Real-time integrity monitoring and fault identification method for self-driving automobile by using PPP-RTK

Abstract

The invention discloses an automatic driving automobile real-time integrity monitoring fault identification method using PPP-RTK, and relates to the field of automatic driving automobiles, and specifically, the method is mainly characterized in that fault satellites are filtered through a ground station network so as to identify hardware faults and position-related faults of a user side. The PPP-RTK is a promising positioning technology, and the positioning availability is not influenced while the calculation load in the IM task is reduced by utilizing the technology. The method is implemented by calculating the ratio between matched undifferentiated and uncombined (UDUC) residuals of a user receiver and the most recent reference station observations. The ratio is used to rank the respective observations, where the observation with the largest ratio is most likely to be erroneous. Thus, it is more possible to identify erroneous observations without generating and testing a large number of subsets. Furthermore, exclusion may be attempted for each observation, which preserves availability of the observation, unlike a grouping technique in which exclusion is performed on a satellite.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving vehicles, and in particular to a real-time integrity monitoring fault identification method for autonomous driving vehicles using PPP-RTK. Background Art

[0002] The Global Navigation Satellite System (GNSS) is a radio navigation system that uses Earth orbit satellites for navigation and positioning. Currently, the GNSS systems that have been built around the world include the United States' GPS (Global Positioning System), Russia's GLONASS (Global Navigation Satellite System), China's BDS (Beidou Satellite Navigation System) and the European Union's Galileo (Galileo Satellite Navigation System).

[0003] In GNSS, precise point positioning (PPP) and real-time kinematic positioning (RTK) are two important positioning technologies. PPP is a method to improve positioning accuracy by correcting the error from satellite to ground receiver. Its principle is to collect data from various sensors, including information such as satellite position and speed, as well as observation data from the receiver, and then use noise filtering, data solution and other methods to process the data. The advantages of PPP are high accuracy and wide coverage, but the algorithm performance is low, resulting in slow convergence speed, and the initialization time is more than 30 minutes, which is too time-consuming and difficult to commercialize. RTK adopts the carrier phase difference method. First, the base station and the vehicle observe the satellite carrier signal at the same time. The base station broadcasts the observation value through the radio. After the vehicle receives the signal sent by the base station, it subtracts it from its own observation value to eliminate the error contained in both, so as to accurately obtain the spatial relative position relationship between the vehicle and the base station, and then convert the relative position relationship into the coordinates required by the user. The advantage of RTK is that the correction number is transmitted in real time through the communication link between the base station and the vehicle, which greatly improves the convergence speed. However, the disadvantage is that after the correction number is lost, the accuracy can only be maintained for 30 seconds, and RTK requires a large number of ground base stations and is highly dependent on a single base station, so the coverage range is severely limited. For a long time, accuracy, convergence time, and coverage range could not be achieved at the same time, until the emergence of PPP-RTK technology, which unified the two at the algorithm level, while having the advantages of both and making up for the disadvantages of both. In this technology, the PPP part is responsible for correcting satellite orbit clock errors and phase deviations, etc., to achieve satellite-based centimeter-level high-precision enhancement. Initialization can be completed within 2 minutes, which greatly shortens the time. The RTK part adjusts the sampling frequency according to different base station densities and bandwidths, and avoids the influence of ionospheric and tropospheric delays. After the correction number is lost, the accuracy retention time is increased to 120 seconds. At the same time, the demand for base station density is relatively small, and the dependence on a single base station is extremely low. Summary of the invention

[0004] In order to solve the problem of filtering faulty satellites through a ground station network during vehicle autonomous driving, thereby identifying hardware faults and location-related faults on the user side, the present invention provides a real-time integrity monitoring fault identification method for an autonomous driving vehicle using PPP-RTK.

[0005] Failures can be divided into two main types according to their source. The first type is failures caused by satellite errors. The second type of failure is caused by user receivers, atmospheric correction anomalies, and position-related errors (such as abnormal multipath). This failure only occurs on the user side and cannot be detected by the ground station network. The present invention identifies the second type of failure. PPP-RTK is a promising positioning technology that reduces the computational load in IM tasks without affecting positioning availability. This is achieved by calculating the ratio between the matched undifferenced and uncombined (UDUC) residuals of the user receiver and the observations of the nearest reference station. This ratio is used to rank the individual observations, where the observation with the largest ratio is most likely to be wrong. Therefore, it is more likely to identify erroneous observations without generating and testing a large number of subsets. In addition, unlike grouping techniques that perform exclusion by satellite, exclusion can be attempted for each observation, which preserves the availability of the observation.

[0006] The present invention is implemented by the following technical solution: a method for real-time integrity monitoring fault identification of an autonomous driving vehicle using PPP-RTK, comprising the following steps:

[0007] S1: The ground network receives the signal, specifically:

[0008] The ground network receives signals including the satellite antenna phase center error (PVO, PCO), precision clock and orbit signals, reference station ground truth and earth direction parameter signals. The satellite antenna phase center error is PVO and PCO.

[0009] S2: The ground network processes the signal and then sends it out. Specifically:

[0010] The ground network uses the known positions of each satellite to check for faulty satellites caused by satellite errors, and sends RTCM messages containing a list of faulty satellites and ionospheric and phase deviation corrections to the network user terminal in real time, allowing the user terminal to eliminate the faulty satellite before the positioning process;

[0011] S3: After the faults caused by satellite errors are resolved in S2, this step is used to resolve the faults caused by user receivers, atmospheric correction anomalies, and position-related errors; this fault only occurs on the user side and cannot be detected by the network, so a receiver-independent integrity monitoring protocol is created, as follows:

[0012] S3-1: One method is to use the chi-square test to scan the calculated residuals of each position solution for faults. First, the test statistic is expressed as:

[0013]

[0014] in is the solution for all views, and is a subset k 1…n where n is the number of test subsets; the difference of the solution for each subset is calculated according to the following two equations Compare with statistical threshold:

[0015]

[0016] Where: H fa is the quantile of the normal distribution of the cumulative distribution function, calculated based on the assigned false alarm probability fa and the total number of failure modes considered, is the standard deviation; the residuals of the estimated position solution are then swept to find the fault as follows:

[0017]

[0018] Where: is the residual vector of the position solution; is the covariance matrix of the observations; α is the significance value determined based on the design and existing applications of the IM process, and df is the degree of freedom;

[0019] S3-2: When the user terminal receives the signal from the ground network and one or more erroneous measurements occur, the number of tests required will be huge, and the corresponding amount of calculation will also be huge, which will affect the real-time positioning performance. By calculating the absolute value of the ratio between the residuals of each observation value of the user position solution and the corresponding residuals of the nearest reference station, the absolute value of the ratio between the corresponding residuals of the nearest reference station is set to coincide with the faulty satellite after eliminating it. It is fault-free, no normalization value is used, an elevation angle weighted model is used, and the distance between network workstations is <100km, so the residuals of the standard deviation mobile station and the reference station used for normalization are the same, and the process is expressed as follows:

[0020]

[0021] Where m is the number of satellites observed by the user after the network excludes suspicious satellites; is the user code and phase observation vector; G U is the design matrix, excluding highly correlated observations from the same satellite, and the correlation coefficient is calculated as follows:

[0022]

[0023] Where ρ is the correlation coefficient between the errors of observations A and B; D is a zero column vector with 1 only at the entries of A and B; is the covariance matrix of user residuals; is the covariance matrix of user observations;

[0024] S3-3: After initially excluding suspicious observations, re-estimate the position solution and repeat the chi-square test; if the detection test continues to fail, try to exclude the second most vulnerable observation and its highly correlated observations without reinserting the previously excluded observations into the model; repeat these processes until the test passes and the erroneous observations are identified. Excluding faulty satellites reported by the ground station network and excluding highly correlated observations can greatly reduce the risk of more than one erroneous observation, and this method can be used to exclude single or two simultaneous observations; whether the observations are from the same satellite or from two different satellites, as long as the two observations with the largest ratio are selected, this method can be applied to the case where two simultaneous erroneous observations are suspected.

[0025] Compared with the prior art, the present invention has the following beneficial effects: the method for real-time integrity monitoring fault identification of an autonomous driving vehicle using PPP-RTK provided by the present invention can be used to exclude a single or two simultaneous observations. Regardless of whether the observations are from the same satellite or from two different satellites, as long as the two observations with the largest ratio are selected, the method can be applied to the case where two simultaneous erroneous observations are suspected. Since a statistical selection process is performed on a limited number of suspicious observations, testing of all subsets is avoided, making the identification process faster. In addition, because UDUC PPP-RTK has the ability to calculate the residual of a single observation, and the exclusion is performed on an individual observation basis, not on a satellite basis like traditional grouping techniques, the identification is performed on a single observation, thereby maintaining the availability of information. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a fault identification flow chart of the present invention. DETAILED DESCRIPTION

[0027] The present invention is further described below in conjunction with specific embodiments.

[0028] A real-time integrity monitoring fault identification method for autonomous vehicles using PPP-RTK, the process is as follows Figure 1 As shown, the following steps are included:

[0029] S1: The ground network receives the signal, specifically:

[0030] The ground network receives signals including the satellite antenna phase center error (PVO, PCO), precision clock and orbit signals, reference station ground truth and earth direction parameter signals. The satellite antenna phase center error is PVO and PCO.

[0031] S2: The ground network processes the signal and then sends it out. Specifically:

[0032] The ground network uses the known positions of each satellite to check for faulty satellites caused by satellite errors, and sends RTCM messages containing a list of faulty satellites and ionospheric and phase deviation corrections to the network user terminal in real time, allowing the user terminal to eliminate the faulty satellite before the positioning process;

[0033] S3: After solving the faults caused by satellite errors in S2, this step is used to solve the faults caused by user receivers, atmospheric correction anomalies, and position-related errors; this fault only occurs on the user side and cannot be detected by the network, so a receiver-autonomous integrity monitoring protocol is created, as follows:

[0034] S3-1: One method is to use the chi-square test to scan the calculated residuals of each position solution for faults. First, the test statistic is expressed as:

[0035]

[0036] in is the solution for all views, and is a subset k 1…n where n is the number of test subsets; the difference of the solution for each subset is calculated according to the following two equations Compare with statistical threshold:

[0037]

[0038] Where: H fa is the quantile of the normal distribution of the cumulative distribution function, calculated based on the assigned false alarm probability fa and the total number of failure modes considered, is the standard deviation; the residuals of the estimated position solution are then swept to find the fault as follows:

[0039]

[0040] Where: is the residual vector of the position solution; is the covariance matrix of the observations; α is the significance value determined based on the design and existing applications of the IM process, and df is the degree of freedom;

[0041] S3-2: When the user terminal receives the signal from the ground network and one or more erroneous measurements occur, the number of tests required will be huge, and the corresponding amount of calculation will also be huge, which will affect the real-time positioning performance. By calculating the absolute value of the ratio between the residuals of each observation value of the user position solution and the corresponding residuals of the nearest reference station, the absolute value of the ratio between the corresponding residuals of the nearest reference station is set to coincide with the faulty satellite after eliminating it. It is fault-free, no normalization value is used, an elevation angle weighted model is used, and the distance between network workstations is <100km, so the residuals of the standard deviation mobile station and the reference station used for normalization are the same, and the process is expressed as follows:

[0042]

[0043] Where m is the number of satellites observed by the user after the network excludes suspicious satellites; is the user code and phase observation vector; G U is the design matrix, excluding highly correlated observations from the same satellite, and the correlation coefficient is calculated as follows:

[0044]

[0045] Where ρ is the correlation coefficient between the errors of observations A and B; D is a zero column vector with 1 only at the entries of A and B; is the covariance matrix of user residuals; is the covariance matrix of user observations;

[0046] S3-3: After initially excluding suspicious observations, re-estimate the position solution and repeat the chi-square test; if the detection test continues to fail, try to exclude the second most vulnerable observation and its highly correlated observations without reinserting the previously excluded observations into the model; repeat these processes until the test passes and the erroneous observations are identified. Excluding faulty satellites reported by the ground station network and excluding highly correlated observations can greatly reduce the risk of more than one erroneous observation, and this method can be used to exclude single or two simultaneous observations; whether the observations are from the same satellite or from two different satellites, as long as the two observations with the largest ratio are selected, this method can be applied to the case where two simultaneous erroneous observations are suspected.

[0047] The scope of protection claimed by the present invention is not limited to the above specific embodiments. For those skilled in the art, the present invention may have various deformations and changes. Any modifications, improvements and equivalent substitutions made within the concept and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for real-time integrity monitoring fault identification of an autonomous driving vehicle using PPP-RTK, characterized in that: The steps include: S1: The ground network receives the signal, specifically: The ground network receives signals including the satellite antenna phase center error, precision clock and orbit signals, reference station ground truth and earth direction parameter signals. The satellite antenna phase center error is PVO and PCO. S2: The ground network processes the signal and then sends it out. Specifically: The ground network uses the known positions of each satellite to check for faulty satellites caused by satellite errors, and sends RTCM messages containing a list of faulty satellites and ionospheric and phase deviation corrections to the network user terminal in real time, allowing the user terminal to eliminate the faulty satellite before the positioning process; S3: After the faults caused by satellite errors are resolved in S2, this step is used to resolve the faults caused by user receivers, atmospheric correction anomalies, and position-related errors; this fault only occurs on the user side and cannot be detected by the network, so a receiver-independent integrity monitoring protocol is created, as follows: S3-1: One method is to use the chi-square test to scan the calculated residuals of each position solution for faults. First, the test statistic is expressed as: in is the solution for all views, and is a subset k 1…n The solution of , where n is the number of test subsets; According to the following two equations, the difference of each subset Compare with statistical threshold: Where: H fa is the quantile of the normal distribution of the cumulative distribution function, calculated based on the assigned false alarm probability fa and the total number of failure modes considered, is the standard deviation; the residuals of the estimated position solution are then swept to find the fault as follows: Where: is the residual vector of the position solution; is the covariance matrix of the observations; α is the significance value determined based on the design and existing applications of the IM process, and df is the degree of freedom; S3-2: When the user terminal receives the signal from the ground network and one or more erroneous measurements occur, the processing flow is as follows: by calculating the absolute value of the ratio between the residual of each observation value of the user position solution and the corresponding residual of the observation value of the nearest reference station, the absolute value of the ratio between the corresponding residual of the observation value of the nearest reference station is set to coincide with the elimination of the faulty satellite, which is fault-free, and the normalized value is not used. The elevation angle weighted model is used, and the distance between the network workstations is <100km, so the residuals of the standard deviation mobile station and the reference station used for normalization are the same. The process is expressed as follows: Where m is the number of satellites observed by the user after the network excludes suspicious satellites; is the user code and phase observation vector; G U is the design matrix, excluding highly correlated observations from the same satellite, and the correlation coefficient is calculated as follows: Where ρ is the correlation coefficient between the errors of observations A and B; D is a zero column vector with 1 only at the entries of A and B; is the covariance matrix of user residuals; is the covariance matrix of user observations; S3-3: After initially excluding suspicious observations, re-estimate the location solution and repeat the chi-square test; if the detection test continues to fail, try to exclude the second most vulnerable observation and its highly correlated observations without reinserting the previously excluded observations into the model; repeat these processes until the test passes and the erroneous observations are identified.

Images