Convolution verification branch scheduling method and device based on GNSS cross-domain physical consistency challenge, equipment and storage medium

Abstract

The present application relates to the field of navigation signal processing, and discloses a convolution verification branch scheduling method and device based on GNSS cross-domain physical consistency challenge, equipment and storage medium. The method comprises: acquiring feature snapshot data including at least two observation domain data; calling a first convolution model to process first GNSS observation domain data, acquiring candidate hypothesis data and hypothesis observation constraint template, comparing the physical constraint relationship between the second GNSS observation domain data and the template, and generating contradiction vector information; acquiring branch descriptor including discrimination code and source code; based on the matching degree of contradiction vector information and the discrimination code of each convolution verification branch and the matching degree of the corresponding source code and the source code of the called convolution verification branch, selecting the most matched target branch from the uncalled convolution verification branch, and acquiring the processing result of the candidate hypothesis data. The technical scheme of the embodiment of the present application improves the output result reliability, reduces the task pressure and hardware power consumption.

Description

Technical Field

[0001] This invention relates to the field of navigation signal processing, and in particular to a method, apparatus, device, and storage medium for convolutional verification branch scheduling based on GNSS cross-domain physical consistency challenges. Background Technology

[0002] Satellite navigation signal receivers can use CNNs to identify spectral interference, correlation peak distortion, multipath propagation, spoofing, tracking anomalies, or observation quality issues in navigation signals. However, simultaneously performing these anomaly detection processes at the chip level of the navigation device adds a large amount of data to be processed, resulting in significant workload and high hardware power consumption.

[0003] Furthermore, existing navigation devices have the following problems when performing local processing: 1. Directly selecting light, medium, or heavy models based on SNR or task level may mix weak signals, multipath interference, suppression interference, and spoofing into data of the same complexity, leading to task scheduling errors.

[0004] 2. In the current navigation data processing workflow, early termination and branching are usually determined based on the confidence level of the model output, the primary feature category, or the resource budget. However, a high confidence level in a CNN does not necessarily mean that its conclusions conform to GNSS physical relationships, leading to unreliable confidence levels. For example, a spectral branch might classify a power drop as suppression interference, while the correlation peak shape and cross-satellite residuals at the same epoch are more consistent with deceptive drag.

[0005] 3. The existing navigation data processing flow cannot judge or detect the reliability of the model output results, and cannot repair or adjust the model when it malfunctions. Summary of the Invention

[0006] This invention provides a method, apparatus, device, and storage medium for convolutional verification branch scheduling based on GNSS cross-domain physical consistency challenges, in order to solve at least one of the above-mentioned problems.

[0007] In a first aspect, embodiments of the present invention provide a convolutional verification branch scheduling method based on the GNSS cross-domain physical consistency challenge, comprising: acquiring feature snapshot data, wherein the feature snapshot data includes at least two different GNSS observation domain data within the same observation epoch; calling the initial convolutional model to process the first GNSS observation domain data to generate candidate hypothesis data in a provisional state; acquiring a hypothesis observation constraint template corresponding to the candidate hypothesis data, wherein the hypothesis observation constraint template includes physical constraint relationships between the candidate hypothesis data and at least one second GNSS observation domain data different from the first GNSS observation domain data; and, based on the hypothesis observation constraint template, configuring the second GNSS observation domain data with... Physical constraints are compared to generate contradiction vector information to identify candidate hypothesis data that does not satisfy physical constraints. Branch descriptors for multiple convolutional validation branches in the initial convolutional model are obtained. Each branch descriptor includes a discriminant code for the contradiction type of each convolutional validation branch and a source code representing the GNSS observation domain to which the input data that each convolutional validation branch can receive belongs. Based on the matching degree between the contradiction vector information and the discriminant code of each convolutional validation branch, and the matching degree between the corresponding source code and the source code of the invoked convolutional validation branches, the most matching target branch is selected from the uninvoked convolutional validation branches. The processing result of the candidate hypothesis data is obtained through the target branch.

[0008] Optionally, the GNSS observation domain of the feature snapshot data includes at least two of the radio frequency domain, correlation domain, tracking domain, and navigation solution domain; wherein, the data in the radio frequency domain includes the fast Fourier transform spectrum and / or automatic gain control status, the data in the correlation domain includes multi-tap correlation values ​​and / or correlation peak shapes, the data in the tracking domain includes code carrier difference and / or Doppler frequency shift difference, and the data in the navigation solution domain includes cross-satellite residual common mode data and / or receiver clock drift signal.

[0009] Optionally, the initial convolutional model includes multiple convolutional verification branches, such as: a correlation domain morphology branch, used to identify whether multi-tap correlation values ​​and / or correlation peak morphology are abnormal, to generate anomaly judgment results; a cross-satellite residual common mode branch, used to identify whether there is a common mode between multiple satellites based on cross-satellite residual common mode data, and / or to identify whether drift occurs based on receiver clock drift signals, to generate drift data; a tracking domain consistency branch, used to obtain the code pseudorange change rate based on code carrier difference, and / or to obtain the carrier phase consistency result of the carrier phase change rate based on Doppler frequency shift difference; and a spectrum consistency branch, used to obtain the spectrum state of the feature snapshot data based on the fast Fourier transform spectrum shape and / or automatic gain control state.

[0010] Optionally, the candidate hypothesis data can be any one of the following: anomaly detection results, drift data, carrier phase consistency results, and spectrum status.

[0011] Optionally, the feature snapshot data further includes a data epoch identifier and a data generation identifier. The method further includes: acquiring the correlator tap interval, coherent integration length, incoherent integration length, tracking mode, and navigation filter reference data of the navigation device; updating the data epoch identifier of the feature snapshot data when at least one of the correlator tap interval, coherent integration length, incoherent integration length, tracking mode, and navigation filter reference data changes; generating a corresponding evidence token after acquiring the processing result, the evidence token including the data epoch identifier and the data generation identifier; comparing whether the updated data epoch identifier and data generation identifier of the feature snapshot data are consistent with the data epoch identifier and data generation identifier of the evidence token; if consistent, verifying the processing result; if inconsistent, generating invalidation information for the evidence token.

[0012] Optionally, the step of verifying the processing result includes: based on the evidence token, verifying whether the processing result meets the preset resolution condition and the independent source quantity condition; if both are met, outputting the verified processing result; if the preset resolution condition or the independent source quantity condition is not met, reselecting a new target branch from the uncalled convolutional verification branch to obtain a new processing result and a new evidence token, and verifying again until both conditions are met, or if the two conditions are still not met after a preset submission time, generating uncertain state information.

[0013] Optionally, the preset resolution condition is that the processing result includes the processed data of each contradiction in the contradiction vector information; the independent source quantity condition is that the processing result includes at least two different GNSS observation domain data, wherein the physical links of the at least two different GNSS observation domain data are independent and are not simultaneously obtained from the same original observation domain data or the same convolution model.

[0014] Optionally, assuming the observation constraint template includes multiple preset physical constraint relationships, the step of comparing the second GNSS observation domain data with the physical constraint relationships based on the assumed observation constraint template to generate contradiction vector information for identifying candidate hypothetical data that does not satisfy the physical constraint relationships includes: obtaining the physical constraint relationships corresponding to the second GNSS observation domain data and the contradiction bit mask based on the assumed observation constraint template; performing a consistency check on the second GNSS observation domain data based on the physical constraint relationships; and setting the bit at the corresponding position in the contradiction bit mask to generate contradiction vector information when the feature snapshot data of the second GNSS observation domain does not conform to the physical constraint relationships.

[0015] Optionally, the physical constraint relationship includes a preset numerical range of candidate hypothesis data and / or a preset expected change direction, and different bits in the contradiction bit mask correspond to different physical constraint relationships; when the second GNSS observation domain data is not in the preset numerical range, and / or when the change direction of the second GNSS observation domain data is inconsistent with the expected change direction after exceeding the dead zone, the bit at the corresponding position in the contradiction bit mask is set.

[0016] Optionally, the step of selecting the most matching target branch from the uncalled convolutional verification branches based on the matching degree between the contradiction vector information and the discriminant code of each convolutional verification branch, and the matching degree between the corresponding source code and the source code of the called convolutional verification branches, includes: when there is an intersection between the contradiction type indicated by the discriminant code of the uncalled convolutional verification branch and the contradiction type of the contradiction vector information, and the number of overlapping bits between the corresponding source code and the set of source codes of all called convolutional verification branches does not exceed a preset overlap threshold, all corresponding convolutional verification branches are taken as candidate convolutional verification branches; or based on the source code of the uncalled convolutional verification branch, it is detected whether the GNSS observation domain data corresponding to the source code is used by the called convolutional verification branches, and if it is not used, all corresponding convolutional verification branches are taken as candidate convolutional verification branches; and the most matching convolutional verification branch is selected from all candidate convolutional verification branches as the target branch.

[0017] Optionally, the branch descriptor also includes the longest execution cycle and energy consumption parameter data of the corresponding convolutional validation branch. The step of selecting the most matching convolutional validation branch as the target branch from all candidate convolutional validation branches includes: obtaining the data recording time in the feature snapshot data; based on the discriminant code, source code, longest execution cycle, and energy consumption parameter data of each candidate convolutional validation branch, obtaining the contradiction coverage of the contradiction vector information, the number of observation domains not used by the called branches, the longest execution cycle, and the energy consumption parameter data of the corresponding candidate convolutional validation branch; calculating the ranking score of each candidate convolutional validation branch based on the contradiction coverage, the number of observation domains not used by the called branches, the data recording time, the longest execution cycle, and the energy consumption parameter data; and selecting the candidate convolutional validation branch with the highest ranking score as the most matching target branch.

[0018] Secondly, embodiments of the present invention provide a navigation chip, comprising: a snapshot controller, configured to acquire feature snapshot data of at least two different GNSS observation domains within the same observation epoch; a first convolution register, configured to store and call a first convolution model to process the feature snapshot data of a first GNSS observation domain, generating candidate hypothesis data in a provisional state; a physical consistency check generator, configured to acquire a hypothesis observation constraint template corresponding to the candidate hypothesis data, wherein the hypothesis observation constraint template includes physical constraint relationships of the candidate hypothesis data in at least one other GNSS observation domain different from the first GNSS observation domain; and based on the hypothesis observation constraint template, to perform a consistency check on the feature snapshot data of a second GNSS observation domain, generating candidate hypotheses. The system includes: a conflict vector information in the data that does not satisfy physical constraints; a branch descriptor memory for storing and calling branch descriptors of multiple convolutional validation branches in the initial convolutional model, wherein the branch descriptor includes a discriminant code for the conflict type corresponding to each convolutional validation branch, and a source code for indicating the GNSS observation domain to which the input data that each convolutional validation branch can receive belongs; a validation branch selector for selecting the most matching target branch from the uncalled convolutional validation branches based on the matching degree between the conflict vector information and the discriminant code of each convolutional validation branch, and the matching degree between the corresponding source code and the source code of the called convolutional validation branches; and an output controller for obtaining the processing results of the candidate hypothesis data through the target branch and outputting the processing results.

[0019] Thirdly, embodiments of the present invention provide an electronic device, including: a processor and a memory, wherein the memory stores instructions; the processor invokes the instructions in the memory to cause the processor to execute the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to any of the foregoing embodiments of the first aspect of the present invention.

[0020] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing instructions that, when executed by a processor, implement the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to any of the foregoing embodiments of the first aspect of the present invention.

[0021] The technical solution of the above embodiments of the present invention uses the output of the initial convolutional model as a candidate hypothesis and generates a contradiction vector based on the hypothesis observation constraint template. Then, it selects the convolutional verification branch as needed according to the discriminant code and the source code. This avoids the situation where the strength of GNSS signal is directly equated with the model complexity, which leads to the selection of the wrong convolutional path. This ensures the interpretability and completeness of the model output results and improves the credibility of the model output results by using data from different observation domains. At the same time, it eliminates the need for the model to call additional branches to process redundant data, thereby reducing the task pressure and power consumption of the navigation equipment. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0023] Figure 1 This is a flowchart of an embodiment of the convolutional verification branch scheduling method based on the GNSS cross-domain physical consistency challenge of the present invention; Figure 2 This is a flowchart of step S140 in one embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention. Figure 3 This is a flowchart illustrating the selection of a target branch in one embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention. Figure 4 This is a flowchart of the verification process in another embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention. Figure 5 This is a schematic diagram of feature snapshot data in one embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention. Figure 6 This is a flowchart illustrating the generation of contradictory vector information in one embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention. Figure 7 This is a flowchart illustrating the selection of convolutional verification branches in one embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenges of the present invention. Figure 8 This is a flowchart illustrating the selection of target branches based on contradiction vector information to generate processing results in one embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention. Figure 9 This is a flowchart illustrating the verification of processing results in one embodiment of the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention. Figure 10 This is a structural block diagram of one embodiment of the navigation chip of the present invention; Figure 11 This is a flowchart illustrating the workflow of one embodiment of the navigation chip of the present invention. Figure 12 This is a structural block diagram of one embodiment of the electronic device of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0025] It should be noted that all directional indications in the embodiments of the present invention, such as up, down, left, right, front, back, etc., are only used to explain the relative positional relationship and movement of the components in a specific posture as shown in the attached figure. If the specific posture changes, the directional indication will also change accordingly.

[0026] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0027] For ease of understanding, the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge of the present invention is described below, such as... Figure 1 As shown, the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge in this embodiment of the invention includes steps S110 to S170.

[0028] In step S110, feature snapshot data is acquired. The feature snapshot data includes data from at least two different GNSS observation domains within the same observation epoch.

[0029] In this embodiment, the GNSS observation domain of the feature snapshot data includes at least two of the radio frequency domain, correlation domain, tracking domain, and navigation solution domain.

[0030] The data in the radio frequency domain includes the fast Fourier transform spectrum and / or automatic gain control status; the data in the correlation domain includes multi-tap correlation values ​​and / or correlation peak shapes; the data in the tracking domain includes code carrier difference and / or Doppler frequency shift difference; and the data in the navigation solution domain includes cross-satellite residual common mode data and / or receiver clock drift signal.

[0031] Specifically, the data in the radio frequency domain can be any one, two, or more of the following: ADC amplitude, AGC code, FFT spectrum, in-band / out-of-band power ratio, and pulse duty cycle.

[0032] The data in the correlation domain can be any one, two, or more of the following: early, immediate, late, and multi-tap correlation values, number of peaks, peak spacing, left-right slope difference, and correlation peak symmetry.

[0033] The data in the tracking domain can be any one, two, or more of the following: code phase error, carrier phase error, code-carrier difference, Doppler information, PLL / DLL lock flag, and integration length.

[0034] The data in the navigation solution domain can be any one, two, or more of the following: pseudorange / carrier residuals of each satellite signal, cross-satellite common modes, receiver clock drift, filtering information, and geometric information.

[0035] like Figure 5 As shown, the feature snapshot data in this embodiment is latched by the snapshot controller at the observation epoch boundary. The feature snapshot data also includes the storage address, data length, valid bits, data age, source code, data epoch identifier EPOCH ID, and data generation identifier GID for each GNSS observation domain data.

[0036] By attaching a data epoch identifier and a data generation identifier to each GNSS observation domain data, it is possible to avoid the erroneous splicing of observation data generated by different observation epochs and different tracking configurations, which could lead to false contradictions or false consistency.

[0037] If the AGC range, correlator tap interval, coherent integration length, tracking mode, or navigation filter reference changes within a data processing cycle, the corresponding domain's generational identifier increments. When the convolutional verification branch returns a result, the data epoch identifier and the generational identifier are compared again; inconsistent results are discarded and a failure message is generated.

[0038] Optionally, the feature snapshot data may also include external observation domain data, which is independent verification input data. External observation domain data includes at least one of the following: inertial measurement unit pre-integration data, wheel speed data, clock reference data, and antenna array output data.

[0039] In step S120, the initial convolutional model is invoked to process the first GNSS observation domain data, generating candidate hypothesis data in a temporary state.

[0040] Specifically, the multiple convolutional validation branches in the initial convolutional model include: The correlation domain morphology branch is used to identify whether the multi-tap correlation values ​​and / or correlation peak morphology are abnormal, so as to generate anomaly judgment results.

[0041] The cross-satellite residual common mode branch is used to identify whether there is a common mode in the pseudorange or carrier phase residuals between multiple satellites based on the cross-satellite residual common mode data, and / or to identify whether drift has occurred based on the receiver clock drift signal, so as to generate drift data.

[0042] The tracking domain consistency branch is used to obtain the code pseudorange change rate based on the code carrier difference and / or the carrier phase change rate based on the Doppler frequency shift difference.

[0043] The spectrum consistency branch is used to obtain the spectrum state of the feature snapshot data based on the fast Fourier transform spectrum shape and / or automatic gain control state.

[0044] Specifically, the candidate hypothesis data can be any one of the following: anomaly detection results, drift data, carrier phase consistency results, and spectrum status. In this embodiment, the anomaly detection results, drift data, carrier phase consistency results, and spectrum status are all uniformly used as evidence tokens for scheduling. The aforementioned data can be, for example, branch_id, candidate_id, constraint_id / contradiction bit, support_or_refute, source code, EPOCH_ID, GID, model_version, quality / fault_flag.

[0045] In this embodiment, the candidate hypothesis data are the temporary category or risk conclusions given by the initial convolutional model, which are only in a temporary state.

[0046] In step S130, the hypothesis observation constraint template corresponding to the candidate hypothesis data is obtained.

[0047] The assumed observation constraint template includes the physical constraint relationships of candidate hypothetical data in at least one second GNSS observation domain data that are different from the first GNSS observation domain data.

[0048] In this embodiment, the convolution branch called by the initial convolution model can be determined based on the currently available GNSS observation domain data with the lowest processing cost, such as performing lightweight convolution classification on the fast Fourier transform spectrum, while the output of the initial convolution model serves as candidate hypothesis data in a temporary state.

[0049] In other words, in this embodiment of the invention, the output of the initial convolutional model is not used as the final processing result of the navigation signal, but as candidate hypothesis data to be verified, in order to avoid the problem that the high confidence result output by the convolutional model does not conform to the GNSS physical relationship.

[0050] In step S140, based on the hypothetical observation constraint template, the second GNSS observation domain data is compared with the physical constraint relationship to generate contradiction vector information used to identify candidate hypothetical data that do not meet the physical constraint relationship.

[0051] In this embodiment, the contradiction vector information is a bit vector, and different bits in the contradiction bit mask correspond to different physical constraint relationships.

[0052] like Figure 2 and Figure 6 As shown, in this embodiment, step S140 includes steps S141 to S142.

[0053] In step S141, based on the assumed observation constraint template, the physical constraint relationship and the contradiction bit mask corresponding to the second GNSS observation domain data are obtained.

[0054] In step S142, based on the physical constraint relationship, a consistency check is performed on the data of the second GNSS observation domain. When the feature snapshot data of the second GNSS observation domain does not conform to the physical constraint relationship, the bits at the corresponding positions in the contradiction bit mask are set to generate contradiction vector information.

[0055] like Figure 6 As shown in this embodiment, specifically, it is assumed that the observation constraint template may include at least two, three or more of the following: constraint number, required observation data source, preset numerical range or expected change direction of feature snapshot data, applicable conditions, protection quantity, maximum data age, minimum number of satellites and minimum number of frequency points.

[0056] Assuming the observation constraint template includes multiple preset physical constraint relationships, the physical constraint relationships include preset numerical ranges of candidate hypothesis data and / or preset expected change directions, and different bits in the contradiction bit mask correspond to different physical constraint relationships.

[0057] When the data in the second GNSS observation domain is not in the preset value range, and / or when the direction of change of the data in the second GNSS observation domain is inconsistent with the expected direction of change after exceeding the dead zone, the bit at the corresponding position in the contradiction bit mask will be set.

[0058] When performing consistency checks on data from the second GNSS observation domain, the following formula can be used for judgment: ; in, The first within the same observation epoch Normalized second GNSS observation domain data, For the first Valid bits of the observation domain data and These are the lower and upper limits of the preset numerical range corresponding to the candidate hypothesis data, respectively. This is a protection factor used for fixed-point quantization, time alignment, and measurement noise.

[0059] exist When =1, it means the first... If a physical constraint cannot be interpreted by the current candidate hypothesis data, the bit position corresponding to that physical constraint in the contradiction bit mask will be set. When = 0, it means that the current candidate hypothesis data conforms to the first... Physical constraint relationships.

[0060] For physical constraints of the desired change direction type, after the absolute value of the change in the second GNSS observation domain data exceeds the dead zone, the actual change direction is compared with the desired change direction in the assumed observation constraint template. If they are inconsistent, the bits at the corresponding positions in the contradictory bit mask are set.

[0061] In step S150, the branch descriptors of multiple convolutional validation branches in the initial convolutional model are obtained.

[0062] The branch descriptor includes a discriminant code for the contradiction type corresponding to each convolutional verification branch, and a source code for indicating the GNSS observation domain to which the input data that each convolutional verification branch can receive belongs.

[0063] In this embodiment, the branch descriptor also includes at least one of the following: branch identifier, input feature address, input scale identifier, preset submission time category, output format, and model version. For example, the branch descriptor may also include branchid, input address or input scale, model version, energy consumption token, and output token format.

[0064] The contradiction vector information, discriminant code, and source code can be represented using 16-bit or 32-bit masks so that the navigation device hardware can filter the convolutional verification branches through bitwise AND, population counting, and priority encoder.

[0065] By encoding the unexplained physical relationships in the candidate hypothesis data into contradictory vector information in the form of bit vectors, we can avoid the problem of directly equating the strength of GNSS signals with model complexity, which would lead to the selection of the wrong convolution path.

[0066] In step S160, based on the degree of matching between the contradiction vector information and the discriminant code of each convolutional verification branch, and the degree of matching between the corresponding source code and the source code of the invoked convolutional verification branch, the most matching target branch is selected from the uninvoked convolutional verification branches.

[0067] In this embodiment, an uncalled convolutional verification branch must meet the following hard constraints to be considered as a candidate convolutional verification branch: 1. The data epoch identifier and data generation identifier of the input data of the convolutional verification branch are consistent with the feature snapshot data.

[0068] 2. The valid bits of the input data are true, and the data age does not exceed the maximum data age in the branch descriptor.

[0069] 3. The sum of the current time and the longest execution cycle of the convolutional verification branch shall not exceed the preset submission time.

[0070] Optionally, the best-matching target branch is selected from the uncalled convolutional validation branches, and the corresponding hard constraint can be: If the discriminant type indicated by the discriminant of the uncalled convolutional verification branch intersects with the contradiction type of the contradiction vector information, and the number of overlapping bits between the corresponding source code and the set of source codes of all called convolutional verification branches does not exceed a preset overlap threshold, then all corresponding convolutional verification branches will be regarded as candidate convolutional verification branches.

[0071] Alternatively, based on the source code of the uncalled convolutional verification branch, it can be detected whether the GNSS observation domain data corresponding to the source code has been used by the called convolutional verification branch. If it has not been used, all corresponding convolutional verification branches can be used as candidate convolutional verification branches.

[0072] After obtaining the candidate convolutional validation branches, the most matching convolutional validation branch is selected as the target branch from all the candidate convolutional validation branches.

[0073] like Figure 3 and Figure 7 As shown, specifically, the branch descriptor also includes the longest execution cycle and energy consumption parameter data of the corresponding convolutional verification branch. The step of selecting the most matching convolutional verification branch as the target branch from all candidate convolutional verification branches includes steps S161 to S164: In step S161, the data recording time in the feature snapshot data is obtained.

[0074] In step S162, based on the discriminant code, source code, longest execution cycle, and energy consumption parameter data of each candidate convolutional verification branch, the contradiction coverage of the contradiction vector information, the number of observation domains not used by the called branches, the longest execution cycle, and energy consumption parameter data of the corresponding candidate convolutional verification branch are obtained.

[0075] In step S163, the ranking score of each candidate convolutional validation branch is calculated based on the contradiction coverage, the number of observation domains not used by the invoked branches, the data recording time, the longest execution cycle, and energy consumption parameter data.

[0076] In step S164, the candidate convolutional verification branch with the highest ranking score is selected as the most matching target branch.

[0077] In this embodiment, the ranking score of the candidate convolutional validation branch can be calculated using the following formula: ; in, The ranking score of branch b is used to verify the candidate convolution. The discriminant code Db of the candidate convolutional verification branch determines the degree of contradiction coverage of the contradiction vector information C, which can be, for example, the number of contradictory bits covered or the weighted sum. The source code Pb for the candidate convolutional verification branch is relative to the set of source codes for all invoked convolutional verification branches. The degree of independence can be measured, for example, by the number of observation domains that have not been used by the invoked branches or the proportion of new sources. The feature freshness is determined by the data recording time and the current time of the feature snapshot data. The longest execution cycle of the candidate convolutional validation branch. Energy consumption parameter data for candidate convolutional validation branches; , , , and These are the corresponding weighting coefficients.

[0078] The weighting coefficients can be freely set according to actual needs, provided that the data sources for the same observation epoch, different observation domains are independent, and the aforementioned hard constraints are met.

[0079] In step S170, the processing results of the candidate hypothesis data are obtained through the target branch.

[0080] In this embodiment, after obtaining the processing result of the target branch, the candidate hypothesis data and outstanding contradictions are updated based on the processing result. When the processing result of the target branch negates the candidate hypothesis data, new candidate hypothesis data can be established based on the processing result.

[0081] For example, in one embodiment, the initial convolutional model outputs candidate hypothesis data for "broadband suppression of interference" based on the Fast Fourier Transform spectral shape. A consistency check reveals a decrease in automatic gain control (AGB) status, but the spectral occupancy does not expand according to the assumed observation constraint template, and the pseudorange residuals of multiple satellites exhibit a common mode. Therefore, the spectral inconsistencies in the inconsistency vector information are matched with the AGB inconsistency bits and the cross-satellite residual common mode inconsistencies. Then, the correlation domain morphology branch, originating from the correlation domain, is selected. If its output anomaly judgment result is bimodal or dragged, the cross-satellite residual common mode branch is selected.

[0082] The data in the relevant domain and the navigation solution domain are independent of the radio frequency domain used by the initial convolutional model. When two independent data sources jointly support the deception candidate, the verified processing result is output, along with the source mask.

[0083] In another embodiment, the loss probability detection result of the navigation device can be the processing result of a convolutional verification branch or an existing result source. If the processing result of the current candidate hypothesis data is to suppress interference, and the abnormal situation of increased loss probability only occurs in a single satellite at a low elevation angle and no cross-satellite residual common mode appears, then the processing result is used to reject the candidate hypothesis data of "global suppression" and does not directly trigger recapture.

[0084] In yet another embodiment, the transition type, probability, or quality level of the carrier phase transition detection output can be encapsulated as an evidence token for the tracking domain.

[0085] For example, the half-cycle slip result of a single satellite can be used to explain the carrier phase change of that satellite, but it cannot explain the clock drift of receivers synchronized with multiple satellites. Therefore, it is still necessary to select the cross-satellite residual common mode branch for processing.

[0086] like Figure 4 As shown, in some optional embodiments, the feature snapshot data also includes data epoch identifiers and data generation identifiers. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenges also includes: Step S210: Obtain the correlator tap interval, coherent integration length, incoherent integration length, tracking mode, and navigation filter reference data of the navigation device.

[0087] Step S220: When at least one of the correlator tap interval, coherent integration length, incoherent integration length, tracking mode, and navigation filter reference data changes, update the data epoch identifier of the feature snapshot data.

[0088] Step S230: After obtaining the processing result, generate the corresponding evidence token, which includes a data epoch identifier and a data generation identifier.

[0089] Step S240: Compare the data epoch identifier and data generation identifier of the updated feature snapshot data with the data epoch identifier and data generation identifier of the evidence token. If they are consistent, verify the processing result. If they are inconsistent, generate invalidation information for the evidence token.

[0090] In this embodiment, the evidence token returned by the convolutional verification branch includes at least the branch identifier, data epoch identifier, data generation identifier, model version, source information, category identifier, confidence level, and fault flag. If the data epoch identifier, data generation identifier, or model version of the evidence token does not match the feature snapshot data, the evidence token is discarded, and invalidation information for the evidence token is generated, recording the corresponding rollback reason. Specifically, the data generation identifier is updated each time a new observation epoch is entered.

[0091] Furthermore, in this embodiment, the preset resolution condition is that the processing result includes processed data for each contradiction in the contradiction vector information, and the independent source quantity condition is that the processing result includes feature snapshot data from at least two different GNSS observation domains. Specifically, the physical links of the at least two different GNSS observation domain data are independent, and they are not simultaneously obtained from the same original observation domain data or the same convolutional model.

[0092] like Figure 8 As shown, the steps for verifying the processing results include: Step S250: Based on the evidence token, verify whether the processing result meets the preset resolution conditions and the independent source quantity conditions. If both are met, output the verified processing result.

[0093] Step S260: If the preset resolution condition or the independent source quantity condition is not met, a new target branch is selected from the uncalled convolutional verification branches to obtain a new processing result and a new evidence token, and verification is performed again until both conditions are met, or if the two conditions are still not met after the preset submission time, indeterminate state information is generated. The preset resolution condition is to clear existing contradictions.

[0094] When the independent source and preset resolution conditions are met, the processing result is output. When the verification result is negative, the candidate hypothesis data is recycled. After the preset submission time is exceeded, uncertain state information is generated, the data becomes invalid, and processing of the candidate hypothesis data is prohibited.

[0095] like Figure 9 As shown, in this embodiment, the submission status of the processing result includes a temporary state, a verified state, a rejected state, and an uncertain state. The output of the initial convolutional model is in the temporary state.

[0096] The processing result is converted to a verified state only when unresolved contradictions are resolved and at least two independent GNSS observation domain data sources provide consistent support for the candidate hypothesis data. When the processing result of the convolutional verification branch negates the candidate hypothesis data, the result can be converted to a negated state, and new candidate hypothesis data can be established. An uncertain state is generated after a preset submission time, if the valid data source is uncertain, the version is inconsistent, or there is a hardware malfunction.

[0097] In this embodiment, the uncertain state is the submission control state. Subsequent processing modules, based on the uncertain state, may downgrade the processing result, maintain the previous trusted state, or request higher-level processing. Only when the processing result is in a verified state is it allowed to be provided to subsequent processing modules for navigation calculation, observation removal, recapture, or alarm.

[0098] Specifically, in this embodiment, the physical links of at least two different GNSS observation domain data are independent, and they are not simultaneously obtained from the same original observation domain data or the same convolutional model. For example, the independent source quantity condition is that at least two GNSS observation domains, predefined as physically link independent, support candidate hypothesis data. These two GNSS observation domain data are not obtained from intermediate features of the same convolutional model or the same original observation domain data, to avoid the model repeatedly voting and forming false high confidence levels.

[0099] The initial convolutional model in this embodiment can be a one-dimensional convolutional neural network, a two-dimensional convolutional neural network, a depthwise separable convolutional network, or other convolutional networks that include convolution operations.

[0100] The convolutional verification branch scheduling method based on the GNSS cross-domain physical consistency challenge provided in this embodiment of the invention includes: acquiring feature snapshot data, wherein the feature snapshot data includes at least two different GNSS observation domain data within the same observation epoch; calling the initial convolutional model to process the first GNSS observation domain data to generate candidate hypothesis data in a temporary state; acquiring a hypothesis observation constraint template corresponding to the candidate hypothesis data, wherein the hypothesis observation constraint template includes the physical constraint relationship between the candidate hypothesis data and at least one second GNSS observation domain data different from the first GNSS observation domain data; and, based on the hypothesis observation constraint template, configuring the second GNSS observation domain data with the physical constraint relationship. The system compares the constraints and generates contradiction vectors to identify candidate hypothesis data that do not meet physical constraints. It then obtains branch descriptors for multiple convolutional validation branches in the initial convolutional model. Each branch descriptor includes a discriminant code for the contradiction type of each validation branch and a source code representing the GNSS observation domain to which the input data received by each validation branch belongs. Based on the matching degree between the contradiction vector information and the discriminant code of each validation branch, and the matching degree between the corresponding source code and the source code of the invoked validation branches, the system selects the most matching target branch from the uninvoked validation branches. Finally, it obtains the processing results of the candidate hypothesis data through the target branch.

[0101] The convolutional verification branch scheduling method based on the GNSS cross-domain physical consistency challenge provided in this invention uses the output of the initial convolutional model as a candidate hypothesis and generates a contradiction vector based on the hypothesis observation constraint template. Then, it selects the convolutional verification branch as needed according to the discriminant code and the source code. This avoids the situation where the strength of GNSS signal is directly equated with the model complexity, which leads to the selection of the wrong convolutional path. This ensures the interpretability and completeness of the model output results and improves the credibility of the model output results by using data from different observation domains. At the same time, it eliminates the need for the model to call additional branches to process redundant data, thereby reducing the task pressure and power consumption of navigation equipment.

[0102] In addition to the above method embodiments, the present invention also provides, for example, Figure 10 and Figure 11 The navigation chip shown includes: a snapshot controller 201, a first convolution register 202, a physical consistency check generator 203, a branch descriptor memory 204, a verification branch selector 205, and an output controller 206.

[0103] The snapshot controller 201 is used to acquire feature snapshot data of at least two different GNSS observation domains within the same observation epoch.

[0104] The initial convolution register 202 is used to store and call the initial convolution model to process the feature snapshot data of the first GNSS observation domain and generate candidate hypothesis data in a temporary state.

[0105] The physical consistency test generator 203 is used to obtain the hypothesis observation constraint template corresponding to the candidate hypothesis data. The hypothesis observation constraint template includes the physical constraint relationships of the candidate hypothesis data in at least one GNSS observation domain different from the first GNSS observation domain, and performs consistency tests on the feature snapshot data of the second GNSS observation domain based on the hypothesis observation constraint template to generate contradictory vector information in the candidate hypothesis data that does not satisfy the physical constraint relationships.

[0106] The branch descriptor memory 204 is used to store and retrieve branch descriptors for multiple convolutional validation branches in the initial convolutional model. Each branch descriptor includes a discriminant code for the conflict type corresponding to each convolutional validation branch, and a source code indicating the GNSS observation domain to which the input data that each convolutional validation branch can receive belongs.

[0107] The verification branch selector 205 is used to select the most matching target branch from the uncalled convolutional verification branches based on the degree of matching between the contradiction vector information and the discriminant code of each convolutional verification branch, and the degree of matching between the corresponding source code and the source code of the called convolutional verification branch.

[0108] The output controller 206 is used to obtain the processing results of the candidate hypothesis data through the target branch and output the processing results.

[0109] In this embodiment, feature snapshot data can be stored in on-chip static random access memory, and tributary descriptors can be stored in read-only memory or verifiable configuration random access memory.

[0110] In the initial convolutional model, each convolutional validation branch can share the convolutional computation engine, and the model weights, input feature addresses, and output evidence tokens of each convolutional validation branch are indexed by the branch identifier.

[0111] If the model version, input scale identifier, or data generation identifier in the evidence token do not match, the evidence token will be discarded and the rollback reason will be set before evidence fusion or submission for verification.

[0112] For example, in one embodiment, the initial convolutional model outputs candidate hypothesis data for "broadband suppression of interference" based on the fast Fourier transform spectrum. The physical consistency test generator finds that the automatic gain control state has decreased, but the spectrum occupancy has not been broadened according to the hypothesis observation constraint template, and the pseudorange residuals of multiple satellites are in a common mode. Therefore, the spectrum in the contradiction vector information is matched with the automatic gain control contradiction bit and the cross-satellite residual common mode contradiction bit.

[0113] The verification branch selector selects the relevant domain morphology branch whose source is the relevant domain. If its output anomaly judgment result is bimodal or dragging, then the cross-satellite residual common mode branch is selected.

[0114] The data in the relevant domain and the navigation solution domain are independent of the radio frequency domain used by the initial convolutional model. When two independent data sources jointly support the deception candidate, the verified processing result is output, along with the source mask.

[0115] In another embodiment, the loss probability detection result of the navigation device can be the processing result of a convolutional verification branch or an existing result source. If the processing result of the current candidate hypothesis data is to suppress interference, and the abnormal situation of increased loss probability only occurs in a single satellite at a low elevation angle and no cross-satellite residual common mode appears, then the processing result is used to reject the candidate hypothesis data of "global suppression" and does not directly trigger recapture.

[0116] In yet another embodiment, the transition type, probability, or quality level of the carrier phase transition detection output can be encapsulated as an evidence token for the tracking domain.

[0117] For example, the half-cycle slip result of a single satellite can be used to explain the carrier phase change of that satellite, but it cannot explain the clock drift of receivers synchronized with multiple satellites. Therefore, it is still necessary to select the cross-satellite residual common mode branch for processing.

[0118] The navigation chip provided in this embodiment of the invention implements the above steps, using the output of the initial convolutional model as a candidate hypothesis, and generating a contradiction vector based on the hypothesis observation constraint template. Then, it selects the convolutional verification branch as needed according to the discriminant code and the source code, thereby avoiding the situation where the strength of GNSS signal is directly equated with the model complexity, which leads to the selection of the wrong convolutional path. This ensures the interpretability and completeness of the model output results, and improves the credibility of the model output results through data from different observation domains. At the same time, it eliminates the need for the model to call additional branches to process redundant data, thereby reducing the task pressure and power consumption of the navigation device.

[0119] In addition to the above method embodiments, the present invention also provides, for example, Figure 12 An electronic device is shown, comprising a processor 301 and a memory 302, wherein the memory 302 stores instructions. The processor 301 is capable of calling the instructions in the memory 302 to execute the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to any of the above embodiments of the present invention.

[0120] The convolutional verification branch scheduling method based on the GNSS cross-domain physical consistency challenge provided in the above embodiments of the present invention includes: acquiring feature snapshot data, wherein the feature snapshot data includes at least two different GNSS observation domain data within the same observation epoch; calling the initial convolutional model to process the first GNSS observation domain data to generate candidate hypothesis data in a temporary state; acquiring a hypothesis observation constraint template corresponding to the candidate hypothesis data, wherein the hypothesis observation constraint template includes the physical constraint relationship between the candidate hypothesis data and at least one second GNSS observation domain data different from the first GNSS observation domain data; and, based on the hypothesis observation constraint template, connecting the second GNSS observation domain data with the physical... Constraint relationships are compared to generate contradiction vector information to identify candidate hypothesis data that does not satisfy physical constraints. Branch descriptors for multiple convolutional validation branches in the initial convolutional model are obtained. Each branch descriptor includes a discriminant code for the contradiction type corresponding to each convolutional validation branch and a source code indicating the GNSS observation domain to which the input data that each convolutional validation branch can receive belongs. Based on the matching degree between the contradiction vector information and the discriminant code of each convolutional validation branch, and the matching degree between the corresponding source code and the source code of the invoked convolutional validation branches, the most matching target branch is selected from the uninvoked convolutional validation branches. The processing result of the candidate hypothesis data is obtained through the target branch.

[0121] The electronic device provided in this embodiment of the invention implements the above method, using the output of the initial convolutional model as a candidate hypothesis, and generating a contradiction vector based on the hypothesis observation constraint template. Then, it selects the convolutional verification branch as needed according to the discriminant code and the source code, thereby avoiding the situation where the strength of GNSS signal is directly equated with the model complexity, which leads to the selection of the wrong convolutional path. This ensures the interpretability and completeness of the model output results, and improves the credibility of the model output results through data from different observation domains. At the same time, it eliminates the need for the model to call additional branches to process redundant data, thereby reducing the task pressure and power consumption of the navigation device.

[0122] Furthermore, the electronic device provided in the embodiments of the present invention may also include a communication interface 303 and a bus 304, wherein the processor 301, the memory 302 and the communication interface 303 are electrically connected through the bus 304.

[0123] The memory 302 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 303 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 304 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 12 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0124] Processor 301 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 301 or by instructions in software form. Processor 301 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a readily available storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 302. The processor 301 reads the information from memory 302 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0125] This invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the above-described convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge.

[0126] The computer-readable storage medium provided in this embodiment of the invention stores data and computer-executable instructions for the above-mentioned convolutional verification branch scheduling method based on the GNSS cross-domain physical consistency challenge. The above-mentioned convolutional verification branch scheduling method based on the GNSS cross-domain physical consistency challenge includes: acquiring feature snapshot data, wherein the feature snapshot data includes at least two different GNSS observation domain data within the same observation epoch; calling the initial convolutional model to process the first GNSS observation domain data to generate candidate hypothesis data in a temporary state; acquiring the hypothesis observation constraint template corresponding to the candidate hypothesis data, wherein the hypothesis observation constraint template includes the physical constraint relationship of the candidate hypothesis data in at least one second GNSS observation domain data different from the first GNSS observation domain data; Based on the hypothetical observation constraint template, the second GNSS observation domain data is compared with the physical constraint relationship to generate contradiction vector information used to identify candidate hypothesis data that does not satisfy the physical constraint relationship. Branch descriptors for multiple convolutional verification branches in the initial convolutional model are obtained. Each branch descriptor includes a discriminant code for the contradiction type corresponding to each convolutional verification branch, and a source code representing the GNSS observation domain to which the input data that each convolutional verification branch can receive belongs. Based on the matching degree between the contradiction vector information and the discriminant code of each convolutional verification branch, and the matching degree between the corresponding source code and the source code of the invoked convolutional verification branches, the most matching target branch is selected from the uninvoked convolutional verification branches. The processing result of the candidate hypothesis data is obtained through the target branch.

[0127] The computer-readable storage medium provided in this embodiment of the invention implements the above method, using the output of the initial convolutional model as a candidate hypothesis, and generating a contradiction vector based on the hypothesis observation constraint template. Then, it selects the convolutional verification branch as needed according to the discriminant code and the source code, thereby avoiding the situation where the strength of GNSS signal is directly equated with the model complexity, which leads to the selection of the wrong convolutional path. This ensures the interpretability and completeness of the model output results, and improves the credibility of the model output results by using data from different observation domains. At the same time, it eliminates the need for the model to call additional branches to process redundant data, thereby reducing the task pressure and power consumption of the navigation device.

[0128] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0129] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0130] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge, characterized in that, The method includes: Acquire feature snapshot data, wherein the feature snapshot data includes at least two different GNSS observation domain data within the same observation epoch; The initial convolutional model is invoked to process the data in the first GNSS observation domain, generating candidate hypothesis data in a provisional state; Obtain the hypothesis observation constraint template corresponding to the candidate hypothesis data, wherein the hypothesis observation constraint template includes the physical constraint relationship of the candidate hypothesis data in at least one second GNSS observation domain data that is different from the first GNSS observation domain data; Based on the hypothetical observation constraint template, the second GNSS observation domain data is compared with the physical constraint relationship to generate contradiction vector information used to identify the candidate hypothetical data that does not satisfy the physical constraint relationship; Obtain branch descriptors for multiple convolutional verification branches in the initial convolutional model, wherein the branch descriptor includes a discriminant code for the contradiction type corresponding to each convolutional verification branch, and a source code for indicating the GNSS observation domain to which the input data that each convolutional verification branch can receive belongs; Based on the degree of matching between the contradiction vector information and the discriminant code of each convolutional verification branch, and the degree of matching between the corresponding source code and the source code of the invoked convolutional verification branch, the most matching target branch is selected from the uninvoked convolutional verification branches. The processing results of the candidate hypothesis data are obtained through the target branch.

2. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 1, characterized in that, The GNSS observation domain of the feature snapshot data includes at least two of the radio frequency domain, correlation domain, tracking domain, and navigation solution domain; The data in the radio frequency domain includes the fast Fourier transform spectrum and / or automatic gain control status; the data in the correlation domain includes multi-tap correlation values ​​and / or correlation peak shapes; the data in the tracking domain includes code carrier difference and / or Doppler frequency shift difference; and the data in the navigation calculation domain includes cross-satellite residual common mode data and / or receiver clock drift signal.

3. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 2, characterized in that, The multiple convolutional validation branches in the initial convolutional model include: The correlation domain morphology branch is used to identify whether the multi-tap correlation value and / or the correlation peak morphology are abnormal, so as to generate an anomaly judgment result; The cross-satellite residual common mode branch is used to identify whether there is a common mode in the pseudorange or carrier phase residuals between multiple satellites based on the cross-satellite residual common mode data, and / or to identify whether drift occurs based on the receiver clock drift signal, so as to generate drift data; The tracking domain consistency branch is used to obtain the code pseudorange change rate based on the code carrier difference and / or the carrier phase consistency result based on the Doppler frequency shift difference value; The spectrum consistency branch is used to obtain the spectrum state of the feature snapshot data based on the fast Fourier transform spectrum shape and / or automatic gain control state.

4. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 3, characterized in that, The candidate hypothesis data is any one of the anomaly judgment result, the drift data, the carrier phase consistency result, and the spectrum state.

5. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 2, characterized in that, The feature snapshot data also includes data epoch identifiers and data generation identifiers, and the method further includes: Acquire the correlator tap interval, coherent integration length, incoherent integration length, tracking mode, and navigation filter reference data of the navigation device; When at least one of the correlator tap interval, the coherent integration length, the incoherent integration length, the tracking mode, and the navigation filter reference data changes, the data epoch identifier of the feature snapshot data is updated; After obtaining the processing result, a corresponding evidence token is generated, which includes a data epoch identifier and a data generation identifier. The data epoch identifier and data generation identifier of the updated feature snapshot data are compared with the data epoch identifier and data generation identifier of the evidence token. If they are consistent, the processing result is verified. If they are inconsistent, invalidation information of the evidence token is generated.

6. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 5, characterized in that, The steps for verifying the processing result include: Based on the evidence token, verify whether the processing result meets the preset resolution conditions and the independent source quantity conditions. If both are met, output the verified processing result. If the preset resolution condition or the independent source quantity condition is not met, a new target branch is selected from the never-called convolutional verification branches to obtain a new processing result and a new evidence token, and verification is performed again until both conditions are met, or If the two conditions are not met after the preset submission time, an uncertain status message is generated.

7. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 6, characterized in that, The preset resolution condition is that the processing result includes the processing data of each contradiction in the contradiction vector information; The independent source quantity condition is that the processing result includes at least two different GNSS observation domain data, wherein the physical links of the at least two different GNSS observation domain data are independent and are not simultaneously obtained from the same original observation domain data or the same convolution model.

8. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 2, characterized in that, The hypothetical observation constraint template includes multiple preset physical constraint relationships. The step of comparing the second GNSS observation domain data with the physical constraint relationships based on the hypothetical observation constraint template to generate contradiction vector information for identifying candidate hypothetical data that does not satisfy the physical constraint relationships includes: Based on the assumed observation constraint template, the physical constraint relationship and contradiction bit mask corresponding to the second GNSS observation domain data are obtained; Based on the physical constraints, a consistency check is performed on the data in the second GNSS observation domain. When the feature snapshot data in the second GNSS observation domain does not conform to the physical constraints, the bits at the corresponding positions in the contradiction bit mask are set to generate the contradiction vector information.

9. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 8, characterized in that, The physical constraint relationship includes a preset numerical range of the candidate hypothesis data and / or a preset expected direction of change, and different bits in the contradiction bit mask correspond to different physical constraint relationships. When the second GNSS observation domain data is not in the preset value range, and / or when the change direction of the second GNSS observation domain data is inconsistent with the expected change direction after exceeding the dead zone, the bit at the corresponding position in the contradictory bit mask will be set.

10. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 1, characterized in that, The step of selecting the most matching target branch from the uncalled convolutional verification branches based on the matching degree between the contradiction vector information and the discriminant code of each convolutional verification branch, and the matching degree between the corresponding source code and the source code of the invoked convolutional verification branches, includes: If the discriminant type indicated by the discriminant code of the uncalled convolutional verification branch intersects with the discriminant type of the discriminant vector information, and the number of overlapping bits between the corresponding source code and the set of source codes of all called convolutional verification branches does not exceed a preset overlap threshold, then all corresponding convolutional verification branches are considered as candidate convolutional verification branches, or... Based on the source code of the uncalled convolutional verification branch, detect whether the GNSS observation domain data corresponding to the source code is used by the called convolutional verification branch. If it is not used, take all the corresponding convolutional verification branches as the candidate convolutional verification branches. The best-matching convolutional verification branch is selected from all the candidate convolutional verification branches as the target branch.

11. The convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge according to claim 10, characterized in that, The branch descriptor also includes the longest execution cycle and energy consumption parameter data of the corresponding convolutional verification branch. The step of selecting the best-matching convolutional verification branch as the target branch from all the candidate convolutional verification branches includes: Obtain the data recording time from the feature snapshot data; Based on the discriminant code, source code, longest execution cycle, and energy consumption parameter data of each candidate convolutional verification branch, the contradiction coverage of the contradiction vector information, the number of observation domains not used by the called branches, the longest execution cycle, and energy consumption parameter data of the corresponding candidate convolutional verification branch are obtained. Based on the contradiction coverage, the number of observation domains not used by the invoked branches, the data recording time, the longest execution cycle, and the energy consumption parameter data, calculate the ranking score for each candidate convolutional verification branch; The candidate convolutional validation branch with the highest ranking score is selected as the most matching target branch.

12. A navigation chip, characterized in that, The navigation chip includes: A snapshot controller is used to acquire characteristic snapshot data of at least two different GNSS observation domains within the same observation epoch. The initial convolution register is used to store and call the initial convolution model to process the feature snapshot data of the first GNSS observation domain and generate candidate hypothesis data in a temporary state. A physical consistency test generator is used to obtain a hypothesis observation constraint template corresponding to the candidate hypothesis data, wherein the hypothesis observation constraint template includes physical constraint relationships of the candidate hypothesis data in at least one other GNSS observation domain different from the first GNSS observation domain; and based on the hypothesis observation constraint template, to perform a consistency test on the feature snapshot data of the second GNSS observation domain, and generate contradictory vector information in the candidate hypothesis data that does not satisfy the physical constraint relationships; A branch descriptor memory is used to store and call branch descriptors of multiple convolutional verification branches in the initial convolutional model. The branch descriptor includes a discriminant code for the contradiction type corresponding to each convolutional verification branch, and a source code for indicating the GNSS observation domain to which the input data that each convolutional verification branch can receive belongs. A verification branch selector is used to select the most matching target branch from the uncalled convolutional verification branches based on the degree of matching between the contradiction vector information and the discriminant code of each convolutional verification branch, and the degree of matching between the corresponding source code and the source code of the invoked convolutional verification branch. An output controller is used to obtain the processing result of the candidate hypothesis data through the target branch and output the processing result.

13. An electronic device, characterized in that, The electronic device includes: a processor and a memory, wherein the memory stores instructions; The processor invokes the instructions in the memory to cause the electronic device to implement the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge as described in any one of claims 1 to 11.

14. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the convolutional verification branch scheduling method based on GNSS cross-domain physical consistency challenge as described in any one of claims 1 to 11.

Images