation vector stack structure with a length of L on the basis of the innovation processing

                            （19）

(19)

为t时刻的奇偶向量，

为

时刻的奇偶向量，

为

时刻的奇偶向量。此时，检测统计量为 in,

is the parity vector at time t,

for

The parity vector of moments,

for

Parity vector of moments. At this point, the detection statistic is

                      （20）

(20)

与检测门限，如果

则表明存在故障，否则无故障。通过该步的完好性监测处理，可以得到各个局部KF的完好性信息，并将完好性信息发送到局部导航状态更新单元中。 Then by comparing the detection statistics

and detection threshold ,if

It indicates that there is a fault, otherwise there is no fault. Through the integrity monitoring process of this step, the integrity information of each local KF can be obtained, and the integrity information is sent to the local navigation state update unit.

, local navigation state update processing

For each network node, a local information fusion filter is further designed to fully fuse the local KF information of other network nodes to update the local navigation status, so as to obtain higher performance navigation system results. Taking the local navigation state update of network node 1, network node 2, and network node 3 as an example, the local navigation state update equation of network node 1 is as follows

          （21）

(twenty one)

和

分别为节点1更新的局部导航状态及其均方差阵；

和

分别为节点1的局部KF估计值及其均方差阵；

和

分别为节点1的局部KF估计值及其均方差阵；

和

分别为节点2的局部KF估计值及其均方差阵；

和分别为节点3的局部KF估计值及其均方差阵；

和

分别为节点2和节点3局部坐标系统到节点1局部坐标系的姿态转换矩阵；和

分别为节点1分别到节点2和节点3的姿态转换矩阵。 in,

and

Respectively, the updated local navigation state of node 1 and its mean square error matrix;

and

are the local KF estimated value of node 1 and its mean square error matrix;

and

are the local KF estimated value of node 1 and its mean square error matrix;

and

Respectively, the local KF estimation value of node 2 and its mean square error matrix;

and are the local KF estimated value of node 3 and its mean square error matrix;

and

are the attitude transformation matrices from the local coordinate system of node 2 and node 3 to the local coordinate system of node 1; and

are the attitude transformation matrices from node 1 to node 2 and node 3 respectively.

When a local KF is detected to be faulty in the system-level integrity monitoring process, the information of the local KF is eliminated in the local navigation state update equation, thereby ensuring the integrity of the final local information fusion filter and improving the overall Integrity of the navigation system.

Claims (3)

1. navigational system completeness monitoring method based on distributed sensor networks, it is characterized in that: adopt the integrity monitoring processing (11) of sensor-level and system-level integrity monitoring to process the classification processing mode of (22), the navigational system based on distributed sensor networks is carried out integrity monitoring; Wherein, navigational system based on distributed sensor networks comprises GNSS receiver, a k SRIMU network node, k is natural number, each network node has identical performance or different performances, the information of all sharing other network node in navigation is processed is carried out information fusion, one of them SRIMU network node also with the information fusion of GNSS receiver, have higher navigation performance, as host node; The treatment step of integrity monitoring and navigation calculation is as follows:

1) in the sensor-level integrity monitoring stage, adopt the RAIM method to carry out integrity monitoring to the GNSS receiver, the FDI processing unit that the metrical information of k SRIMU network node is sent to respectively k SRIMU network node carries out resultant fault detection and the isolation processing based on MW-PV method and wavelet analysis method;

2) through the inertia information after the FDI processing unit processes of k SRIMU network node, be input to respectively in k inertia measurement integrated unit, inertia information through the SRIMU of sensor-level integrity monitoring is carried out fusion treatment, obtain the calculating inertia information with respect to three axle orthogonal coordinate systems;

3) with the calculating inertia information after k inertia measurement fusion treatment, in k local KF of input, carry out Local Navigation information and resolve; Wherein, each local KF receives the inertia measurement fuse information that all are shared; In the local KF of host node, also merge through the GNSS receiver information after the RAIM monitoring, have higher performance than the navigation calculation of other wave filter;

4) the new breath with k local KF is input in system-level integrity monitoring processing unit (22), employing is based on the completeness monitoring method of residual test and the new breath moving window method of the new breath of filtering, carry out the system-level integrity monitoring of navigational system, and integrity information is sent in k Local Navigation state updating unit;

5) last, k Local Navigation state updating unit, the navigational state information of the same type of k local KF of reception is carried out fusion treatment, obtains the navigation information of final renewal; In this k Local Navigation state updating unit, the integrity information that provides is provided according to system-level integrity monitoring, if there is fault in certain local KF, reject the navigational state information of this part KF in fusion treatment.

2. the navigational system completeness monitoring method based on distributed sensor networks according to claim 1, it is characterized in that, described sensor-level integrity monitoring is processed in (11), supposes that its a SRIMU network node is made of the redundant configuration of n inertial sensor by angle mount, wherein, a=1,2 ... k, n is natural number, and n＞3, and the integrity monitoring treatment step of SRIMU comprises:

2-1) the n of an angle mount redundancy Inertial Measurement Unit sensor information, at first send in the fault detect processing unit based on the MW-PV method, set up observation equation, calculate successively parity vector, detection statistic, detection threshold, fault to n sensor detects, and effectively detects the step fault; Then adopt moving window to process, effectively detect the slope fault; When fault being detected, further metrical information is sent to based in the resultant fault of MW-PV method and wavelet analysis method diagnosis isolation processing unit, process according to parity vector and moving window, diagnose out respectively step fault or slope fault, fault-signal is isolated, and reject fault sensor information, construction observation equation again in observation equation;

2-2) when the 1st fault occurs, according to step 2-1), calculate n-3 dimension parity vector by observation equation, can effectively detect the step fault according to this n-3 dimension parity vector, further adopt moving window to process, effectively detect the slope fault; When fault being detected, further process according to this n-3 dimension parity vector and moving window, diagnose out respectively step fault or slope fault, fault-signal is isolated, reject fault sensor information, construction observation equation again in observation equation;

2-3) when the 2nd fault occurs, calculate n-4 dimension parity vector by observation equation, with step 2-2) similar, directly tie up parity vector detection and isolation step fault by n-4, process further detection and isolate the slope fault by moving window;

2-3) when the 3rd fault, and follow-up fault adopts similar method to carry out fault detect and isolation when occurring successively;

2-4) when n-3 fault occurs, the parity vector of this moment only has 1 dimension, when fault being detected, in the fault diagnosis isolation processing, adopts the wavelet transform method come tracing trouble and isolate.

3. the navigational system completeness monitoring method based on distributed sensor networks according to claim 1, it is characterized in that, described system-level integrity monitoring is processed in (22), process based on the new breath of each local KF wave filter of distributed navigation system and carry out integrity monitoring, treatment step comprises:

3-1) will calculate inertia measurement information through the k batch total that the inertia measurement fusion treatment obtains, be input in k local KF and carry out the local message fusion treatment, adopt the mode of differential filtering, observation information in each local KF is the difference processing that all k batch totals are calculated inertia measurement information, has also merged GNSS pseudorange information in the observed quantity of the local KF of described host node;

3-2) in k local KF resolves process, their new breath is sent in the integrity monitoring unit of processing based on new breath; At first carry out the step fault detect according to the residual test of the new breath of filtering; If the fault of not detecting further adopts new breath moving window method to carry out the slope fault detect; When two kinds of integrity detections are all passed through, think that the result of this part KF is believable, otherwise show that this part KF breaks down;

The integrity information that 3-3) will obtain based on the integrity monitoring that new breath is processed, be input in the Local Navigation state updating unit based on local message fused filtering device, further merge the Local Navigation information of the same type of k local KF, improve the navigation performance of each network node; When the new breath of certain local KF does not satisfy the integrity requirement, in local message fused filtering device, it is isolated.

Images