gnss anomaly grading and early warning method based on observation domain multi-evidence phase evolution

Abstract

The application discloses a GNSS anomaly grading early warning method based on an observation domain multi-evidence stage evolution, and relates to the technical field of satellite navigation anomaly monitoring and anti-interference; the method comprises the following steps: analyzing acquired GNSS observation data; constructing an observation domain multi-feature set comprising power degradation features, observation consistency features and continuity failure features; performing robust standardization and evidence mapping on each feature to obtain feature evidence strength; constructing adaptive weights determined by observation reliability factors, anomaly stage matching factors, mutation sensitivity factors and redundancy suppression factors; constructing a joint early warning risk score according to the feature evidence strength and the adaptive weights; determining an anomaly event level based on a dynamic grading threshold and outputting early warning information; the application represents the GNSS anomaly stage evolution process from multiple observation levels, reduces the influence of transient noise and single feature fluctuation on the early warning result, and improves the stability and interpretability of the anomaly grading early warning.

Description

Technical Field

[0001] This invention relates to the field of satellite navigation anomaly monitoring and anti-interference technology, and in particular to a GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain. Background Technology

[0002] Global Navigation Satellite Systems (GNSS) are widely used in precise positioning, timing synchronization, intelligent transportation, unmanned systems, aviation and maritime navigation, and safety-critical infrastructure. These applications typically require receivers to continuously output reliable positioning, velocity, and timing results in complex environments. However, GNSS signals reach ground receivers at relatively low power, and their open service signal structure makes them susceptible to external electromagnetic interference, spoofing signals, changes in the propagation environment, and abnormal receiver tracking. Therefore, timely detection and tiered early warning of GNSS observation anomalies are crucial technical aspects for ensuring the reliability and integrity of navigation services.

[0003] In real-world observation environments, GNSS anomalies are often multi-source and phased. Suppressive radio frequency interference typically raises the receiver front-end noise floor, causing a synchronous decrease in the carrier-to-noise ratio of multiple satellites; deceptive interference may disrupt the kinematic consistency between pseudorange and Doppler by dragging pseudorange or constructing pseudo-satellite observations; ionospheric scintillation causes rapid carrier phase disturbances and signal amplitude fluctuations, thus affecting phase tracking stability; obstruction, non-line-of-sight propagation, and local multipath propagation may cause degradation of some satellite observations. As the anomaly process intensifies, observational anomalies may further develop into cycle slips, loss of lock, or observation interruptions. Therefore, GNSS anomalies do not necessarily manifest as a single, instantaneous transgression of observation limits, but rather can exhibit a phased evolutionary process from signal power degradation and disruption of observation consistency to continuity failure.

[0004] Existing GNSS anomaly detection methods typically rely on a single or limited set of indicators. For instance, methods based on carrier-to-noise ratio (CNR) changes can reflect signal power attenuation but struggle to identify deceptive anomalies that don't involve significant power drops. Methods based on pseudorange or positioning residuals can detect measurement anomalies, but their response usually lags behind observation layer disturbances and is susceptible to geometric configuration and positioning model errors. Methods based on carrier phase or loss-of-lock indicators are sensitive to phase continuity changes but struggle to distinguish the cause of anomalies and are also affected by single-satellite cycle slips, low-elevation observations, or localized obstructions. Furthermore, some methods use fixed thresholds for decision-making, which are applicable when the observation background is stable but struggle to adapt to background fluctuations at different stations, under different satellite visibility conditions, and in different electromagnetic environments.

[0005] The aforementioned existing technologies have at least the following shortcomings: First, a single observation feature is insufficient to cover multiple anomaly scenarios, including suppression interference, deceptive interference, ionospheric scintillation, obstruction, and localized observation degradation. Second, existing methods primarily focus on whether anomalies occur, lacking a phased characterization of the progression from early power degradation to observational consistency disruption and then to continuous failure. Third, fixed thresholds or single statistics are easily affected by background noise, changes in the number of satellites, and short-term outliers, leading to false alarms or missed alarms. Fourth, some detection results lack interpretable observational evidence, making it difficult to determine whether the anomalies primarily originate from the power layer, code observation layer, phase observation layer, or continuous failure layer, thus limiting subsequent anti-interference decisions and integrity assessments. Summary of the Invention

[0006] This invention provides a GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain. It comprehensively covers various anomaly scenarios through multi-level evidence, and characterizes the evolution process of anomalies from power degradation to observation consistency disruption and then to continuity failure in stages. This reduces the false alarm rate and missed alarm rate, and improves the stability and interpretability of the early warning results.

[0007] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0008] A GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain includes the following steps:

[0009] Acquire GNSS observation data, analyze and preprocess the GNSS observation data, extract carrier-to-noise ratio, pseudorange, carrier phase, Doppler observation values ​​and loss-of-lock flag, and construct an observation data matrix according to epoch and satellite number;

[0010] Based on the observation data matrix, a multi-feature set of the observation domain is constructed. The multi-feature set of the observation domain includes at least power degradation features, observation consistency features, and continuous failure features.

[0011] Robust standardization and evidence mapping are performed on each feature in the multi-feature set of the observation domain to obtain the feature evidence strength corresponding to each feature.

[0012] An adaptive weight is constructed based on the feature evidence strength, the observation validity of each feature, the abnormal stage attributes, the temporal changes of evidence, and the correlation between features. The adaptive weight is determined by a combination of the observation reliability factor, the abnormal stage matching factor, the mutation sensitivity factor, and the redundancy suppression factor.

[0013] A joint early warning risk score is constructed based on the strength of the feature evidence and adaptive weights.

[0014] Based on the joint early warning risk score, a dynamic hierarchical threshold is constructed, and the early warning level of the abnormal event is determined by combining the event persistence constraint and the event-level evidence judgment, and the early warning information of the abnormal event is output.

[0015] A further improvement to the above technical solution is as follows:

[0016] Preferably, the method for constructing the power degradation feature includes: establishing a carrier-to-noise ratio (CNR) sliding background baseline for each satellite, calculating the degradation amount of the satellite's current CNR relative to the sliding background baseline, and marking the satellite as an affected satellite when the degradation amount is greater than a single-satellite degradation threshold; applying multi-satellite synchronization constraints based on the number of affected satellites and the number of effective satellites; constructing at least a common-mode edge drop sub-quantum based on the positive CNR decrease, a common-mode proportional sub-quantum based on the proportion of affected satellites, and a common-mode platform drop sub-quantum based on the CNR degradation amount of affected satellites, and weighted fusing the three sub-quantums to obtain the power degradation feature.

[0017] Preferably, the observation consistency class feature includes a first observation consistency class feature and a second observation consistency class feature;

[0018] The first observation consistency category feature constructs pseudorange-Doppler consistency residuals based on pseudorange time difference scores and Doppler observations. After median centering, it is constructed based on the residual robust scale and the residual abnormal satellite ratio.

[0019] The second observation consistency feature converts carrier phase observations into metric carrier phase quantities, constructs carrier phase-Doppler consistency residuals based on metric carrier phase time difference scores and Doppler observations, and then constructs them based on the residual robust scale and the residual abnormal satellite ratio after median centering.

[0020] Preferably, the method for constructing the continuous failure class feature includes:

[0021] Cycle slip status is detected based on the adjacent epoch abrupt change in the carrier phase-Doppler consistency residual;

[0022] The single-satellite loss status is determined by combining at least one of the following information: loss of lock flag, carrier phase observation missing status, observation time discontinuity status, continuous weak tracking status, and lock time reset status.

[0023] The cycle slip state and the unlock state are unified into a single-satellite continuous failure flag, and the proportion of continuously failing satellites is calculated within the reference satellite set to obtain the continuous failure class characteristics.

[0024] Preferably, the robust normalization process includes:

[0025] For the Features Calculate the background median and median absolute deviation within a sliding background window, and construct the standardized features according to the following formula:

[0026]

[0027] in, Indicates the first One feature in the epoch robust normalization characteristics Indicates feature index, Indicates the first One feature in the epoch eigenvalues, Indicates the first One feature in the sliding background window The background median, Indicates the epoch index within the window. Represents the epoch Sliding background window, Indicates the first The feature is within the background window. eigenvalues; Indicates the first The median absolute deviation of each feature within the background window; To prevent positive numbers with a denominator of zero, This represents the positive part of the function.

[0028] Preferably, the evidence mapping process includes mapping the standardized features to feature evidence strength, using the following formula:

[0029]

[0030] in, Indicates the first One feature in the epoch The strength of characteristic evidence, Indicates the first One feature in the epoch robust normalization characteristics Indicates the first The mapping sensitivity parameter of each feature, This represents the evidence mapping function.

[0031] Preferably, the method for constructing the adaptive weights includes:

[0032] According to the The ratio of the number of valid satellites participating in the statistics at the current epoch to the number of visible satellites at the current epoch is used to construct the observation reliability factor.

[0033] An anomaly stage index is constructed based on the depth of the anomaly stage corresponding to each feature and the current strength of evidence. And construct an abnormal stage matching factor based on the abnormal stage index;

[0034] Based on the positive increment of the strength of characteristic evidence between adjacent epochs, a mutation sensitivity factor is constructed;

[0035] Based on the correlation between the evidence corresponding to the first observation consistency class feature and the evidence corresponding to the second observation consistency class feature within a short time window, a redundancy suppression factor is constructed.

[0036] The adaptive weights are obtained by multiplying and normalizing the observation reliability factor, anomaly phase matching factor, mutation sensitivity factor, and redundancy suppression factor.

[0037] Preferably, the method for constructing the joint early warning risk score includes:

[0038] Based on the strength of feature evidence and adaptive weights, construct the main risk aggregation term:

[0039]

[0040] in, Represents the epoch The main risk aggregation item, Indicates the first One feature in the epoch Adaptive weights, Indicates the first One feature in the epoch The strength of characteristic evidence, Indicates feature index;

[0041] Synergistic enhancement terms are constructed based on the stage evolution relationship between power degradation, observation consistency violation, and continuity failure.

[0042] The main risk aggregation term and the synergistic enhancement term are combined to obtain an instantaneous joint anomaly score, and the instantaneous joint anomaly score is then subjected to time smoothing to obtain a joint early warning risk score.

[0043] Preferably, determining the warning level of the abnormal event and outputting the warning information of the abnormal event specifically includes:

[0044] Based on the median and median absolute deviation of the joint early warning risk score within the background window, a three-level dynamic classification threshold is constructed.

[0045] The instantaneous warning level is determined based on the comparison between the joint early warning risk score and the three-level dynamic grading threshold.

[0046] The instantaneous warning level is subjected to event persistence constraint processing to obtain risk candidate event segments;

[0047] Based on the event-level evidence thresholds for the strength of each characteristic evidence, candidate segments of characteristic evidence are determined;

[0048] The core segment of the abnormal event is determined based on the risk candidate event segment and the feature evidence candidate segment;

[0049] The core segment of the abnormal event is subjected to event-level quantile judgment to determine the final warning level of the abnormal event;

[0050] Output the start time, end time, duration, warning level, and corresponding feature evidence results of the abnormal event.

[0051] The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain provided by this invention has the following advantages compared with the prior art:

[0052] (1) The GNSS anomaly hierarchical early warning method based on multi-evidence stage evolution in the observation domain of this invention constructs three multi-level evidences in the observation domain: power degradation features, observation consistency features (including pseudorange-Doppler consistency and carrier phase-Doppler consistency), and continuity failure features. These features can simultaneously reflect signal power degradation, deviations in consistency between code / phase observations and Doppler constraints, and interruptions in carrier tracking continuity, significantly broadening the coverage of anomaly detection and reducing the false negative rate. By corresponding the power degradation features, observation consistency features, and continuity failure features to the early, middle, and late stages of anomaly evolution, respectively, and constructing anomaly stage indices and anomaly stage matching factors, the early warning results can reflect the current evolution stage of the anomaly. This mechanism enables this invention not only to detect the existence of anomalies but also to determine the degree of anomaly development, providing more refined information for subsequent anti-interference decisions and integrity assessments.

[0053] (2) The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain of this invention constructs a three-level dynamic classification threshold based on the median and median absolute deviation of the joint early warning risk score within a sliding background window, enabling the threshold to adapt to changes in background noise levels. Simultaneously, robust normalization reduces the impact of outliers on feature scales, and temporal smoothing suppresses instantaneous fluctuations. These mechanisms work together to significantly reduce the false alarm rate and false negative rate of the fixed threshold method in complex environments.

[0054] (3) The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain of this invention introduces multiple mechanisms, including event persistence constraints, event-level evidence thresholds, and event-level quantile decisions, when determining the final early warning level. The event persistence constraint requires that multiple consecutive epochs meet the conditions to form a candidate event, effectively suppressing false alarms caused by single-epoch peaks and short-term fluctuations; the event-level quantile decision determines the level based on the quantiles of multiple epochs within the core segment of the event rather than a single extreme value, avoiding the excessive influence of a single extreme epoch on the event level. These mechanisms together ensure the stability and reliability of the early warning results. Attached Figure Description

[0055] Figure 1 This is a schematic diagram illustrating the construction of the multi-feature set in the observation domain and the evolution of the anomaly stage in this invention.

[0056] Figure 2 This is a flowchart of the joint risk calculation and event-level hierarchical decision-making process in this invention.

[0057] Figure 3 The results are multi-feature time series results of the observation domain based on measured GNSS observation data, where (a) is the common-mode carrier-to-noise ratio anomaly feature, (b) is the pseudorange and Doppler consistency feature, (c) is the carrier phase and Doppler consistency feature, and (d) is the cycle slip or lockout ratio feature.

[0058] Figure 4 This is the result of the joint early warning risk score and dynamic grading threshold determination.

[0059] Figure 5 It is a time series of real weighted contributions of multiple features and statistical results of high contribution epochs. Detailed Implementation

[0060] The following provides a detailed description of specific embodiments of the present invention. It should be understood that the specific embodiments described herein are for illustrative and explanatory purposes only and are not intended to limit the scope of the invention.

[0061] like Figure 1 and Figure 2 As shown, the GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain of the present invention specifically includes the following steps:

[0062] Step S1: Acquisition and preprocessing of GNSS observation data.

[0063] Acquire GNSS observation data, analyze and preprocess the GNSS observation data, extract carrier-to-noise ratio, pseudorange, carrier phase, Doppler observation values ​​and loss-of-lock flags, construct the corresponding observation data matrix according to epoch and satellite number, and determine the corresponding carrier wavelength according to satellite system or signal frequency.

[0064] S1-1, GNSS observation data acquisition. Reads RINEX observation files or receives real-time observation data streams. If it is a RINEX file, it parses the observation file header information to obtain the observation epoch, satellite number, satellite system, observation type, and sampling interval; if it is a real-time data stream, it parses the observation data packets according to the corresponding protocol.

[0065] S1-2, Data Analysis and Preprocessing. Carrier-to-noise ratio, pseudorange, carrier phase, Doppler observations, and loss-of-lock indicators are extracted from the observation data, and corresponding observation data matrices are constructed according to epoch and satellite number. Rows in the matrix correspond to epochs. The corresponding satellite number The matrix elements are the observations, for each epoch. For satellites that are not visible in the table, the corresponding element is filled with an invalid value.

[0066] Record satellite label In the calendar The carrier-to-noise ratio, pseudorange, carrier phase, Doppler observations, and loss-of-lock indicators are as follows: ,in, Used to reflect signal power and tracking quality, Used to reflect the distance information of code observation. Used to reflect carrier phase tracking information, Used to reflect the relative radial velocity change between the satellite and the receiver. Used to reflect the state of loss of lock or abnormal continuous tracking.

[0067] Determine the carrier wavelength corresponding to each satellite based on the satellite system or signal frequency:

[0068] (1)

[0069] in, For satellite The carrier wavelength of the corresponding signal At the speed of light, This corresponds to the signal frequency. The carrier wavelength is used to unify carrier phase observations and Doppler observations to the metric scale, providing a foundation for subsequent consistency feature construction.

[0070] Step S2: Construct a multi-feature set for the observation domain.

[0071] A multi-feature set of the observation domain is constructed based on the observation data matrix. The multi-feature set of the observation domain includes at least power degradation features, observation consistency features, and continuity failure features. Power degradation features are used to reflect the synchronous decline in signal quality of multiple satellites relative to the historical background; observation consistency features are used to reflect the degree of deviation of code observations and phase observations from the Doppler constraint; continuity failure features are used to reflect the degree of failure of the continuous tracking state of carrier phase observations.

[0072] This embodiment constructs a multi-feature set of the observation domain based on the preprocessed observation data matrix. The multi-feature set of the observation domain is organized according to the hierarchical representation of GNSS anomalies in the observation domain.

[0073] The observation domain multi-feature set includes power degradation features, observation consistency features, and continuous failure features. The specific construction method is as follows:

[0074] S2-1, Construction of power degradation class features.

[0075] Power degradation features are used to characterize the synchronous decline in signal quality of multiple satellites relative to the historical background. These features are constructed from carrier-to-noise ratio (CNR) observations. Since the CNR of different satellites is significantly affected by satellite elevation angle, antenna pattern, and propagation environment, this embodiment does not directly use a fixed CNR threshold, but instead establishes a sliding background baseline for each satellite.

[0076] S2-1-1, establish a carrier-to-noise ratio (CNR) sliding background baseline for each satellite and calculate the degradation of the satellite's current CNR relative to the sliding background baseline, using the following formula:

[0077] (2)

[0078] in, Indicates satellite In the calendar The carrier-to-noise ratio degradation Indicates the satellite number, Indicates epoch index, Indicates the epoch index within the window. Represents the epoch The corresponding background window, Indicates satellite epochs within the window The observed carrier-to-noise ratio; This indicates the operation of taking the median. Indicates the window Carrier-to-noise ratio observations for all epochs Take the median as the satellite The background reference value at the current moment; Indicates satellite In the current epoch The observed carrier-to-noise ratio, This represents the positive part of the function.

[0079] when When the degradation threshold for a single satellite is exceeded, the satellite is designated as an affected satellite. The set of affected satellites is as follows:

[0080] (3)

[0081] in, For the calendar The set of affected satellites Indicates the satellite number, Indicates epoch index, For the calendar The set of visible satellites, Indicates satellite In the calendar The carrier-to-noise ratio degradation; This represents the single-satellite degradation threshold, which is a pre-set positive threshold value, set to 2~4 dB based on the standard deviation of noise ratio fluctuation under normal environmental conditions.

[0082] S2-1-2, to avoid misjudgments caused by single-satellite obstruction, low-elevation fading, or occasional tracking fluctuations, this embodiment further introduces multi-satellite synchronization constraints:

[0083] (4)

[0084] in, Represents the epoch The number of affected satellites Represents the epoch The number of effective satellites, Indicates the minimum number of satellites affected. Indicates the minimum affected percentage; The up rounding operation is used to convert the number of affected satellites calculated proportionally into an integer number of satellites; epochs that satisfy the conditions of formula (4) are considered to have multi-satellite synchronous power degradation characteristics.

[0085] In this embodiment, The value is 3. The value is set to 0.25, and can be adaptively adjusted based on the total number of visible satellites and scene requirements: when the number of visible satellites is small (e.g., less than 8), the value can be appropriately reduced. Up to 2; in scenarios with high anti-interference requirements, the value can be appropriately increased. Up to 0.3.

[0086] S2-1-3 characterizes the degree of power degradation from three aspects: calculating the edge abrupt change intensity of multi-satellite synchronous decline in a short period of time, calculating the proportion of affected satellites, and calculating the decline magnitude of the common-mode platform of affected satellites.

[0087] Define satellite The positive decrease in carrier-to-noise ratio between adjacent epochs is:

[0088] (5)

[0089] in, Indicates satellite In the calendar The positive decrease in carrier-to-noise ratio, Indicates the satellite number, Indicates epoch index, Indicates satellite In the previous epoch The observed carrier-to-noise ratio, Indicates satellite In the current epoch The observed carrier-to-noise ratio, This represents the positive part of the function.

[0090] Based on the positive decrease in the carrier-to-noise ratio, a common-mode edge drop subquantity is constructed:

[0091] (6)

[0092] in, Represents the epoch The common-mode edge sudden drop in quantum quantity, Indicates epoch index, Indicates the epoch index within the window. Indicates a short-term statistical window. Indicates the starting epoch of the short-time window. Indicates satellite In the calendar The positive decrease in carrier-to-noise ratio, Indicates the epoch within the window The affected satellite set; Indicates the epoch The affected satellite set Above, the positive decrease in carrier-to-noise ratio for each satellite. Take the median.

[0093] The proportion of affected satellites is constructed as follows:

[0094] (7)

[0095] in, Represents the epoch The effective proportion of affected satellites Represents the epoch The proportion of satellites affected Indicates the baseline threshold for the affected proportion. This represents the positive part of the function.

[0096] The common-mode platform descent sub-quantum is constructed as follows:

[0097] (8)

[0098] in, Represents the epoch The common-mode platform dropout quantity represents the common-mode level of the satellite carrier-to-noise ratio degradation affected in the current epoch; For the calendar The set of affected satellites Indicates satellite In the calendar The carrier-to-noise ratio degradation Indicates the satellite number, Indicates epoch index, Represents a set All satellites Take the median.

[0099] Robust normalization is performed on the three subquantities mentioned above. For any subquantity... Standardize within the sliding background window as follows:

[0100] (9)

[0101] in, Subquant The value after robust normalization Indicates the subquantity in the sliding background window The background median, Indicates the epoch index within the window. Represents the epoch Sliding background window, This indicates that the subquantity is located at a certain historical epoch within the background window. The value at time; Indicates the quantity in the current epoch The value; This indicates the median absolute deviation of the sub-quantity within the background window; To prevent positive numbers with a denominator of zero, This represents the positive part of the function.

[0102] Substituting the common-mode edge drop subquant, the affected satellite proportional subquant, and the common-mode platform drop subquant into formula (9) for robust normalization, we obtain... , and And construct power degradation class features according to the following formula:

[0103] (10)

[0104] in, Represents the epoch Power degradation characteristics The fusion weights represent the common-mode edge drop subquantities. Indicates the common-mode edge drop subquantity The value after robust normalization The fusion weights represent the proportion of affected satellites. Indicates the effective proportion of affected satellites The value after robust normalization This indicates the fusion weight of the common-mode platform's reduced sub-quantity. Indicates the common-mode platform decreasing quantum quantity The value after robust standardization.

[0105] in, , and These are the weights of the common-mode edge drop subquant, the affected satellite proportion subquant, and the common-mode platform drop subquant, respectively, and they satisfy the following:

[0106] (11)

[0107] The normalization constraints of the three weights ensure that the weighted sum is not biased due to different weight scales. It can simultaneously reflect the suddenness, coverage, and degree of sustained power degradation, and is suitable for identifying multi-satellite common-mode degradation phenomena caused by suppression interference, strong obstruction, or receiver front-end impairment.

[0108] S2-2, Construction of observation consistency class features.

[0109] Observation consistency features are used to characterize the kinematic consistency deviations between pseudorange observations, carrier phase observations, and Doppler observations. These features include a first observation consistency feature and a second observation consistency feature. The first feature characterizes the kinematic consistency deviation between pseudorange and Doppler observations, while the second feature characterizes the kinematic consistency deviation between carrier phase and Doppler observations.

[0110] S2-2-1, For the first observation consistency class feature, construct the pseudorange-Doppler consistency residual:

[0111] (12)

[0112] in, Indicates satellite In the calendar The pseudorange-Doppler consistency residuals are used to quantify the degree of deviation between the pseudorange rate of change and the Doppler observations. Indicates satellite In the calendar pseudorange observations, Indicates satellite In the previous epoch pseudorange observations, Indicates satellite The carrier wavelength of the corresponding signal Indicates satellite In the calendar Doppler observations, This indicates the time interval between adjacent epochs.

[0113] Under normal circumstances, the pseudorange time difference should be consistent with the distance rate reflected by Doppler observations. This consistency can be disrupted when code observations are affected by deception, correlation peak distortion, or local propagation anomalies.

[0114] S2-2-2, For the second observation consistency class feature, first convert the carrier phase observations into metric carrier phase quantities:

[0115] (13)

[0116] in, Indicates satellite In the calendar The metric carrier phase quantity. Indicates satellite The carrier wavelength of the corresponding signal Indicates satellite In the calendar The carrier phase observation values.

[0117] S2-2-3, Constructing carrier phase-Doppler consistency residuals:

[0118] (14)

[0119] in, express, Indicates satellite In the calendar The metric carrier phase quantity. Indicates satellite In the previous epoch The metric carrier phase quantity. Indicates satellite The carrier wavelength of the corresponding signal Indicates satellite In the calendar Doppler observations, This indicates the time interval between adjacent epochs.

[0120] The median centering process is applied to the consistency residuals of multiple satellites within the same epoch to eliminate the effects of receiver clock bias and common errors.

[0121] (15)

[0122] in, Indicates satellite In the calendar The centralized consistency residual, Indicates satellite In the calendar The Class-consistent residuals; Indicates feature index, This indicates pseudorange-Doppler consistency (a first observation consistency feature). Indicates carrier phase-Doppler consistency (second observation consistency class feature); Represents a set Consistency residuals of all satellites The median is taken as a robust estimate of the receiver clock bias and common error at that epoch. Represents the epoch China participated in the The effective set of satellites for class-consistent residual calculation. Represents a set China Satellite The Class-consistent residuals Represents a set Satellite numbers within the region.

[0123] Residual robust scaling based on median absolute bias of centered residuals:

[0124] (16)

[0125] in, Represents the epoch No. Robust scaling estimation of class-consistent residuals; Indicates the epoch The following are participants in the first All valid satellites in the class consistency calculation Calculate their centralized consistency residuals respectively. Find the absolute values ​​of the values, and then take the median of these absolute values; Indicates satellite In the calendar The centralized consistency residual, Represents the epoch China participated in the The effective set of satellites for class-consistent residual calculation. This indicates that the median absolute deviation (MAD) is converted to a correction factor consistent with the standard deviation scale. Indicates an index of observation consistency type. This indicates pseudorange-Doppler consistency (a first observation consistency feature). This indicates carrier phase-Doppler consistency (a second observation consistency feature).

[0126] When the single-star centered residual satisfies At that time, the satellite was identified as a satellite with abnormal observation consistency, among which This is the threshold coefficient for consistency anomaly judgment.

[0127] Based on the residual robustness scale and the proportion of residual abnormal satellites, construct the corresponding observation consistency class features:

[0128] (17)

[0129] in, Represents the epoch No. Class-based observation consistency features are used to quantify the degree of consistency deviation of pseudorange or carrier phase observations relative to Doppler constraints; Represents the epoch No. Robust scaling estimation of class-consistent residuals Represents the epoch No. The number of residual abnormal satellites in class consistency Represents the epoch China participated in the The total number of valid satellites for class-consistency calculation.

[0130] S2-3, Construction of features for continuous failure types.

[0131] Continuous failure features are used to characterize cycle slips, loss of lock, or carrier tracking interruption states. Continuous failure features are used to characterize whether an anomaly is associated with a cycle slip, loss of lock, or tracking interruption phase.

[0132] First, cycle slip status is detected based on the adjacent epoch abrupt change in carrier phase-Doppler consistency residual:

[0133] (18)

[0134] in, Indicates satellite In the calendar The number of cycle slip detections, Indicates satellite In the calendar The centered carrier phase-Doppler consistency residual, Indicates satellite In the previous epoch The centered carrier phase-Doppler consistency residual, This indicates an epochal index.

[0135] When the mutation amount of adjacent epochs satisfies the following formula:

[0136] (19)

[0137] in, Indicates satellite In the calendar The cycle slip continuity mutation flag has a value of 1 indicating that a mutation has been detected and a value of 0 indicating that no mutation has been detected. This indicates that the flag is constructed based on carrier phase-Doppler consistency. This indicates an indicator function that takes the value 1 when the condition within the parentheses is true and 0 when it is false. Indicates satellite In the calendar The number of cycle slip detections; This represents the cycle slip detection threshold, which is a pre-set positive threshold value. When the threshold is exceeded, it is determined that a phase continuity abrupt change has occurred; that is, when formula (18) is valid, it is determined that the satellite has a phase continuity abrupt change.

[0138] Furthermore, the single-satellite loss status is determined by combining at least one of the following information: loss of lock flag, carrier phase observation missing status, observation time discontinuity status, continuous weak tracking status, and lock time reset status.

[0139] The loss-of-lock flag is a hardware / software loss-of-lock indication signal output by the receiver's baseband signal processing channel, directly reflecting whether the phase-locked loop (PLL) or delay-locked loop (DLL) is in a lost-lock state. When this flag is set, it indicates that the receiver can no longer stably track the satellite signal, and both carrier phase and pseudorange observations are unavailable. This flag is usually generated in real time by the receiver's internal state machine based on loop errors, phase detector output, etc., and is the most direct indication of tracking interruption.

[0140] A missing carrier phase observation is a state where no carrier phase observation is output at the current epoch, or the observation is invalid. This state typically occurs when the signal is blocked, the signal-to-noise ratio is too low for the receiver to extract reliable carrier phase measurements, or the receiver actively discards unreliable observations during data recording. Unlike the loss-of-lock flag, missing observations primarily reflect gaps in the data layer.

[0141] A discontinuous observation time state occurs when the satellite's observations are not continuous in the time series; that is, there were observations in the previous epoch but not in the current epoch, or observations were missing for several intermediate epochs and then resumed in the current epoch. This state reflects an interruption in the observation data stream, which may be caused by signal blockage, receiver channel scheduling, data recording frame loss, etc. This state alone is insufficient to determine the source of the anomaly, but it can serve as supplementary evidence of a continuity failure.

[0142] Continuous weak tracking occurs when the carrier-to-noise ratio remains below the receiver's tracking threshold for a certain period (e.g., more than 5 consecutive epochs). Although the receiver does not report a loss of lock, the tracking quality has severely degraded, and carrier phase measurement noise has increased significantly. This state reflects a slow deterioration in tracking quality and typically occurs in scenarios where the signal is gradually blocked or interference gradually increases.

[0143] The lock-time reset state occurs when the carrier phase lock-time counter maintained internally by the receiver is reset to zero or significantly shortened. The lock-time counter is reset to zero and restarted when the receiver detects a cycle slip, loss of lock, or reconvergence of the tracking loop. This state is one of the most reliable indicators of whether continuous phase tracking has been interrupted, as it directly reflects the receiver's internal decision regarding the continuity of carrier phase integer ambiguity.

[0144] The cycle slip state and the loss of lock state are unified into a single-star continuous failure flag. The proportion of continuously failing satellites within the reference satellite set is calculated to obtain the characteristics of the continuously failing satellite category, expressed by the following formula:

[0145] (20)

[0146] in, Represents the epoch The continuous failure category characteristic represents the proportion of satellites in the reference satellite set that have experienced continuous failures; Represents the epoch Reference satellite set, Indicates satellite In the calendar The single-star continuity failure indicator.

[0147] The above characteristics can reflect the impact of aggravated anomalies on carrier tracking continuity, providing direct evidence for severe early warning.

[0148] Step S3, Robust Standardization and Evidential Mapping.

[0149] Robust standardization and evidence mapping are performed on each feature in the multi-feature set of the observation domain to obtain the evidence strength of each feature. Robust standardization is based on the median and median absolute deviation within the sliding background window, while evidence mapping is used to unify features with different dimensions and statistical distributions into a fusionable evidence scale.

[0150] For the Features Calculate the background median and median absolute deviation within a sliding background window, and construct standardized features:

[0151] (twenty one)

[0152] in, Indicates the first One feature in the epoch robust normalization characteristics Indicates feature index, Indicates the first One feature in the epoch eigenvalues, Indicates the first One feature in the sliding background window The background median, Indicates the epoch index within the window. Represents the epoch Sliding background window, Indicates the first The feature is within the background window. eigenvalues; Indicates the first The median absolute deviation of each feature within the background window; To prevent positive numbers with a denominator of zero, The positive part function is represented. Formula (9) and Formula (21) use the same statistical principle. Both are based on robust scaling estimation of the median and median absolute deviation within the sliding background window. However, Formula (9) is an independent standardization operation performed on the three constituent components of the power degradation class feature, while Formula (21) is a standardization process performed on each feature that has been constructed. The two belong to different levels of processing.

[0153] Evidence-based mapping of standardized features yields the feature evidence strength:

[0154] (twenty two)

[0155] in, Indicates the first One feature in the epoch The characteristic evidence strength, with a value range of [value missing]. ; Indicates the first One feature in the epoch robust normalization characteristics Indicates the first The mapping sensitivity parameter of each feature, This represents the evidence mapping function.

[0156] Robust standardization can reduce the impact of background fluctuations and outliers on feature scale, while evidence mapping can unify features with different dimensions and statistical distributions into a range of values. The strength of the evidence that can be fused between them.

[0157] Step S4: Construct adaptive weights.

[0158] An adaptive weighting system is constructed based on the strength of feature evidence. This adaptive weighting is jointly determined by an observation reliability factor, an anomaly phase matching factor, a mutation sensitivity factor, and a redundancy suppression factor. The observation reliability factor characterizes the effective satellite support of each feature; the anomaly phase matching factor characterizes the matching relationship between the current anomaly state and each anomaly phase; the mutation sensitivity factor enhances the response during the rising phase of feature evidence; and the redundancy suppression factor suppresses duplicate contributions from evidence of consistency with the same source observations.

[0159] S4-1, according to the... The observation reliability factor is constructed by comparing the ratio of the number of valid satellites participating in the statistics to the number of visible satellites at the current epoch.

[0160] (twenty three)

[0161] in, Indicates the first One feature in the epoch The observation reliability factor Indicates feature index, ; Indicates the first One feature in the epoch The number of valid satellites included in the statistics Represents the epoch The number of visible satellites, To prevent positive numbers with a denominator of zero. When the number of satellites included in the statistics is small, the reliability of the corresponding features decreases.

[0162] S4-2, Construct an anomaly stage index based on the anomaly stage depth corresponding to each feature and the current evidence strength:

[0163] (twenty four)

[0164] in, Represents the epoch Abnormal phase index, Indicates the first One feature in the epoch The strength of characteristic evidence, To prevent positive numbers with a denominator of zero; Indicates the first The depth of the anomaly stage corresponding to each feature is a fixed constant. Indicates a power degradation category, in the early stages; Indicates pseudorange-Doppler consistency, mid-range; Indicates carrier phase-Doppler consistency, mid-term; This indicates a continuous failure type, which is late stage; pseudorange-Doppler consistency and carrier phase-Doppler consistency both belong to the intermediate stage of observation consistency failure, but they describe the same level of anomalies from the perspectives of different observations (code observation and phase observation).

[0165] Pseudorange-Doppler consistency and carrier phase-Doppler consistency both belong to the intermediate stage of observation consistency disruption. They are at the same depth in the anomalous evolution hierarchy, but may not occur strictly synchronously in actual events. This invention uses short-term correlation analysis in a redundancy suppression factor to jointly process both. When they are asynchronous, the correlation decreases, and redundancy suppression weakens, thus allowing each to contribute risk evidence independently, without ignoring their temporal differences due to the same stage depth.

[0166] Formula (23) is used to calculate the stage position of the current epoch anomaly. The anomaly stage depth reflects the position of the feature in the anomaly evolution hierarchy, where power degradation features correspond to the early stage, observation consistency features correspond to the middle stage, and continuous failure features correspond to the late stage. When the evidence strength of all features is 0, the anomaly stage index is calculated. Defined as 0, at this point the anomaly stage matching factor for each feature is... It remains a finite value, but since the strength of evidence is 0 for all of them, the main risk aggregation term... Joint early warning risk score Unaffected. Therefore, this situation will not lead to false alarms.

[0167] S4-3, Constructing the matching factor for the anomaly phase:

[0168] (25)

[0169] in, Indicates the first One feature in the epoch The abnormal phase matching factor, Indicates the first The depth of the abnormal stage corresponding to each feature. Represents the epoch Abnormal phase index, Indicates feature index; This indicates the stage matching scale parameter. The larger the value, the slower the decay (loose matching). The smaller the value, the faster the decay (strict matching); This represents the natural exponential function. The anomaly phase matching factor assigns higher weight to evidence strength that matches its phase depth.

[0170] Based on the positive increment of the strength of characteristic evidence between adjacent epochs, a mutation sensitivity factor is constructed:

[0171] (26)

[0172] in, Indicates the first One feature in the epoch The mutation sensitivity factor has a value range of 100%. This is used to enhance the response during the rising phase of characteristic evidence; Indicates the parameter for regulating the mutation response. Indicates the first One feature in the epoch The strength of characteristic evidence, Indicates the first These features in the previous epoch The strength of characteristic evidence, Indicates epoch index, Indicates the previous epoch index. This indicates the operation of finding the maximum value.

[0173] Mutation-sensitive factors can enhance the response to abnormal rising edges, making early warnings more sensitive to the onset of abnormalities.

[0174] A redundancy suppression factor is constructed based on the correlation between the evidence corresponding to the first observation consistency class feature and the evidence corresponding to the second observation consistency class feature within a short time window.

[0175] The short-term window correlation is:

[0176] (27)

[0177] in, Represents the epoch The short-time window correlation coefficient, with a value range of [value range missing]. Used to quantify evidence of consistency in first observations. Evidence consistent with second observation The degree of linear correlation within a short time window, the closer the value is to 1, the more synchronous the two changes, indicating that the two types of consistent evidence reflect the same source of anomaly and there is redundancy. Indicates the current epoch. Before A short window consisting of epochs, Indicates the short-time window length. Indicates short window Calculate the Pearson correlation coefficient. This indicates that the second feature (pseudorange-Doppler consistency feature) is in the epoch. The strength of characteristic evidence, This indicates that the third feature (carrier phase-Doppler consistency feature) is in the epoch. The strength of characteristic evidence.

[0178] The corresponding redundancy suppression factor is:

[0179] and (28)

[0180] in, The redundancy suppression factor represents the second feature (pseudorange-Doppler consistency). The redundancy suppression factor represents the third feature (carrier phase-Doppler consistency). This represents the redundancy suppression strength coefficient. Indicates the relevance threshold. Represents the epoch The short-term window correlation coefficient.

[0181] When two types of consistent evidence are highly correlated, it indicates that they reflect the same source of anomaly, and their duplication contribution should be reduced.

[0182] Multiplying the observation reliability factor, the anomaly stage matching factor, the mutation sensitivity factor, and the redundancy suppression factor together yields the comprehensive support quantity:

[0183] (29)

[0184] in, Indicates the first One feature in the epoch The overall support volume Indicates the first One feature in the epoch The observation reliability factor Indicates the first One feature in the epoch The abnormal phase matching factor, Indicates the first One feature in the epoch Mutation-sensitive factors Indicates the first One feature in the epoch Redundancy suppression factor.

[0185] The comprehensive support values ​​of each feature are then normalized to obtain adaptive weights:

[0186] (30)

[0187] in, Indicates the first One feature in the epoch Adaptive weights, Indicates the first One feature in the epoch The overall support volume Indicates the first One feature in the epoch The overall support capacity; Represents the epoch The sum of the combined support of all four characteristics, The temporary feature index used for summation has a value range of {1, 2, 3, 4}; To prevent positive numbers with a denominator of zero.

[0188] Step S5: Construct a joint early warning risk score.

[0189] The main risk aggregation term is constructed based on the strength of feature evidence and the adaptive weights. A co-enhancement term is constructed based on the stage evolution relationship between power degradation, observation consistency disruption and continuity failure represented by the multi-feature set in the observation domain. The main risk aggregation term and the co-enhancement term are combined and then subjected to time smoothing to obtain the joint early warning risk score.

[0190] like Figure 2 As shown, the main risk aggregation term is constructed based on the feature evidence strength and the adaptive weights:

[0191] (31)

[0192] in, Represents the epoch The main risk aggregation item, Indicates the first One feature in the epoch Adaptive weights, Indicates the first One feature in the epoch The strength of characteristic evidence, Indicates the feature index.

[0193] The main risk aggregation term reflects the weighted sum of the evidence strength of each feature, with the weights determined by adaptive weights.

[0194] Based on the stage evolution relationship between power degradation, observation consistency violation, and continuity failure characterized by a multi-feature set in the observation domain, feature evidence with time delay compensation is constructed:

[0195] (32)

[0196] in, Indicates the first One feature in the epoch Evidence with time delay compensation features Indicates the first The feature in the first The strength of characteristic evidence for each epoch, Indicates the epoch number that is lagging behind. Indicates the first The maximum time delay compensation length for each feature.

[0197] Time delay compensation is used to capture the temporal sequence of events during the evolution of abnormal phases. In this embodiment, the maximum time delay compensation length for each feature is set as follows: power degradation features Each epoch, observation consistency class feature Individual epochs, characteristics of continuous failures Each epoch. The value is determined based on the typical timescale of GNSS anomaly evolution: power degradation to observation consistency failure typically requires several epochs, and observation consistency failure to continuity failure may take even longer. In practical applications, this can be adjusted according to the data sampling rate; the higher the sampling rate, the more... The value can be increased accordingly.

[0198] Further build collaborative enhancements:

[0199] (33)

[0200] in, Represents the epoch Synergistic enhancement terms; , , The weighting coefficients of the synergistic enhancement terms are pre-defined non-negative constants that control the strength of the three synergistic effects respectively. This indicates that the first feature (power degradation class) is in the epoch. Evidence with time-delay compensation characteristics; This indicates that the second feature (pseudorange-Doppler consistency) and the third feature (carrier phase-Doppler consistency) are in epoch. The average value of evidence with time-delay compensation characteristics; This indicates that the fourth characteristic (continuous failure type) is in the epoch. Evidence with time-delay compensation characteristics, This indicates that the second feature is in the epoch. The strength of characteristic evidence, This indicates that the third feature is in the epoch. The strength of characteristic evidence.

[0201] The synergistic enhancement term utilizes the physical laws governing the evolution of the anomaly phase: when power degradation and observation consistency disruption occur simultaneously, the anomaly risk is higher than the sum of the risks of the two occurring individually; similarly, when observation consistency disruption and continuity failure occur simultaneously, it also indicates that the anomaly has developed in a serious direction.

[0202] Combining the main risk aggregation term with the synergistic enhancement term yields the instantaneous joint anomaly score:

[0203] (34)

[0204] in, Represents the epoch Instantaneous joint anomaly score, Represents the epoch The main risk aggregation item, Indicates the weight of collaborative enhancement. Represents the epoch Synergistic enhancements.

[0205] The joint early warning risk score is obtained by performing time smoothing on the instantaneous joint anomaly score:

[0206] (35)

[0207] in, Represents the epoch Joint early warning risk score; This is the time smoothing coefficient. The larger the value, the stronger the smoothing and the slower the response. It is generally taken as 0.7~0.9. Indicates the previous epoch Joint early warning risk score, Represents the epoch The instantaneous joint anomaly score. Time smoothing can suppress instantaneous fluctuations and make the risk score changes more stable.

[0208] Step S6: Classification and early warning of abnormal events.

[0209] A dynamic grading threshold is constructed based on the joint early warning risk score, and the core segment of the abnormal event is extracted by combining the event persistence constraint and event-level evidence judgment. The event-level quantile judgment is performed on the core segment of the abnormal event to determine the early warning level of the abnormal event, and the start time, end time, duration, early warning level and corresponding feature evidence results of the abnormal event are output.

[0210] Based on the median and median absolute deviation of the joint early warning risk score within the background window, a three-level dynamic classification threshold is constructed:

[0211] (36)

[0212] in, Represents the epoch The Level dynamic grading threshold, Represents the epoch The background median of the joint early warning risk score, Indicates the first The multiplier corresponding to the threshold level, Represents the epoch Background robust scaling estimation of joint early warning risk score This represents the level index, with a value range of {1, 2, 3}, representing the warning thresholds for primary, intermediate, and advanced levels, respectively.

[0213] (37)

[0214] (38)

[0215] in, Represents the epoch The background median of the joint early warning risk score, Indicates in the background window All epochs within Take the median above, Represents the epoch Background robust scaling estimation of joint early warning risk score Represents the epoch Sliding background window, Indicates in the background window Inside, calendar Joint early warning risk score, This indicates the epoch index within the window. The dynamic grading threshold can adaptively change according to the current background risk level.

[0216] The instantaneous warning level is determined by comparing the joint early warning risk score with the three-level dynamic grading threshold.

[0217] The instantaneous warning level is subjected to event persistence constraints to obtain risk candidate event segments. Event persistence constraints are used to suppress single-epoch spikes and short-term fluctuations, ensuring a certain degree of temporal continuity for abnormal events. Specifically, a risk candidate event segment is formed only when the instantaneous warning level is greater than 0 for multiple consecutive epochs. When the instantaneous warning level is at least [number missing] consecutive [epochs missing]... When the instantaneous warning level of each epoch is greater than 0, the continuous interval is marked as a risk candidate event segment. This represents the minimum duration epoch of the event persistence constraint. In this embodiment... The value is set to 3 epochs. This value can be adjusted according to the sampling rate. When the sampling rate is higher than 1Hz, it can be increased appropriately. .

[0218] Based on the event-level evidence thresholds for the strength of each characteristic evidence, candidate segments of characteristic evidence are determined. The event-level evidence thresholds are as follows:

[0219] (39)

[0220] in, Indicates the first The event-level evidence threshold for each feature, Indicates taking the first The sequence of evidence strength of each feature in all historical data The median of Represents a sequence of characteristic evidence strength. Indicates taking the first The sequence of evidence strength of each feature in all historical data The median absolute deviation, This represents the event-level threshold multiplier. This represents the clipping function. and These represent the lower and upper limits of the event-level evidence threshold, respectively, in this embodiment. , .

[0221] In this embodiment, for the first One characteristic, when the epoch Characteristic Evidence Strength Greater than the event-level evidence threshold At that time, mark this epoch as the 1st epoch. There are candidate epochs for feature evidence. A minimum duration constraint is imposed on the candidate epochs for feature evidence: no less than [number] consecutive [durations]. An interval of epochs that satisfies the above conditions constitutes a candidate segment of feature evidence for that feature. The minimum candidate segment duration epoch is preset, and the preferred value is... .

[0222] The core segment of anomalies is determined based on the risk candidate event segment and the characteristic evidence candidate segment. Specifically, the intersection of the two is taken as the core segment of anomalies to ensure that the epochs within the core segment have both a sustained level of risk and clear observational evidence to support them.

[0223] Perform event-level quantile determination on the core segment of the abnormal event:

[0224] (40)

[0225] (41)

[0226] in, Indicates the first The core segment of the abnormal event Within, the joint early warning risk score for all eras. of quantiles; Indicates calculation Quantile operators, Indicates quantiles, Represents the epoch Joint early warning risk score, Indicates epoch index, Indicates the first The core detection range for each abnormal event. Indicates the first The core segment of the abnormal event within, no. Strength of evidence for each feature of quantiles; Indicates the first One feature in the epoch The strength of characteristic evidence, Indicates feature index, Indicates the abnormal event number.

[0227] The final warning level of an abnormal event is determined based on the event-level quantile decision. Specifically, the quantile of each characteristic evidence strength within the core segment of the event is compared with the corresponding characteristic evidence threshold to comprehensively determine the event level. Dynamic grading thresholds can adaptively change according to the current background risk level; event persistence constraints can suppress transient false alarms; event-level evidence thresholds can ensure that abnormal events are supported by clear observational evidence; and event-level quantile decisions can avoid the excessive influence of a single extreme epoch on the event level.

[0228] Finally, the start time, end time, duration, warning level, and corresponding feature evidence results of the abnormal event are output.

[0229] Experimental verification:

[0230] like Figure 3 As shown in (a) to (d), this invention constructs four types of observation domain features based on measured GNSS observation data, including common-mode carrier-to-noise ratio anomaly features. (like Figure 3 (a) Pseudorange, Doppler consistency characteristics (like Figure 3 (b) Carrier phase and Doppler consistency characteristics (like Figure 3 (c) and cycle slip or loss of lock ratio characteristics (like Figure 3 (d)). By Figure 3 It is evident that during the occurrence of abnormal events, The first significant increase indicates that the carrier-to-noise ratio of multiple satellites is simultaneously degrading; The high level of activity in the early stages of the event indicates that some satellites experienced continuous disruptions such as cycle slips or loss of lock; as the event progressed, and The presence of varying degrees of enhancement indicates that the consistency between pseudorange observations, carrier phase observations, and Doppler observations has been disrupted. This result demonstrates that this invention, by constructing multiple types of observation domain features, can characterize GNSS anomalies from four aspects: power degradation, code observation consistency, carrier phase consistency, and observation continuity, avoiding the problem of incomplete anomaly characterization caused by relying on only a single feature.

[0231] like Figure 4 As shown, after obtaining the above four types of observation domain characteristics, this invention further constructs a joint early warning risk score. And combined with dynamic hierarchical thresholds , , The severity level of abnormal events is determined. The red curve in the graph represents... The temporal changes are shown, with dashed lines representing the judgment thresholds corresponding to different risk levels. Figure 3 Compared to the dispersed responses of each feature, Figure 4 In Compressing multi-source observed anomalies into a single continuous risk curve makes the onset, duration, and intensification phases of the anomalous process clearer. During the anomalous event, The level has remained high for an extended period and has exceeded the high-level warning threshold multiple times. This indicates that the event was not caused by a single epochal peak or individual characteristic fluctuations, but rather by a sustained high-risk process supported by multiple types of observational evidence. Furthermore, the 90th percentile of the joint early warning risk score within the core segment of the event... The value is higher than the high-level threshold, therefore the event is classified as a Level Lv3 warning event. Using event-level quantiles instead of single-point maximum values ​​for the decision reduces the impact of instantaneous spikes, short-term dips, and local noise on the classification results, making the final warning level more reflective of the overall risk level of the event.

[0232] like Figure 5 As shown, this invention further calculates the true weighted contribution of various features within anomaly events. ,in, Indicates the first Adaptive weights for class features Indicates the first The strength of evidence for class features Indicates the first The contribution of class features to the joint early warning risk score. Figure 5 As can be seen, the contribution value of the first type of feature is 0.530, the contribution value of the second type of feature is 0.409, the contribution value of the third type of feature is 0.234, and the contribution value of the fourth type of feature is 0.182. This indicates that the first type of feature contributes the most to the joint risk score of this event, while the second type of feature also has a significant contribution. This result demonstrates that this invention, while providing an event-level warning level, can further quantify the contribution of different observation domain features to the joint risk score, thereby explaining the main sources of evidence for the formation of the warning level and improving the interpretability and credibility of GNSS anomaly classification warning results.

[0233] The above embodiments are merely preferred examples of the present invention and are not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention should fall within the protection scope of the present invention.

Claims

1. A GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain, characterized in that, Includes the following steps: Acquire GNSS observation data, analyze and preprocess the GNSS observation data, extract carrier-to-noise ratio, pseudorange, carrier phase, Doppler observation values ​​and loss-of-lock flag, and construct an observation data matrix according to epoch and satellite number; Based on the observation data matrix, a multi-feature set of the observation domain is constructed. The multi-feature set of the observation domain includes at least power degradation features, observation consistency features, and continuous failure features. Robust standardization and evidence mapping are performed on each feature in the multi-feature set of the observation domain to obtain the feature evidence strength corresponding to each feature. An adaptive weight is constructed based on the feature evidence strength, the observation validity of each feature, the abnormal stage attributes, the temporal changes of evidence, and the correlation between features. The adaptive weight is determined by a combination of the observation reliability factor, the abnormal stage matching factor, the mutation sensitivity factor, and the redundancy suppression factor. A joint early warning risk score is constructed based on the strength of the feature evidence and adaptive weights. Based on the joint early warning risk score, a dynamic hierarchical threshold is constructed, and the early warning level of the abnormal event is determined by combining the event persistence constraint and the event-level evidence judgment, and the early warning information of the abnormal event is output.

2. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 1, characterized in that, The method for constructing the power degradation feature includes: establishing a carrier-to-noise ratio (CNR) sliding background baseline for each satellite, calculating the degradation amount of the satellite's current CNR relative to the sliding background baseline, and marking the satellite as an affected satellite when the degradation amount is greater than the single-satellite degradation threshold; applying multi-satellite synchronization constraints based on the number of affected satellites and the number of effective satellites; constructing at least a common-mode edge drop sub-quantum based on the positive CNR decrease, a common-mode proportional sub-quantum based on the proportion of affected satellites, and a common-mode platform drop sub-quantum based on the CNR degradation amount of affected satellites, and weighted fusing the three sub-quantums to obtain the power degradation feature.

3. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 1, characterized in that, The observation consistency class features include a first observation consistency class feature and a second observation consistency class feature; The first observation consistency category feature constructs pseudorange-Doppler consistency residuals based on pseudorange time difference scores and Doppler observations. After median centering, it is constructed based on the residual robust scale and the residual abnormal satellite ratio. The second observation consistency feature converts carrier phase observations into metric carrier phase quantities, constructs carrier phase-Doppler consistency residuals based on metric carrier phase time difference scores and Doppler observations, and then constructs them based on the residual robust scale and the residual abnormal satellite ratio after median centering.

4. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 3, characterized in that, The method for constructing the continuous failure class features includes: Cycle slip status is detected based on the adjacent epoch abrupt change in the carrier phase-Doppler consistency residual; The single-satellite loss status is determined by combining at least one of the following information: loss of lock flag, carrier phase observation missing status, observation time discontinuity status, continuous weak tracking status, and lock time reset status. The cycle slip state and the unlock state are unified into a single-satellite continuous failure flag, and the proportion of continuously failing satellites is calculated within the reference satellite set to obtain the continuous failure class characteristics.

5. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 4, characterized in that, The robust normalization process includes: For the Features Calculate the background median and median absolute deviation within a sliding background window, and construct the standardized features according to the following formula: ； in, Indicates the first One feature in the epoch robust normalization characteristics Indicates feature index, Indicates the first One feature in the epoch eigenvalues, Indicates the first One feature in the sliding background window The background median, Indicates the epoch index within the window. Represents the epoch Sliding background window, Indicates the first The feature is within the background window. eigenvalues; Indicates the first The median absolute deviation of each feature within the background window; To prevent positive numbers with a denominator of zero, This represents the positive part of the function.

6. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 5, characterized in that, The evidence mapping process includes mapping the standardized features to feature evidence strength, as shown in the formula: ； in, Indicates the first One feature in the epoch The strength of characteristic evidence, Indicates the first One feature in the epoch robust normalization characteristics Indicates the first The mapping sensitivity parameter of each feature, This represents the evidence mapping function.

7. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 6, characterized in that, The method for constructing the adaptive weights includes: According to the The ratio of the number of valid satellites participating in the statistics at the current epoch to the number of visible satellites at the current epoch is used to construct the observation reliability factor. An anomaly stage index is constructed based on the depth of the anomaly stage corresponding to each feature and the current strength of evidence. And construct an abnormal stage matching factor based on the abnormal stage index; Based on the positive increment of the strength of characteristic evidence between adjacent epochs, a mutation sensitivity factor is constructed; Based on the correlation between the evidence corresponding to the first observation consistency class feature and the evidence corresponding to the second observation consistency class feature within a short time window, a redundancy suppression factor is constructed. The adaptive weights are obtained by multiplying and normalizing the observation reliability factor, anomaly phase matching factor, mutation sensitivity factor, and redundancy suppression factor.

8. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 7, characterized in that, The method for constructing the joint early warning risk score includes: Based on the strength of feature evidence and adaptive weights, construct the main risk aggregation term: ； in, Represents the epoch The main risk aggregation item, Indicates the first One feature in the epoch Adaptive weights, Indicates the first One feature in the epoch The strength of characteristic evidence, Indicates feature index; Synergistic enhancement terms are constructed based on the stage evolution relationship between power degradation, observation consistency violation, and continuity failure. The main risk aggregation term and the synergistic enhancement term are combined to obtain an instantaneous joint anomaly score, and the instantaneous joint anomaly score is then subjected to time smoothing to obtain a joint early warning risk score.

9. The GNSS anomaly classification and early warning method based on multi-evidence stage evolution in the observation domain according to claim 8, characterized in that, The process of determining the warning level of an abnormal event and outputting the warning information for the abnormal event specifically includes: Based on the median and median absolute deviation of the joint early warning risk score within the background window, a three-level dynamic classification threshold is constructed. The instantaneous warning level is determined based on the comparison between the joint early warning risk score and the three-level dynamic grading threshold. The instantaneous warning level is subjected to event persistence constraint processing to obtain risk candidate event segments; Based on the event-level evidence thresholds for the strength of each characteristic evidence, candidate segments of characteristic evidence are determined; The core segment of the abnormal event is determined based on the risk candidate event segment and the feature evidence candidate segment; The core segment of the abnormal event is subjected to event-level quantile judgment to determine the final warning level of the abnormal event; Output the start time, end time, duration, warning level, and corresponding feature evidence results of the abnormal event.

Images