Vehicle-mounted navigation system fault monitoring method

Abstract

The invention provides a vehicle-mounted navigation system fault monitoring method, which comprises the following steps of: 1, receiving a signal by using a GNSS receiver and an IMU device in a vehicle-mounted navigation system, positioning by adopting a Kalman filtering method, and obtaining measurement domain information at the same time; step 2, constructing state domain information and a covariance matrix thereof according to the Kalman filtering parameters and the measurement domain information; step 3, separating and judging GNSS faults from the state domain information and the covariance matrix thereof by using a projection operator; step 4, positioning the judged GNSS fault by adopting an orthogonal matching pursuit method and removing the fault, if all satellites are removed, warning that the GNSS is unavailable, otherwise, executing step 5; step 5, according to the state domain information and the covariance matrix of the state domain information after the GNSS fault is eliminated, separating and judging an IMU fault by using a projection operator; and step 6, correcting the IMU fault, and completing fault monitoring of the vehicle-mounted navigation system.

Description

Technical Field

[0001] This invention relates to a fault monitoring method, specifically a fault monitoring method for an in-vehicle navigation system. Background Technology

[0002] This section provides only background information relevant to this disclosure and is not necessarily prior art.

[0003] With the rapid development of intelligent connected vehicles and autonomous driving technologies, higher demands are placed on the reliability of PNT (Progressive Navigation Target) information in vehicle navigation systems. However, in scenarios such as urban canyons, mountain tunnels, and forest roads, GNSS signals often become intermittent or distorted due to obstruction, multipath effects, and non-line-of-sight (NLOS) effects, affecting the quantity and quality of GNSS observations and leading to a decrease in the vehicle's positioning accuracy, continuity, and integrity. To compensate for the contradiction between high accuracy in short-term and long-term continuity in single GNSS positioning, tightly coupled GNSS / IMU integrated navigation systems have been widely used due to the reduction in cost and improvement in accuracy of inertial measurement units (IMUs). This architecture, within the Kalman filter (KF) framework, fuses GNSS pseudorange observations with IMU outputs. On the one hand, it achieves centimeter-level accuracy when GNSS signals are available; on the other hand, it maintains navigation continuity through inertial extrapolation when GNSS signals are weakened or lost. However, in complex environments such as urban roads, multiple satellites often experience simultaneous step or gradual failures; simultaneously, small biases or historical undetected faults of the IMU accumulate in the filtering and prediction stage, thus affecting subsequent state estimation. Existing fault detection methods based on the measurement domain of AIME typically assume that the visible satellites within the visible sliding window remain unchanged, resulting in insufficient adaptability when the number of satellites dynamically changes. Furthermore, traditional deconstruction-based fault elimination methods often target single-satellite faults, performing poorly in multi-fault scenarios. In addition, traditional subset enumeration-based fault combination detection and sliding window re-filtering strategies introduce exponential online computational overhead, making it difficult to meet the dual requirements of real-time performance and high reliability for vehicles.

[0004] In summary, the existing technology has the following drawbacks:

[0005] a) Satellite number assumption is fixed. Traditional AIME fault detection algorithms require the number of satellites observed in each epoch within a sliding window to remain constant, or to forcibly remove redundant pseudoranges in the measurement domain. During vehicle operation, the number of visible satellites frequently changes due to obstructions such as sunroof vibration and building obstruction, leading to the discarding of effective observations and severely reducing the applicability and continuity of fault detection.

[0006] b) Limitations of the single-fault model. Traditional separation methods are generally designed only for step errors of a single satellite. They cannot provide accurate separation and positioning solutions for situations where multiple satellites may have step or gradual faults simultaneously, making it difficult to meet the high safety requirements of vehicle navigation.

[0007] c) The impact of IMU bias on fault detection is not considered. Existing methods mostly focus on fault detection of GNSS observations, while ignoring IMU prediction bias. This means that when the IMU itself drifts or small faults in the early stages are not eliminated in time, the overall state estimation error will continue to accumulate, eventually affecting the positioning accuracy of the entire filter.

[0008] d) High computational complexity. Combinatorial enumeration of multiple fault subsets, sliding window re-filtering, or multiple batch processing lead to an exponential increase in computational load, making it impossible to respond quickly in real-time vehicle systems. Especially in scenarios with extremely high real-time and reliability requirements, such as urban expressways and highways, these algorithms struggle to meet the demands of continuous online monitoring.

[0009] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0010] Purpose of the invention: The technical problem to be solved by the present invention is to provide a fault monitoring method for vehicle navigation systems, addressing the shortcomings of the existing technology.

[0011] To address the aforementioned technical problems, this invention discloses a fault monitoring method for an in-vehicle navigation system, comprising the following steps:

[0012] Step 1: Use the GNSS receiver and IMU device in the vehicle navigation system to receive signals, use the Kalman filter method for positioning, and obtain measurement domain information at the same time;

[0013] Step 2: Construct the state domain innovation and its covariance matrix based on the Kalman filter parameters and the measurement domain innovation;

[0014] Step 3: Use projection operators to separate and determine GNSS faults from the state domain innovation and its covariance matrix;

[0015] Step 4: Use the orthogonal matching pursuit method to locate and eliminate the identified GNSS fault. If all satellites are eliminated, an alarm will be issued that GNSS is unavailable; otherwise, proceed to step 5.

[0016] Step 5: Based on the state domain information and its covariance matrix after excluding GNSS faults, use the projection operator to separate and determine IMU faults;

[0017] Step 6: Correct the IMU fault and complete the vehicle navigation system fault monitoring.

[0018] Beneficial effects:

[0019] 1. This invention maps the fault detection statistics of the GNSS / IMU integrated navigation system from the measurement domain to the filter state domain. By constructing the state domain information vector and its covariance matrix, fault monitoring can be achieved in scenarios where the number of visible satellites changes dynamically.

[0020] 2. Within the state domain framework, this invention employs a sparse recovery algorithm that combines orthogonal matching pursuit (OMP) and successive projection culling (MP) to locate step or gradual faults in multiple satellites at once, avoiding the exponential computational overhead caused by combined enumeration.

[0021] 3. This invention designs a two-stage projection operator to first eliminate GNSS pseudorange observation faults, and then separate and compensate for IMU prediction biases, thereby realizing independent detection and online correction of GNSS observation faults and IMU output biases after mechanical arrangement.

[0022] 4. This invention transforms multiple fault elimination, fault monitoring and deviation correction into incremental matrix operations and a small number of iterative steps, which significantly improves online operating efficiency and meets the strict requirements of vehicle navigation systems for real-time performance and continuity. Attached Figure Description

[0023] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0024] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0025] Figure 2 This is a schematic diagram of the position, velocity, and attitude data output by the IMU after mechanical orchestration in one embodiment.

[0026] Figure 3 This is a schematic diagram of a measurement matrix in one embodiment.

[0027] Figure 4 This is a schematic diagram of the state domain information vector in one embodiment.

[0028] Figure 5 This is a schematic diagram of an information vector in a state domain that contains only GNSS components, as shown in one embodiment.

[0029] Figure 6 This is a schematic diagram of GNSS fault detection results in one embodiment.

[0030] Figure 7 This is a schematic diagram illustrating the satellite's contribution to the residuals in one embodiment.

[0031] Figure 8 This is a schematic diagram of an information vector in a state domain that contains only IMU components, as shown in one embodiment.

[0032] Figure 9 This is a schematic diagram of IMU fault detection results in one embodiment. Detailed Implementation

[0033] This invention proposes a fault monitoring method for vehicle navigation systems, the overall idea of ​​which is as follows:

[0034] a) To address the problem that existing AIME algorithms require the number of observed satellites to be exactly the same at each epoch within the sliding window, which leads to the discarding of additional pseudorange observations caused by changes in the road environment during vehicle operation, thus reducing the continuity and sensitivity of fault monitoring, the fault monitoring method for vehicle navigation systems proposed in this invention performs fault monitoring in the state domain, without requiring the number of satellites at each epoch to remain constant, and can fully utilize the observations of all visible satellites within the sliding window.

[0035] b) In view of the fact that traditional deseparation methods are mostly based on the assumption of "single satellite failure" and that it is difficult to accurately separate and locate multiple satellites when multiple satellites experience step or gradual failures at the same time, this invention designs a multi-fault satellite detection strategy that integrates orthogonal matching pursuit (OMP) and successive projection elimination (MP), which enables accurate detection of multiple satellite faults and one-time location.

[0036] c) Addressing the problem that existing methods focus on GNSS observation fault detection while neglecting IMU prediction bias, leading to the long-term accumulation of small IMU drift, this invention constructs a unified state-domain projection framework. It first monitors and eliminates GNSS observation faults, then monitors and compensates for IMU biases, achieving independent separation and online compensation for the two types of faults to ensure the long-term stability and accuracy continuity of the integrated navigation system.

[0037] d) To address the issue that subset enumeration or sliding window re-filtering for multiple faults results in an exponential increase in computational complexity, which is unsuitable for the stringent real-time requirements of vehicle systems, this invention proposes a projection operator and a simplified iterative process, enabling multiple fault elimination, integrity detection, and deviation correction to all be achieved in O(n log n) time. sat n d The implementation of matrix multiplication and sparse iteration on the order of magnitude significantly improves online running efficiency.

[0038] The technical solution proposed in this invention is as follows: Figure 1 As shown, the details are as follows:

[0039] The sensors selected for the vehicle navigation system of this invention are a GNSS receiver (capable of receiving BeiDou and GPS signals) and an IMU.

[0040] Step (1) GNSS / IMU compact combination solution

[0041] 1) First, construct the state equations for the GNSS / IMU compact combination, with the 25-dimensional system error state X input to the Kalman filter as follows:

[0042] X=[δr IMU T δv IMU T δφ IMU T b g T b a T s g T s a T t b δt b ] T

[0043] In the formula, δr IMU δv IMU and δφ IMU These are the IMU's three-dimensional position error vector, velocity error vector, and attitude error vector, respectively; b g b a s g and s a These are the three-axis zero bias of the gyroscope, the three-axis zero bias of the accelerometer, the three-axis scaling factor of the gyroscope, and the three-axis scaling factor error of the accelerometer, respectively; t b and δt b These are the clock bias and clock drift of a GNSS receiver, respectively.

[0044] 2) Secondly, construct the measurement equations for the filter. Using the receiver position obtained after mechanical arrangement of the IMU and the satellite position determined by the satellite ephemeris, the distance from the receiver to the satellite is retrieved. The measurement vector is composed of the difference between the retrieved pseudorange prediction value and the GNSS pseudorange measurement value:

[0045] Z = δz ρ =ρ G -ρ I

[0046] In the formula, Z represents the measurement matrix, and δz ρ ρ represents the pseudo-distance difference. I ρ is the pseudorange prediction value retrieved by the IMU inversion. G These are pseudorange measurements from GNSS.

[0047] The dynamic equations of a discrete linear Kalman system, formed by the state equations and the measurement equations, are as follows:

[0048]

[0049] Among them, X k Z represents the state vector at time k, where k represents the k-th epoch. k Let φ represent the measurement matrix at time k. k,k-1 W represents the system state transition matrix from time k-1 to time k. k-1 Γ represents the system noise matrix. k,k-1 H represents the system noise driving matrix. k V represents the observation matrix. k This represents the measurement noise matrix. Kalman filtering generally includes the prediction process:

[0050]

[0051] ∑ k,k-1 =Φ k,k-1 ∑ k-1,k-1 Φ k,k-1 T +Γ k,k-1 Q k-1 Γ k,k-1 T

[0052] Measurement update process:

[0053]

[0054]

[0055] ∑ k,k =[IK k H k ]∑ k,k-1 [IK k H k ] T +K k R k K k T

[0056] in, This represents the posterior state estimate at time k-1. Represents the posterior state estimate at time k-1. Estimate the prior state estimate at time k, ∑ k,k-1 Q represents the covariance of the prediction error estimate from time k-1 to time k. k-1 K represents the system noise variance at time k-1. k Let ∑ denote the Kalman filter gain matrix. k,k R represents the error estimation covariance matrix at time k. k This represents the variance of the measurement noise at time k.

[0057] Step (2) Constructing the state domain information

[0058] 1) Computational measurement domain information

[0059] According to Kalman filtering, the innovation r can be obtained. k The expression:

[0060]

[0061] Among them, new information r k Z represents k Includes pseudorange error observations and This represents the difference between the predicted pseudorange error values.

[0062] When the Kalman filter is a closed-loop corrected filter, the prior state variables are... If the value is zero, then the expression for the new information becomes:

[0063] r k =Z k =ρ G -ρ I

[0064] 2) Constructing the state-domain information and covariance matrix

[0065] Construct the state domain innovation vector d k as follows:

[0066]

[0067]

[0068] In the formula, d k For prior state quantities With posterior state quantity The difference; d k The covariance matrix is ​​P d,k , For d k The expected value, E[] is the expected value of d. k Seeking expectations.

[0069] P d,k =P′ x,k -P x,k =K k P′ z,k (K k ) T

[0070]

[0071] According to the error propagation law and for ease of calculation, P d,kLet P' be the covariance matrix of the prior state variables. x,k With posterior state variable P x,k The difference of covariance matrices, where P′ z,k Let be the prior covariance matrix of the observed vectors.

[0072] Step (3) Isolate GNSS faults

[0073] 1) GNSS and IMU Fault Analysis

[0074] According to the filter's measurement equation, the measured value Z is the difference between the GNSS observed pseudorange value and the IMU inverted pseudorange value, where:

[0075] ρ G =ρ R +ε G

[0076] ρ I =ρ R +ε I =ρ R +H·ε y

[0077] In the formula, ρ R ε represents the true value of the pseudorange. G For GNSS pseudorange measurement noise; ε I The pseudorange inversion noise output by the IMU after mechanical orchestration; ε y H represents the noise in the navigation solution output by the IMU after mechanical orchestration; H is the measurement matrix. Based on these two equations, the state-domain innovation vector can be expressed as:

[0078]

[0079] Because vehicle navigation systems may be affected by deviations in the positioning solution output by the IMU after mechanical orchestration in the previous epoch, as well as deviations in GNSS and IMU measurements in the current epoch, fault monitoring of the navigation system requires separate monitoring of faults to facilitate alarm or fault repair by the monitoring system. Therefore, considering faults, ρ G With ρ I It can be modeled as:

[0080] ρ G =ρ R +ε G +b G

[0081] ρ i =ρ R +H·(ε y +b y )

[0082] b Gb is the GNSS fault vector; y Let d be the deviation vector of the navigation solution output by the IMU after mechanical orchestration. Therefore, the state-domain innovation vector d k and the inverse of its covariance matrix It can be represented as:

[0083] d k =K k r k =K k (ε G +b G )-K k H(ε y +b y )

[0084]

[0085] 2) Projection constructs GNSS fault components

[0086] To isolate faults in GNSS measurements, K is constructed. k The 25-dimensional projection matrix π of H G , so that π G T K k H = 0.

[0087]

[0088]

[0089] Let A represent an identity matrix of dimension 25, and let A represent the gain-observation sensitivity matrix. This indicates that the matrix dimension is n. d ×n y At this point, the innovation vector in the state domain containing only GNSS components is d. G,k :

[0090] d G,k =π G T K k r k =π G T K k (ε G +b G )

[0091] According to the law of error propagation, d G,k covariance matrix P G,k for:

[0092] P G,k =π G T·K k P′ z,k K k T ·π G

[0093] 3) Design an AIME method for detecting GNSS faults in the state domain.

[0094] Based on the traditional Autonomous Integrity Extrapolation (AIME, see: Jiang Yingying, Pan Shuguo, Ye Fei, et al. A Slowly Varying Fault Detection Method Based on Robust Estimation and Improved AIME [J]. Systems Engineering and Electronics, 2022, 44(09):2894-2902.), firstly, batch processing is used to determine the average innovation vector r. avg As shown below:

[0095]

[0096] in, V represents the inverse of the average innovation covariance matrix. i -1 Let r represent the inverse of the information covariance matrix at time i. i Let m represent the innovation vector at time i, and m represent the sliding window length. It can be calculated using the following formula:

[0097]

[0098] Secondly, according to the average innovation vector r avg inverse of the average innovation covariance matrix To construct the test statistic s avg s avg The calculation formula is as follows:

[0099]

[0100] If no fault occurs, s avg It follows a chi-square distribution with degrees of freedom equal to the dimension of the measurement domain. Therefore, based on the traditional AIME algorithm, the GNSS test statistic S can be constructed similarly. G,avg,k :

[0101] S G,avg,k =(d G,avg,k ) T (P G,avg,k ) -1 d G,avg,k ,S G,avg,k ~p(S G,avg,k )

[0102]

[0103] dG,avg,k The average innovation vector that contains only GNSS components in the state domain. S is the inverse of the average innovation covariance matrix in the state domain containing only GNSS components. If no fault occurs, S G,avg,k The degrees of freedom are equal to the dimension (N) of the state domain. d -3-N const The chi-square distribution of (3 location parameters and the number of satellite constellations N) const ), d G,k-m+i P represents the innovation vector in the state domain at time k-m+i that contains only GNSS components. G,k-m+i Let represent the inverse of the innovation covariance matrix containing only GNSS components in the state domain at time k-m+i. Therefore, according to the false alarm rate P FA Calculate the corresponding threshold And S G,avg,k In comparison:

[0104]

[0105] In the formula, F(·) is the cumulative distribution function of the chi-square distribution; P FA For a predetermined false alarm rate, inf{·} represents taking the minimum threshold value. If If the GNSS pseudorange observation is normal, then no fault has occurred; otherwise, a fault has occurred.

[0106] Step (4) Detect and troubleshoot GNSS faults

[0107] If a GNSS pseudorange observation malfunctions, it indicates a fault in the GNSS observation, requiring the separation of the faulty satellite. Based on the aforementioned state domain framework, the Orthogonal Matching Pursuit (OMP) algorithm can be introduced to locate one or more faulty satellites in a single operation.

[0108] 1) Construct the state domain dictionary. Based on the current filter gain K... k and projection matrix π G Calculate the mapping vector of the l-th satellite in the state domain.

[0109]

[0110] And Concatenate the columns to form a dictionary matrix Φ:

[0111]

[0112] At this point, if only the l-th satellite malfunctions, the introduced state deviation is expressed as: b G,l This represents the deviation when the l-th satellite malfunctions.

[0113] 2) Initialization:

[0114]

[0115] In the formula, Let represent the residual vector of the t-th iteration, with the same dimension as the number of rows in the innovation vector; Let represent the set of faulty satellite indices identified in the t-th iteration; t represents the iteration number.

[0116] 3) Contribution calculation. For all The satellite, with weighted Mahalanobis inner product c l Measuring the contribution of the l-th satellite to the residual:

[0117]

[0118] 4) Selection of the most suspicious satellites. Find the index of the satellite with the greatest contribution and update the set of faulty satellites:

[0119]

[0120] Where, j (t) This represents the faulty satellite selected in the t-th iteration.

[0121] 5) Residual Update. Calculate the satellite projection coefficients to be removed and remove the corresponding residual terms:

[0122]

[0123] In the formula, β (t) The projection coefficients of the most suspicious satellite at present are shown on the residuals.

[0124] 6) Fault detection and termination conditions. Based on the fault satellite index set Λ (t) Eliminating satellites in r k 、P′ z,k Recalculate the corresponding rows and columns.

[0125]

[0126] In the formula, This represents the state-domain information vector after removing the row corresponding to the faulty satellite at time k and in the t-th iteration. This represents the information vector in the state domain containing only GNSS components after removing the row corresponding to the faulty satellite at time k and in the t-th iteration. Let represent the covariance matrix of the innovation vectors containing only GNSS components in the state domain after removing the rows corresponding to faulty satellites at time k and in the t-th iteration. Then, recalculate the GNSS test statistic at time k and in the t-th iteration.

[0127]

[0128] In the formula, This represents the information vector in the state domain containing only GNSS components after removing the row corresponding to the faulty satellite at time k-m+i and in the t-th iteration. Let represent the covariance matrix of the innovation vectors in the state domain that contains only GNSS components after removing the row corresponding to the faulty satellite at time k-m+i and in the t-th iteration. This represents the average innovation vector in the state domain containing only GNSS components after removing the row corresponding to the faulty satellite at time k and in the t-th iteration. It represents the inverse of the covariance matrix of the average innovation vector containing only GNSS components in the state domain after the row corresponding to the faulty satellite is removed at time k and in the t-th iteration.

[0129] If the test statistic If the number of satellites is less than or equal to the threshold, the elimination process stops; if the number of satellites is greater than the threshold and the remaining number is 0, an alarm is issued and GNSS information is unavailable; if the number of satellites is greater than the threshold and the remaining number is greater than 0, then t←t+1 is set, and the next most suspicious satellite is selected, and steps 3)-6 are repeated.

[0130] Step (5) Isolate IMU fault

[0131] To isolate the faults of the IMU, a weighted pseudo-inverse K is constructed. + and the state domain projection matrix π F :

[0132]

[0133] The innovation vector d of the IMU component at time k after removing the GNSS component is obtained. F,k and its covariance matrix P F,k :

[0134] d F,k =π F d k

[0135]

[0136] Next, the IMU test statistic s is constructed. F,k :

[0137]

[0138] Wherein, the degree of freedom v equals P F,k The rank. Based on the false alarm rate P. FA Calculate the corresponding threshold and s F,kIn comparison. If If the IMU is functioning correctly, then it is not malfunctioning; otherwise, it is malfunctioning.

[0139] Step (6) Correcting IMU fault deviations

[0140] At this point, since the GNSS fault has been resolved, the fault originates solely from the IMU. Therefore, the remaining residuals can be used for bias estimation.

[0141]

[0142] This represents the IMU bias predicted at time k, and the navigation solution output by the IMU after mechanical orchestration can be updated online.

[0143]

[0144] A more accurate positioning result was obtained after repairing the IMU and GNSS faults.

[0145] Example:

[0146] The sensors selected for the vehicle navigation system of this invention are a GNSS receiver (capable of receiving BeiDou and GPS signals) and an IMU. The following practical example illustrates the above process:

[0147] (1) GNSS / IMU compact combination solution

[0148] 1) First, the specific force and angular velocity information output by the IMU are converted into position, velocity, and attitude through mechanical programming, such as... Figure 2 As shown.

[0149] Based on the Kalman filter principle, the state equation is constructed using the position, velocity, and attitude output by the IMU.

[0150] 2) Construct the Kalman filter measurement equation. Calculate the distance from the receiver to the satellite using the receiver position output by the IMU and the satellite position calculated from the ephemeris; this is the pseudorange reference value retrieved by the IMU. Then, subtract the pseudorange reference value from the GNSS-observed pseudorange value after corrections for ionospheric delay, tropospheric delay, and receiver clock error. This difference yields the pseudorange error value required for the measurement matrix. The measurement matrix Z is as follows: Figure 3 As shown.

[0151] (2) Calculate state domain information

[0152] 1) Computational measurement domain information

[0153]

[0154] Calculate the measurement domain information based on the above formula.

[0155] 2) Constructing the state-domain information and covariance matrix

[0156] Construct the state domain innovation vector d k as follows:

[0157]

[0158] P d,k =P′ x,k -P x,k =K k P′ z,k (K k ) T

[0159]

[0160] Where the state domain innovation vector d k like Figure 4 As shown.

[0161] Step (3) Isolate GNSS faults

[0162] 2) Projection constructs GNSS fault components

[0163] To isolate faults in GNSS measurements, K is constructed. k H's 25-dimensional left-zero matrix π G , so that π G T K k H = 0.

[0164]

[0165] At this point, the innovation vector in the state domain containing only GNSS components is d. G,k and its covariance matrix d G,k :

[0166] d G,k =π G T K k r k =π G T k k (ε G +b G )

[0167] P G,k =π G T ·K k P′ z,k K k T ·π G

[0168] The state domain contains only the innovation vector d of the GNSS components. G,k like Figure 5 As shown.

[0169] 3) Detecting GNSS faults

[0170] Construct the GNSS test statistic S G,avg,k :

[0171] S G,avg,k =(d G,avg,k ) T (P G,avg,k ) -1 d G,avg,k ,S G,avg,k ~p(S G,avg,k )

[0172]

[0173] According to the false alarm rate P FA Calculate the corresponding threshold And S G,avg,k In comparison:

[0174]

[0175] If the sliding window size m is set to 10, the false alarm rate P FA 10 -5 / hr, the threshold can be calculated to be approximately 80.88. At this point, S G,avg,k Compared to 80.88, if S G,avg,k If the value is ≤80.88, then the GNSS pseudorange observation is not faulty; otherwise, a fault has occurred. Figure 6 As shown, it can be seen that GNSS observations malfunctioned at epochs of approximately 450-500, 650, 900, and 1200.

[0176] Step (4) Detect and troubleshoot GNSS faults

[0177] If a GNSS pseudorange observation malfunctions, it indicates a fault in the GNSS observation, requiring the separation of the faulty satellite. Based on the aforementioned state domain framework, the Orthogonal Matching Pursuit (OMP) algorithm can be introduced to locate one or more faulty satellites in a single operation.

[0178] 1) Construct the state domain dictionary. Based on the current filter gain K... k and projection matrix π G Calculate the mapping vector of the l-th satellite in the state domain.

[0179]

[0180] And Concatenate the columns to form a dictionary matrix Φ:

[0181]

[0182] At this point, if only the l-th satellite malfunctions, the introduced state deviation is expressed as: b G,l This represents the deviation when the l-th satellite malfunctions.

[0183] 2) Initialization:

[0184]

[0185] In the formula, Let represent the residual vector of the t-th iteration, with the same dimension as the number of rows in the innovation vector; Let represent the set of faulty satellite indices identified in the t-th iteration; t represents the iteration number.

[0186] 3) Contribution calculation. For all The satellite, with weighted Mahalanobis inner product c l Measure the contribution of the i-th satellite to the residual:

[0187]

[0188] like Figure 7 As shown, there are a total of 20 satellites in this epoch, and the 13th satellite contributed the most to the fault, with a contribution value of 1.9392.

[0189] 4) Selection of the most suspicious satellites. Find the index of the satellite with the greatest contribution and update the set of faulty satellites:

[0190]

[0191] Where, j (t) This represents the faulty satellite selected in the t-th iteration.

[0192] 5) Residual Update. Calculate the satellite projection coefficients to be removed and remove the corresponding residual terms:

[0193]

[0194] In the formula, β (t) The projection coefficients of the most suspicious satellite on the residuals are β in the first iteration of this epoch. (1) The value is 1.3688.

[0195] 6) Fault detection and termination conditions. Based on the fault satellite index set Λ(t) Eliminating satellites in r l 、P′ z,k Recalculate the corresponding rows and columns.

[0196]

[0197] At this point, the GNSS test statistic at time k and the t-th iteration is recalculated. If the test statistic is less than or equal to the threshold, the elimination process stops. If it is greater than the threshold and the number of remaining satellites is 0, an alarm is issued, and the GNSS information is unusable. If it is greater than the threshold and the number of remaining satellites is greater than 0, then t←t+1 is set, and the next most suspicious satellite is selected, and steps 3)-6 are repeated.

[0198] Step (5) Isolate IMU fault

[0199] To isolate the faults of the IMU, a weighted pseudo-inverse K is constructed. + and the state domain projection matrix π F :

[0200]

[0201] The innovation vector d of the IMU component at time k after removing the GNSS component is obtained. F,k and its covariance matrix P F,k :

[0202] d F,k =π d d k

[0203]

[0204] like Figure 8 As shown, d is the innovation vector in the state domain that contains only IMU components. F,k .

[0205] Next, the IMU test statistic s is constructed. F,k :

[0206]

[0207] Wherein, the degree of freedom v equals P F,k The rank. Based on the false alarm rate P. FA Calculate the corresponding threshold and s F,k In comparison. If If the IMU is functioning correctly, then it is not malfunctioning; otherwise, it is malfunctioning. Figure 9 As shown, the IMU malfunctioned in some epochs.

[0208] Step (6) Correcting filter fault deviation

[0209] At this point, since the GNSS fault has been resolved, the fault originates solely from the IMU. Therefore, the remaining residuals can be used for bias estimation.

[0210]

[0211] This represents the IMU bias predicted at time k, and the navigation solution output by the IMU after mechanical orchestration can be updated online.

[0212]

[0213] A more accurate positioning result was obtained after repairing the IMU and GNSS faults.

[0214] In its specific implementation, this application provides a computer storage medium and a corresponding data processing unit. The computer storage medium is capable of storing a computer program, which, when executed by the data processing unit, can run the invention's content regarding a fault monitoring method for an in-vehicle navigation system, as well as some or all of the steps in various embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0215] Those skilled in the art will clearly understand that the technical solutions in the embodiments of the present invention can be implemented using computer programs and their corresponding general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of computer programs, i.e., software products. These computer program software products can be stored in a storage medium and include several instructions to cause a device containing a data processing unit (which may be a personal computer, server, microcontroller, MCU, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0216] This invention provides a concept and method for fault monitoring in vehicle navigation systems. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A method for fault monitoring in a vehicle navigation system, characterized in that, Includes the following steps: Step 1: Use the GNSS receiver and IMU device in the vehicle navigation system to receive signals, use the Kalman filter method for positioning, and obtain measurement domain information at the same time; Step 2: Construct the state domain innovation and its covariance matrix based on the Kalman filter parameters and the measurement domain innovation; Step 3: Use projection operators to separate and determine GNSS faults from the state domain innovation and its covariance matrix; Step 4: Use the orthogonal matching pursuit method to locate and eliminate the identified GNSS fault. If all satellites are eliminated, an alarm will be issued that GNSS is unavailable; otherwise, proceed to step 5. Step 5: Based on the state domain information and its covariance matrix after excluding GNSS faults, use the projection operator to separate and determine IMU faults; Step 6: Correct the IMU fault and complete the vehicle navigation system fault monitoring.

2. The method for fault monitoring of a vehicle navigation system according to claim 1, characterized in that, The localization using the Kalman filter method described in step 1 includes: The state equation and measurement equation of the vehicle navigation system are used together to form the dynamic equation of the discrete linear Kalman system, which is expressed as follows: Among them, X k Z represents the system state vector at time k, where k represents the k-th epoch. k Let φ represent the measurement matrix at time k. k,k-1 W represents the system state transition matrix from time k-1 to time k. k-1 Γ represents the system noise matrix. k,k-1 H represents the system noise driving matrix. k V represents the observation matrix. k Represents the measurement noise matrix; The prediction and measurement update processes of an in-vehicle navigation system are represented as follows: ∑ k,k-1 =Φ k,k-1 ∑ k-1,k-1 F k,k-1 T +C k,k-1 Q k-1 C k,k-1 T ∑ k,k =[I-K k H k ]∑ k,k-1 [I-K k H k ] T +K k R k K k T in, This represents the posterior state estimate at time k. Represents the posterior state estimate at time k-1. Estimate the prior state estimate at time k, ∑ k,k-1 Q represents the covariance of the prediction error estimate from time k-1 to time k. k-1 K represents the system noise variance at time k-1. k Let ∑ denote the Kalman filter gain matrix. k,k R represents the error estimation covariance matrix at time k. k Let k represent the measurement noise variance at time k, and I represent the identity matrix.

3. The method for fault monitoring of a vehicle navigation system according to claim 2, characterized in that, The construction of the state-domain information and its covariance matrix in step 2 includes: Step 2-1, calculate the measurement domain information r k The details are as follows: Among them, new information r k Represents the measurement matrix Z k Includes pseudorange error observations and pseudorange error predictions. difference; When the Kalman filter in the Kalman filtering method is a closed-loop correction, the prior state variable If the value is zero, then the new information is: r k =Z k =ρ G -r I Where, ρ I ρ is the pseudorange prediction value retrieved by the IMU inversion. G The pseudorange measurement value is from GNSS. Step 2-2, construct the state-domain innovation and covariance matrix, as follows: Construct the state domain innovation vector d k , means as follows: Where, d k For prior state quantities With posterior state quantity difference; For d k The expected value, E[] is the expected value of d. k Seeking expectations; State-domain information vector d k covariance matrix P d,k , means as follows: P d,k =P′ x,k -P x,k =K k P′ z,k (K k ) T Among them, P d,k Let P' be the covariance matrix of the prior state variables. x,k With posterior state variable P x,k The difference in covariance matrices, P′ z,k Let be the prior covariance matrix of the observed vectors.

4. The method for fault monitoring of a vehicle navigation system according to claim 3, characterized in that, Step 3, which describes using projection operators to determine GNSS faults from state-domain innovation and its covariance matrix, includes: Step 3-1: Perform GNSS and IMU fault analysis, and convert the state domain innovation vector d... k and the inverse of its covariance matrix It is expressed as follows: d k =K k r k =K k (e G +b G )-K k H(e y +b y ) Among them, b G b is the GNSS fault vector; y ε is the deviation vector of the navigation solution output by the IMU after mechanical orchestration; G For GNSS pseudorange measurement noise; ε y The noise in the navigation solution output by the IMU after mechanical orchestration; Step 3-2: Project and construct GNSS fault components to isolate faults in GNSS measurements; Step 3-3: Design an autonomous integrity extrapolation method to detect GNSS faults in the state domain.

5. A method for fault monitoring of a vehicle navigation system according to claim 4, characterized in that, The projection construction of GNSS fault components described in step 3-2 is as follows: Construction parameter K k The 25-dimensional projection matrix π of H G , so that π G T K k H = 0 means the following: Among them, I nd Let A be an identity matrix of dimension 25, and let A be the gain observation sensitivity matrix. This indicates that the matrix dimension is n. d ×b y ; At this point, the state domain contains only the innovation vector D of the GNSS components. G,k , means as follows: D G,k =π G T K k r k =π G T K k (e G +b G ) According to the law of error propagation, d G,k covariance matrix P G,k for: P G,k =π G T ·K k P′ z,k K k T ·p G 。 6. The method for fault monitoring of a vehicle navigation system according to claim 5, characterized in that, The detection of GNSS faults in the state domain as described in step 3-3 is as follows: Step 3-3-1: Calculate the average innovation vector r using the traditional autonomous integrity extrapolation method. avg The details are as follows: in, Denotes the inverse of the average information covariance matrix. Let r represent the inverse of the information covariance matrix at time i. i Let m represent the innovation vector at time i, and m represent the sliding window length. The calculation method is as follows: Step 3-3-2, based on the average innovation vector r avg inverse of the average innovation covariance matrix Construct the test statistic s avg Used for fault detection, specifically as follows: If no fault occurs, the test statistic s avg It conforms to a chi-square distribution with degrees of freedom equal to the dimension of the measurement domain; Step 3-3-3: Construct the GNSS test statistic S based on the traditional autonomous integrity extrapolation method. G,avg,k Used for GNSS fault detection, specifically as follows: S G,avg,k =(d G,avg,k ) T (P G,avg,k ) -1 d G,avg,k ,S G,avg,k ~p(S G,avg,k ) Where, d G,svg,k The average innovation vector that contains only GNSS components in the state domain. It is the inverse of the average innovation covariance matrix that contains only GNSS components in the state domain; If no fault occurs, then the GNSS test statistic S G,avg,k The degrees of freedom are equal to the dimension N of the state domain. d -3-N const The chi-square distribution, where N d d G,avg,k The dimension, N const Indicates the number of satellite constellations; d G,k-m+i P represents the innovation vector in the state domain at time k-m+i that contains only GNSS components. G,k-n+i It represents the inverse of the new information covariance matrix that contains only GNSS components in the state domain at time k-m+i; According to the false alarm rate P FA Calculate the corresponding threshold And S G,avg,k In comparison, if If the GNSS pseudorange observation is normal, it is considered that no fault has occurred; otherwise, it is considered that a fault has occurred.

7. A method for fault monitoring of a vehicle navigation system according to claim 6, characterized in that, In step 3-3-3, the threshold It is expressed as follows: Where F(·) is the cumulative distribution function of the chi-square distribution; P FA For a predetermined false alarm rate, inf{·} represents taking the minimum threshold value.

8. A method for fault monitoring of a vehicle navigation system according to claim 7, characterized in that, Step 4, which describes locating and eliminating the identified GNSS fault using the orthogonal matching pursuit method, includes: Step 4-1: Construct a state domain dictionary to represent the state deviations introduced by the fault, as follows: Calculate the mapping vector of the l-th satellite in the state domain. It is expressed as follows: Where, N sat This represents the total number of observable satellites at the current epoch. This represents the gain matrix corresponding to the l-th satellite at time k. Mapping vector Concatenate the columns to form a dictionary matrix Φ: The state deviation introduced when the l-th satellite malfunctions is represented as: b g,l Wherein represents the deviation when the l-th satellite malfunctions; Step 4-2, calculate the contribution, as follows: For all The satellite, with weighted Mahalanobis inner product c l The contribution of the l-th satellite to the residual is measured as follows: in, Let n represent the residual vector of the (t-1)th iteration, with dimension n. d The number of rows is equal to that of the innovation vector; Let represent the set of faulty satellite indices identified in the t-th iteration; t represents the iteration number, and the residual vector satisfies: Step 4-3: Select the most suspicious satellites and add them to the set of faulty satellites, as follows: Find the index of the satellite with the largest contribution and update the set of faulty satellites, as shown below: Where, j (t) This represents the faulty satellite selected in the t-th iteration; Step 4-4, update the residuals, as follows: The satellite projection coefficients are calculated and the corresponding residual terms are removed, as shown below: Where, β (t) The projection coefficients of the most suspicious satellite at present onto the residuals; Steps 4-5: Fault location and troubleshooting, based on the fault satellite index set Λ (t) Remove satellites from the new information k The prior covariance matrix P′ of the observation vectors z,k The corresponding rows and columns are recalculated as follows: in, This represents the state-domain information vector after removing the row corresponding to the faulty satellite in the t-th iteration at time k. This represents the information vector in the state domain containing only GNSS components after removing the row corresponding to the faulty satellite in the t-th iteration at time k. Let represent the covariance matrix of the information vectors containing only GNSS components in the state domain after the row corresponding to the faulty satellite is removed in the t-th iteration at time k. Recalculate the GNSS test statistic for the t-th iteration at time k. It is expressed as follows: in, This represents the information vector in the state domain containing only GNSS components after the row corresponding to the faulty satellite is removed in the t-th iteration at time k-m+i. Let represent the covariance matrix of the information vectors in the state domain that contains only GNSS components after the row corresponding to the faulty satellite is removed in the t-th iteration at time k-m+i. This represents the average innovation vector in the state domain containing only GNSS components after removing the rows corresponding to faulty satellites in the t-th iteration at time k. It represents the inverse of the covariance matrix of the average innovation vector containing only GNSS components in the state domain after the row corresponding to the faulty satellite is removed in the t-th iteration at time k. If the test statistic If the value is less than or equal to the threshold, the fault location and elimination are complete, and step 5 is executed. If the number of satellites exceeds the threshold and the number of remaining satellites is 0, then the GNSS information is deemed unavailable and an alarm is issued. If the number of satellites exceeds the threshold and the number of remaining satellites is greater than 0, then t←t+1 is set, and the next most suspicious satellite is selected, and step 4-3 is repeated.

9. A method for fault monitoring of a vehicle navigation system according to claim 8, characterized in that, Step 5, which involves using the projection operator to separate and determine IMU faults, includes: Step 5-1, construct the weighted pseudo-inverse K + and the state domain projection matrix π F , means as follows: Step 5-2: Calculate the innovation vector d of the IMU component at time k after removing the GNSS component. F,k and its covariance matrix P F,k , means as follows: d F,k =π F d k Step 5-3, construct the IMU test statistic s F,k , means as follows: Wherein, the degree of freedom v equals P F,k rank, χ 2 (v) represents a chi-square distribution with v degrees of freedom; Step 5-4, based on the false alarm rate P FA Calculate the corresponding threshold and s F,k In comparison, the following steps are used to determine if an IMU is faulty: if If the IMU is functioning correctly, then it is not malfunctioning; otherwise, it is malfunctioning.

10. A method for fault monitoring of a vehicle navigation system according to claim 9, characterized in that, The IMU fault correction described in step 6 includes: Step 6-1, use the remaining residuals to estimate the bias, as shown below: in, This represents the IMU bias predicted at time k; Step 6-2: Update the navigation solution output by the IMU after mechanical orchestration online. It is expressed as follows: in, This indicates the updated exported solution, completing the IMU fault correction.

Images