Tight coupling fusion positioning method and system applying Beidou high-precision positioning and inertial navigation

Abstract

The invention provides a tight coupling fusion positioning method and system applying Beidou high-precision positioning and inertial navigation, and the method comprises the steps: accessing a Beidou high-precision positioning signal and inertial navigation original data, building a parallel transmission link, and generating an initial coupling link parameter; inputting the initial coupling link parameters into a pre-trained artificial intelligence coupling link parameter iteration module, and outputting optimized coupling link parameters; constructing a feature interaction graph according to the optimized coupling link parameters and generating a feature association weight sequence; constructing a dynamic coupling rule set based on the feature correlation weight sequence, and performing tight coupling processing on the Beidou high-precision positioning signal and the inertial navigation original data; and finally, outputting tight coupling fusion positioning data through an artificial intelligence coupling verification module. According to the invention, internal association between Beidou high-precision positioning and inertial navigation is fully utilized, high-precision and high-reliability positioning is realized, and the method is suitable for positioning requirements in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to a tightly coupled fusion positioning method and system that applies BeiDou high-precision positioning and inertial navigation. Background Technology

[0002] In the current field of positioning technology, single positioning methods often have their own limitations. While BeiDou high-precision positioning technology can provide high-precision location information, satellite signals are easily blocked in complex environments, such as urban canyons with tall buildings or dense forests, leading to signal loss or decreased positioning accuracy. Inertial navigation technology, on the other hand, relies on inertial measurement units (such as accelerometers and gyroscopes) to measure the motion state of objects and achieve autonomous navigation. It has advantages such as not relying on external signals and providing continuous navigation information. However, inertial navigation suffers from the problem of error accumulation over time, with positioning errors gradually increasing after prolonged use. To overcome the shortcomings of single positioning methods, existing technologies have developed methods that fuse BeiDou high-precision positioning with inertial navigation. However, most common fusion positioning methods currently employ a loosely coupled approach, processing BeiDou high-precision positioning data and inertial navigation data separately and then simply fusing the results. This loosely coupled approach cannot fully utilize the inherent correlation between the two types of positioning data, resulting in limited fusion effects and failing to meet the requirements for high-precision and high-reliability positioning. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a tightly coupled fusion positioning method applying BeiDou high-precision positioning and inertial navigation, the method comprising: By accessing BeiDou high-precision positioning signals and raw inertial navigation data, a parallel transmission link for BeiDou high-precision positioning signals and raw inertial navigation data is established, and initial coupled link parameters are generated through link adaptation processing. The initial coupling link parameters are input into the pre-trained artificial intelligence coupling link parameter iteration module, which outputs optimized coupling link parameters based on the inherent correlation between BeiDou high-precision positioning signals and inertial navigation raw data. Based on the optimized coupling link parameters, a feature interaction map of BeiDou high-precision positioning signal and inertial navigation raw data is constructed, and a feature association weight sequence is generated by parsing the feature interaction map; A dynamic coupling rule set is constructed based on the feature association weight sequence. The Beidou high-precision positioning signal and the raw inertial navigation data are input into the dynamic coupling rule set for tight coupling processing. The tightly coupled fusion positioning data is output through the artificial intelligence coupling verification module. The tightly coupled fusion positioning data carries the accuracy characteristics of Beidou high-precision positioning and the continuous characteristics of inertial navigation.

[0004] In another aspect, embodiments of the present invention also provide a tightly coupled fusion positioning system that applies BeiDou high-precision positioning and inertial navigation, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or code. The processor is used to execute the programs, instructions or code in the machine-readable storage medium to implement the above-mentioned method.

[0005] Based on the above, this embodiment of the invention establishes a parallel transmission link and generates initial coupling link parameters through adaptive processing, laying a solid foundation for subsequent fusion processing and ensuring the stability and synchronization of data transmission. Utilizing a pre-trained artificial intelligence coupling link parameter iteration module, optimized coupling link parameters are output based on the inherent correlation between BeiDou high-precision positioning signals and inertial navigation raw data. This allows for in-depth exploration of the potential connections between the two types of data, achieving precise parameter optimization. A feature interaction map is constructed based on the optimized coupling link parameters, generating a feature association weight sequence, clearly presenting the degree of correlation and importance between the two types of data features. A dynamic coupling rule set is constructed based on the feature association weight sequence for tight coupling processing. This allows for dynamic adjustment of the coupling method according to real-time data, fully leveraging the accuracy advantage of BeiDou high-precision positioning and the continuity advantage of inertial navigation. The final tightly coupled fused positioning data carries the advantages of both positioning methods, significantly improving positioning accuracy and reliability. It can provide stable and accurate positioning information even in complex environments, possessing broad application prospects and significant practical value. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of the execution flow of the tightly coupled fusion positioning method of BeiDou high-precision positioning and inertial navigation provided in the embodiments of the present invention.

[0007] Figure 2 This is a schematic diagram of exemplary hardware and software components of a tightly coupled fusion positioning system that applies BeiDou high-precision positioning and inertial navigation, provided in an embodiment of the present invention. Detailed Implementation

[0008] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation, provided in one embodiment of the present invention. The following is a detailed description of this tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation.

[0009] Step S110: Access the BeiDou high-precision positioning signal and the raw inertial navigation data, establish a parallel transmission link for the BeiDou high-precision positioning signal and the raw inertial navigation data, and generate the initial coupled link parameters through link adaptation processing.

[0010] A parallel transmission link is established between the BeiDou high-precision positioning signal and the raw inertial navigation data. Initial coupling link parameters are generated through link adaptation processing. The BeiDou high-precision positioning signal, containing positioning coordinates and signal strength, is received by the BeiDou receiving module; the raw inertial navigation data, including acceleration and angular velocity, is output by the inertial measurement module. The parallel transmission link must meet the synchronous transmission requirements of both data streams. Link adaptation processing adjusts for issues such as latency and bandwidth during transmission. The generated initial coupling link parameters are used in subsequent parameter optimization stages.

[0011] In step S110, before accessing the BeiDou high-precision positioning signal and inertial navigation raw data, the legal and regulatory requirements for data collection and authorization must be strictly followed. The collection of BeiDou high-precision positioning signals must be conducted through a legally qualified BeiDou receiving module. The module must comply with the relevant technical specifications and usage licenses of the BeiDou Satellite Navigation System. Before collection, it must be ensured that the BeiDou receiving module has completed equipment registration and frequency license application. The relevant registration information is stored in the module's built-in license storage unit and can be verified through command queries. The collection of inertial navigation raw data must clearly define the purpose and scope of data collection. Only motion data related to positioning fusion, such as acceleration and angular velocity, should be collected; other sensitive data unrelated to positioning should not be collected. Before collection, the purpose, scope, storage method, and usage period of data collection must be displayed to the user through the system settings interface of the carrier device, and explicit authorization from the user must be obtained. The authorization information is stored in encrypted form in the secure storage area of ​​the carrier device. The authorization record includes the authorization time, authorization content, and user identifier, and is traceable and verifiable. During data transmission, both BeiDou high-precision positioning signals and raw inertial navigation data employ encrypted transmission protocols. Communication content within the transmission link is processed using symmetric encryption algorithms, and encryption keys are generated through a secure key negotiation mechanism and updated periodically to prevent data theft or tampering during transmission. For data storage, the collected BeiDou high-precision positioning signals and raw inertial navigation data are stored in storage devices that meet national information security level protection requirements. These storage devices have access control, data encryption, and auditing functions. Only authorized modules and processes can access the stored data, and the storage period strictly adheres to the user-authorized usage period, automatically deleting data upon expiration. During data usage, the collected BeiDou high-precision positioning signals and raw inertial navigation data are used solely for tightly coupled fusion positioning purposes and not for any other unauthorized uses. Data usage records are meticulously maintained through a log system, including the data usage time, the module used, and the purpose of use. Log information is backed up periodically for monitoring and auditing. These measures ensure that the collection, transmission, storage, and use of BeiDou high-precision positioning signals and raw inertial navigation data comply with relevant laws and regulations, protecting users' data security and privacy rights.

[0012] Step S111: Access the BeiDou high-precision positioning signal through the BeiDou receiving module and access the raw inertial navigation data through the inertial measurement module.

[0013] The system receives high-precision BeiDou positioning signals via a BeiDou receiver module and raw inertial navigation data via an inertial measurement module. The BeiDou receiver module uses equipment supporting the BeiDou-3 global satellite navigation system, receiving positioning signals from multiple BeiDou satellites and outputting high-precision BeiDou positioning signals containing information such as timestamps, positioning coordinates, and signal-to-noise ratio. The inertial measurement module includes a three-axis accelerometer and a three-axis gyroscope, acquiring real-time acceleration and angular velocity data of the carrier and outputting raw inertial navigation data containing information such as sampling time, acceleration components, and angular velocity components.

[0014] Step S112: Establish a dual-path parallel transmission link, configure the transmission protocol and data interaction timing of the dual-path parallel transmission link, and realize the synchronous transmission of BeiDou high-precision positioning signals and inertial navigation raw data.

[0015] A dual-path parallel transmission link was established, and its transmission protocol and data interaction timing were configured to achieve synchronous transmission of BeiDou high-precision positioning signals and inertial navigation raw data. The dual-path parallel transmission link uses independent physical transmission channels to carry the BeiDou high-precision positioning signals and inertial navigation raw data respectively. A high-reliability data transmission protocol was selected, and parameters such as data frame format and verification method were configured. The data interaction timing was achieved through a clock synchronization mechanism to ensure that the sampling time deviation between the two data streams is controlled within a preset range, meeting the requirements for synchronous transmission.

[0016] Step S113: Extract the link attribute information of the dual-path parallel transmission link. The link attribute information includes transmission delay characteristics, bandwidth occupancy characteristics, and signal attenuation characteristics.

[0017] Step S113: Extract the link attribute information of the dual-path parallel transmission link. The link attribute information includes transmission delay characteristics, bandwidth occupancy characteristics, and signal attenuation characteristics.

[0018] Step S114: Input the link attribute information into the link adaptation processing module. The link adaptation processing module outputs the link adaptation adjustment scheme, which covers transmission timing calibration and bandwidth allocation ratio.

[0019] The link attribute information is input into the link adaptation processing module, which then outputs a link adaptation adjustment scheme. This scheme covers transmission timing calibration and bandwidth allocation ratio. The link adaptation processing module incorporates a built-in link attribute analysis algorithm to analyze the input transmission delay characteristics, bandwidth occupancy characteristics, and signal attenuation characteristics. It outputs timing calibration parameters for transmission timing deviations and bandwidth allocation ratio adjustment parameters for uneven bandwidth occupancy, thus forming a complete link adaptation adjustment scheme.

[0020] Step S115: Optimize the dual-path parallel transmission link according to the link adaptation adjustment scheme, and collect the transmission parameters of the optimized dual-path parallel transmission link. The transmission parameters include synchronization error characteristics and data integrity characteristics.

[0021] The dual-path parallel transmission link was optimized according to the link adaptation and adjustment scheme. Transmission parameters of the optimized dual-path parallel transmission link were collected, including synchronization error characteristics and data integrity characteristics. The optimization process involved adjusting the link's clock synchronization mechanism according to the timing calibration parameters in the link adaptation and adjustment scheme, and adjusting the link's bandwidth usage according to the bandwidth allocation ratio. The optimized synchronization error characteristics were obtained by comparing the timestamp differences between the two data streams, and the data integrity characteristics were obtained by statistically analyzing the proportion of data frames lost during transmission to the total number of data frames.

[0022] Step S116: Construct an initial coupled link parameter set based on transmission parameters. The initial coupled link parameter set integrates the association information between the link adaptation and adjustment scheme and the transmission parameters.

[0023] An initial coupled link parameter set is constructed based on transmission parameters. This initial coupled link parameter set integrates the correlation information between the link adaptation and adjustment scheme and the transmission parameters. The construction of the initial coupled link parameter set is achieved through a parameter correlation algorithm, which associates the timing calibration parameters and bandwidth allocation ratio in the link adaptation and adjustment scheme with the synchronization error characteristics and data integrity characteristics in the transmission parameters, forming an initial coupled link parameter set containing parameter names, parameter values, and parameter correlation relationships.

[0024] Step S117: Perform parameter encoding processing on the initial coupled link parameter set to generate standardized initial coupled link parameters. The standardization process adopts the general encoding specification for link parameters.

[0025] The initial coupled link parameter set is encoded to generate standardized initial coupled link parameters. The standardization process adopts a general encoding standard for link parameters. Following the encoding rules in the general encoding standard, each parameter in the initial coupled link parameter set is converted into a uniformly formatted encoded value. The encoded value includes a parameter type identifier, a parameter length identifier, and a parameter content identifier. The generated standardized initial coupled link parameters are used as subsequent input to the artificial intelligence coupled link parameter iteration module.

[0026] Step S120: Input the initial coupling link parameters into the pre-trained artificial intelligence coupling link parameter iteration module. The artificial intelligence coupling link parameter iteration module outputs optimized coupling link parameters based on the inherent correlation between Beidou high-precision positioning signals and inertial navigation raw data.

[0027] The initial coupled link parameters are input into a pre-trained AI coupled link parameter iteration module. This module outputs optimized coupled link parameters based on the inherent correlation between BeiDou high-precision positioning signals and raw inertial navigation data. The AI ​​coupled link parameter iteration module employs a deep neural network structure, including an input layer, a correlation mining layer, an iterative adjustment layer, and an output layer. The pre-training process uses a large number of BeiDou high-precision positioning signal and raw inertial navigation data samples to learn the inherent correlation between the two. After inputting the initial coupled link parameters, the module extracts the correlation features between the parameters through the correlation mining layer, and the iterative adjustment layer iterates and optimizes the parameters, ultimately outputting the optimized coupled link parameters.

[0028] Step S121: The input layer of the AI ​​coupling link parameter iteration module receives the initial coupling link parameters, performs feature parsing on the initial coupling link parameters, and decomposes them to obtain the basic link parameters and associated adaptation parameters.

[0029] The input layer of the AI-coupled link parameter iteration module receives initial coupled link parameters, performs feature parsing on these parameters, and decomposes them into basic link parameters and associated adaptation parameters. The input layer includes a parameter receiving interface and a feature parsing unit. The parameter receiving interface adapts to the standardized format of the initial coupled link parameters. The feature parsing unit decomposes the initial coupled link parameters into basic link parameters and associated adaptation parameters based on the parameter type identifier. The basic link parameters include fundamental link attribute parameters such as transmission delay and bandwidth usage, while the associated adaptation parameters include adaptation adjustment parameters such as timing calibration and bandwidth allocation.

[0030] Step S122: The parsed link basic parameters and associated adaptation parameters enter the association mining layer of the artificial intelligence coupled link parameter iteration module. The association mining layer calls the pre-trained association feature extraction submodule to mine the inherent association rules between Beidou high-precision positioning signals and inertial navigation raw data.

[0031] The parsed link basic parameters and associated adaptation parameters enter the association mining layer of the AI-coupled link parameter iteration module. The association mining layer calls a pre-trained association feature extraction submodule to uncover the inherent correlation patterns between the BeiDou high-precision positioning signal and the original inertial navigation data. The association mining layer includes a feature mapping unit and an association analysis unit. The feature mapping unit maps the link basic parameters and associated adaptation parameters into feature vectors. The association analysis unit calls the association feature extraction submodule to perform association analysis on the feature vectors, uncovering the inherent correlation patterns between the positioning coordinate features of the BeiDou high-precision positioning signal and the motion trajectory features of the original inertial navigation data, as well as the signal stability features of the BeiDou high-precision positioning signal and the motion attitude features of the original inertial navigation data.

[0032] Step S1221: The associated feature extraction submodule loads the pre-trained feature extraction weights and receives BeiDou high-precision positioning signal samples and inertial navigation raw data samples.

[0033] The associated feature extraction submodule loads pre-trained feature extraction weights and receives BeiDou high-precision positioning signal samples and inertial navigation raw data samples. The associated feature extraction submodule adopts a convolutional neural network structure, and the pre-trained feature extraction weights are obtained through training with a large number of samples. The received BeiDou high-precision positioning signal samples contain multiple sets of positioning coordinates and signal strength data, and the inertial navigation raw data samples contain multiple sets of acceleration and angular velocity data.

[0034] Step S1222: Decompose the signal features of the BeiDou high-precision positioning signal sample to obtain positioning coordinate features, signal matching degree features and signal stability features.

[0035] Signal feature decomposition was performed on BeiDou high-precision positioning signal samples to obtain positioning coordinate features, signal matching degree features, and signal stability features. Signal feature decomposition was achieved through feature extraction algorithms. The positioning coordinate features included latitude, longitude, and altitude data from the samples; the signal matching degree features were obtained by comparing the matching degree between the satellite signal and the receiving module; and the signal stability features were obtained by statistically analyzing the variation amplitude of signal strength in the samples.

[0036] Step S1223: Perform data feature decomposition on the original inertial navigation data sample to obtain motion attitude features, motion trajectory features and motion acceleration features.

[0037] The raw inertial navigation data samples are decomposed into motion attitude features, motion trajectory features, and motion acceleration features. The data feature decomposition is achieved through a feature extraction algorithm. Motion attitude features are calculated from the angular velocity data in the samples; motion trajectory features are obtained from the integral acceleration data; and motion acceleration features include the three-axis acceleration components in the samples.

[0038] Step S1224: Construct feature pairing units, pairing positioning coordinate features with motion trajectory features, and pairing signal stability features with motion posture features to form a feature pairing set.

[0039] Feature pairing units are constructed to pair positioning coordinate features with motion trajectory features, and signal stability features with motion attitude features, forming a feature pairing set. The feature pairing units pair features based on their physical meaning. Positioning coordinate features and motion trajectory features both reflect the position information of the carrier, while signal stability features and motion attitude features both reflect the stability of the carrier's motion state. After pairing, a feature pairing set containing multiple feature pairs is formed.

[0040] Step S1225: Perform correlation analysis on the feature pairing set to generate a feature correlation sequence. The correlation analysis adopts a correlation evaluation algorithm based on sample training.

[0041] A correlation analysis is performed on the feature pair set to generate a feature correlation sequence. The correlation analysis employs a sample-trained correlation evaluation algorithm. The algorithm calculates the correlation between feature pairs by comparing their changing trends. The correlation value reflects the closeness of the association between feature pairs, and the generated feature correlation sequence contains the correlation values ​​for each feature pair.

[0042] Step S1226: Extract feature pairs in the feature correlation sequence that have reached a preset matching degree, and construct the basic framework of the correlation model based on the changing pattern of the feature pairs that have reached the preset matching degree.

[0043] Feature pairs that achieve a preset matching degree in the feature correlation sequence are extracted. Based on the changing patterns of these feature pairs, a basic framework for the correlation model is constructed. The preset matching degree is set according to the correlation of pre-trained samples. After extracting feature pairs that achieve the preset matching degree, the changing patterns of these feature pairs are analyzed to construct a basic framework for the correlation model that includes feature pair correlation rules and descriptions of changing trends.

[0044] Step S1227: Input more BeiDou high-precision positioning signal samples and inertial navigation raw data samples to optimize the basic framework of the correlation model, supplement the indirect correlation information of feature pairs that have not reached the preset matching degree, and finally form a complete description of the internal correlation rules.

[0045] By inputting more BeiDou high-precision positioning signal samples and inertial navigation raw data samples, the basic framework of the correlation model is optimized, and indirect correlation information of feature pairs whose correlation does not reach the preset matching degree is supplemented, ultimately forming a complete description of the inherent correlation rules. By inputting more samples, the correlation rules in the basic framework of the correlation model are verified and adjusted, and indirect correlation information transmitted through intermediate features for feature pairs whose correlation does not reach the preset matching degree is supplemented, forming a complete description of the inherent correlation rules that includes both direct and indirect correlations.

[0046] Step S123: Construct a parameter association model based on the inherent correlation rules, input the basic parameters of the link and the correlation adaptation parameters into the parameter association model, and generate a parameter association matrix.

[0047] A parameter association model is constructed based on inherent correlation patterns. The basic link parameters and associated adaptation parameters are input into the parameter association model to generate a parameter association matrix. The parameter association model adopts a rule-based model structure. The association rules of the model are set according to the mined inherent correlation patterns. After inputting the basic link parameters and associated adaptation parameters into the model, the model outputs a parameter association matrix containing the correlation strength and direction between parameters. The element values ​​in the matrix represent the degree of correlation between corresponding parameter pairs.

[0048] Step S124: The parameter correlation matrix is ​​passed to the iterative adjustment layer of the AI ​​coupling link parameter iteration module. The iterative adjustment layer starts the parameter iteration process. The first iteration calculates the initial parameter deviation value based on the parameter correlation matrix.

[0049] The parameter correlation matrix is ​​passed to the iterative adjustment layer of the AI ​​coupling link parameter iteration module. The iterative adjustment layer initiates the parameter iteration process. The first iteration calculates the initial parameter deviation value based on the parameter correlation matrix. The iterative adjustment layer includes a deviation calculation unit and a parameter adjustment unit. In the first iteration, the deviation calculation unit compares the initial coupling link parameters with the standard parameters in the pre-trained model based on the correlation strength in the parameter correlation matrix, and calculates the initial parameter deviation value. The parameter deviation value reflects the degree of difference between the initial coupling link parameters and the standard parameters.

[0050] Step S125: Generate the first round of parameter adjustment scheme based on the initial parameter deviation value, adjust the initial coupling link parameters, and generate the second round of iteration parameters.

[0051] The initial parameter adjustment scheme is generated based on the initial parameter deviation value. The initial coupling link parameters are adjusted, and the parameters for the second round of iteration are generated. The parameter adjustment unit generates the initial parameter adjustment scheme, which includes the parameter adjustment range and direction, based on the magnitude and direction of the initial parameter deviation value. The initial coupling link parameters are adjusted according to the adjustment scheme to generate the parameters for the second round of iteration. The parameters for the second round of iteration are used in the next round of iteration.

[0052] Step S126: Repeat the parameter deviation calculation, adjustment scheme generation and parameter update process until the deviation value of the iterative parameter meets the preset iteration termination condition of the artificial intelligence coupling link parameter iteration module.

[0053] The process of calculating parameter deviation, generating adjustment schemes, and updating parameters is repeated until the deviation value of the iterative parameters meets the preset iteration termination condition of the AI ​​coupling link parameter iteration module. The iteration termination condition is set as follows: the parameter deviation value is less than a preset threshold and the rate of change of the deviation value is less than a preset proportion for multiple consecutive iterations. During each iteration, the deviation calculation unit calculates the deviation value of the current iteration parameter, the parameter adjustment unit generates the corresponding adjustment scheme, and updates the iteration parameters until the termination condition is met.

[0054] Step S127: The parameters obtained from the final iteration are used as the optimized coupling link parameters and output through the output layer of the artificial intelligence coupling link parameter iteration module. Before output, the optimized coupling link parameters are formatted.

[0055] The parameters obtained from the final iteration are used as the optimized coupled link parameters and output through the output layer of the AI ​​coupled link parameter iteration module. Before output, the optimized coupled link parameters are formatted. The output layer includes a parameter formatting unit and a parameter output interface. The formatting unit adjusts the format of the optimized coupled link parameters according to the general coding standard of link parameters to ensure that the parameter format is consistent with the initial coupled link parameters. The parameter output interface outputs the formatted optimized coupled link parameters to subsequent stages.

[0056] Step S130: Construct a feature interaction map of BeiDou high-precision positioning signal and inertial navigation raw data based on optimized coupling link parameters, and generate feature association weight sequence through feature interaction map parsing.

[0057] Based on optimized coupling link parameters, a feature interaction graph of BeiDou high-precision positioning signals and inertial navigation raw data is constructed. Feature association weight sequences are generated through feature interaction graph parsing. The feature interaction graph represents features as nodes and interactions between features as edges. Feature interaction rules in the optimized coupling link parameters are used to determine the edge connection methods between nodes. Graph parsing calculates the association weights of each feature by traversing the nodes and edges in the graph. The generated feature association weight sequences are used for subsequent construction of dynamic coupling rule sets.

[0058] Step S131: Analyze and optimize the coupling link parameters, extract the feature interaction rules and parameter association thresholds, and use them as the basis for constructing the feature interaction graph.

[0059] The coupling link parameters are analyzed and optimized, and the feature interaction rules and parameter association thresholds are extracted as the basis for constructing the feature interaction graph. The optimization of coupling link parameters is achieved through a parameter analysis algorithm. The extracted feature interaction rules include information such as the interaction type and direction between features; the parameter association threshold includes a minimum threshold for the strength of feature association, and feature interaction relationships below this threshold are not included in the graph construction.

[0060] Step S132: Build a graph node generation unit to convert the core features of the BeiDou high-precision positioning signal and the core features of the original inertial navigation data into feature interaction graph nodes.

[0061] A graph node generation unit is constructed to convert the core features of BeiDou high-precision positioning signals and the core features of inertial navigation raw data into feature interaction graph nodes. The graph node generation unit includes a feature classification module and a node encoding module. The feature classification module classifies the core features of BeiDou high-precision positioning signals (such as positioning coordinates and signal stability) and the core features of inertial navigation raw data (such as motion trajectory and motion attitude). The node encoding module assigns a unique node identifier to each feature, generating feature interaction graph nodes.

[0062] Step S133: Construct the connection relationship of feature interaction graph nodes based on feature interaction rules. Feature interaction graph nodes with the same association type use the same type of connection lines, and nodes with different association types use different lines.

[0063] The feature interaction graph node connection relationships are constructed based on feature interaction rules. Feature interaction graph nodes with the same association type use the same type of connection line, while nodes with different association types use different lines. The interaction types in the feature interaction rules include direct association, indirect association, and feedback association. Direct association is connected with a solid line, indirect association is connected with a dashed line, and feedback association is connected with a solid line with an arrow. The thickness of the connection line reflects the strength of the feature association.

[0064] Step S1331: Analyze the feature interaction rules and classify the feature association types. The feature association types include direct association types, indirect association types, and feedback association types.

[0065] The feature interaction rules are analyzed to classify feature association types, including direct association, indirect association, and feedback association. Feature interaction rule analysis is implemented using a rule analysis algorithm. Direct association refers to features that can interact without the need for intermediate features; indirect association refers to features that need to interact through intermediate features; and feedback association refers to features that influence each other.

[0066] Step S1332: Assign a unique connection line type to different feature association types. Use solid lines for direct association types, dashed lines for indirect association types, and solid lines with arrows for feedback association types.

[0067] Different feature association types are assigned their own unique connection line types: direct association types use solid lines, indirect association types use dashed lines, and feedback association types use solid lines with arrows. The assignment of connection line types is achieved through line mapping rules: direct association types correspond to solid lines, indirect association types correspond to dashed lines, and feedback association types correspond to solid lines with arrows, where the arrow direction indicates the direction of feedback.

[0068] Step S1333: Traverse all BeiDou high-precision positioning signal feature interaction map nodes and inertial navigation raw data feature interaction map nodes, determine the feature association type between any two feature interaction map nodes, and match the corresponding connection line type.

[0069] The algorithm iterates through all feature interaction graph nodes of BeiDou high-precision positioning signals and inertial navigation raw data, determines the feature association type between any two feature interaction graph nodes, and matches the corresponding connecting line type. Node traversal is implemented using a graph traversal algorithm, feature association type determination is based on the interaction type information in the feature interaction rules, and the corresponding connecting line type is matched and lines are drawn between the nodes.

[0070] Step S1334: Connect the feature interaction graph nodes of the directly associated type, and mark the direct association basis on the connection line. The direct association basis comes from the direct association clause in the optimized coupling link parameters.

[0071] For directly related feature interaction graph node connections, the direct association basis is marked on the connection lines. The direct association basis comes from the direct association clauses in the optimized coupling link parameters. The direct association basis marking is implemented through a text annotation algorithm, and the annotation content includes the direct association clause number and core content, which is convenient for reference during subsequent graph analysis.

[0072] Step S1335: Connect the feature interaction graph nodes of the indirect association type, trace the intermediate association feature interaction graph nodes, mark the names of the intermediate association feature interaction graph nodes next to the connecting lines, and clarify the transmission path of indirect association.

[0073] For indirect association type feature interaction graph node connections, the intermediate association feature interaction graph nodes are traced, and the names of the intermediate association feature interaction graph nodes are labeled next to the connecting lines to clarify the transmission path of the indirect association. The tracing of intermediate association feature interaction graph nodes is achieved through a path finding algorithm, and the labeling includes the name and number of intermediate nodes, which facilitates understanding the formation process of indirect associations.

[0074] Step S1336: Connect the feature interaction graph nodes of the feedback association type, mark the feedback direction with the arrow direction, and clarify the interaction direction between the Beidou high-precision positioning signal features and the inertial navigation raw data features.

[0075] For feature interaction graph nodes of feedback association types, the feedback direction is marked with arrows to clarify the interaction direction between BeiDou high-precision positioning signal features and inertial navigation raw data features. The feedback direction marking is achieved through a direction labeling algorithm, and the label content includes "BeiDou to Inertial" or "Inertial to BeiDou", clearly indicating the direction of feedback interaction between features.

[0076] Step S1337: After completing the connection of all feature interaction graph nodes, perform a redundancy check on the connection lines, delete duplicate connection lines, and retain the connection relationships that best reflect the essence of the association.

[0077] After connecting all feature interaction graph nodes, a redundancy check is performed on the connecting lines. Duplicate connecting lines are deleted, and the connections that best reflect the essence of the relationship are retained. The redundancy check is implemented through a line comparison algorithm, which compares the type, direction, and correlation basis of all connecting lines, deletes duplicate lines, and retains the lines with the strongest correlation or those that best reflect the essence of the feature interaction.

[0078] Step S134: Assign the association matching degree information in the optimized coupling link parameters to the connection lines of the feature interaction graph nodes to form a weighted feature interaction graph, where the weight values ​​directly map the association matching degree.

[0079] The association matching degree information from the optimized coupling link parameters is assigned to the connection lines of the feature interaction graph nodes, forming a weighted feature interaction graph. The weight value directly maps to the association matching degree. The association matching degree information is extracted from the optimized coupling link parameters and includes the association strength value of each feature pair. This value is assigned to the corresponding connection line to form a weighted feature interaction graph. The larger the weight value, the stronger the association between features.

[0080] Step S135: Build a graph parsing module. The graph parsing module loads a graph parsing algorithm to traverse and analyze the connection relationships and weight distribution of feature interaction graph nodes in the feature interaction graph.

[0081] A graph analysis module is constructed, which loads a graph analysis algorithm to traverse and analyze the connection relationships and weight distribution of nodes in the feature interaction graph. The graph analysis module includes a graph traversal unit and a weight analysis unit. The graph traversal unit traverses all nodes and connecting lines in the feature interaction graph in a preset order, and the weight analysis unit calculates the incoming and outgoing edge weights of each node.

[0082] Step S136: Extract the incoming and outgoing edge weights of all feature interaction graph nodes, and calculate the comprehensive association weight of the feature interaction graph nodes. The comprehensive association weight reflects the importance of the feature interaction graph nodes in the feature interaction graph.

[0083] Extract the incoming and outgoing edge weights of all nodes in the feature interaction graph, and calculate the comprehensive association weight of the nodes. The comprehensive association weight reflects the importance of each node in the feature interaction graph. The comprehensive association weight calculation is implemented through a weight fusion algorithm, which merges the incoming and outgoing edge weights of the nodes. The fusion method is determined according to the node type to generate a comprehensive association weight value for each node.

[0084] Step S137: Classify and organize the comprehensive association weights according to the feature interaction graph node type to generate the BeiDou high-precision positioning signal feature association weight subsequence and the inertial navigation raw data feature association weight subsequence.

[0085] The comprehensive association weights are categorized and organized according to the node types of the feature interaction graph, generating subsequences of feature association weights for BeiDou high-precision positioning signals and raw inertial navigation data. The categorization is achieved using a feature classification algorithm, grouping the comprehensive association weights of nodes belonging to BeiDou high-precision positioning signals into one category, generating a subsequence of feature association weights for BeiDou high-precision positioning signals; and grouping the comprehensive association weights of nodes belonging to raw inertial navigation data into another category, generating a subsequence of feature association weights for raw inertial navigation data.

[0086] Step S138: Concatenate the BeiDou high-precision positioning signal feature association weight subsequence and the inertial navigation original data feature association weight subsequence according to the feature interaction order to generate a complete feature association weight sequence.

[0087] The feature association weight subsequence of BeiDou high-precision positioning signal and the feature association weight subsequence of inertial navigation raw data are concatenated according to the feature interaction order to generate a complete feature association weight sequence. The feature interaction order is extracted from the feature interaction graph, and the two subsequences are concatenated according to the interaction order between nodes to generate a complete sequence containing all feature association weights.

[0088] Step S140: Construct a dynamic coupling rule set based on the feature association weight sequence, and input the Beidou high-precision positioning signal and the original inertial navigation data into the dynamic coupling rule set for tight coupling processing.

[0089] A dynamic coupling rule set is constructed based on the feature association weight sequence. BeiDou high-precision positioning signals and raw inertial navigation data are input into the dynamic coupling rule set for tight coupling processing. The dynamic coupling rule set contains coupling rules for different feature association weights. The tight coupling processing performs feature matching and interaction on the two data streams according to the rules in the rule set, generating coupled intermediate data.

[0090] Step S141: Parse the feature association weight sequence, extract feature pairs whose weights meet the core criteria and feature pairs whose weights meet the auxiliary criteria from the feature association weight sequence, and divide the first coupled feature into the second coupled feature.

[0091] The feature association weight sequence is parsed, and feature pairs whose weights meet the core criterion and those whose weights meet the auxiliary criterion are extracted. These are then divided into first-coupled features and second-coupled features. The feature association weight sequence parsing is implemented using a sequence analysis algorithm. The core criterion is set to a weight value higher than a preset core threshold, and the auxiliary criterion is set to a weight value higher than a preset auxiliary threshold but lower than the core threshold. The first-coupled features are feature pairs whose weights meet the core criterion, and the second-coupled features are feature pairs whose weights meet the auxiliary criterion.

[0092] Step S142: Construct the first coupling rule unit, formulate the first coupling clause based on the association law of feature pairs whose weights conform to the core standard, and clarify the interaction mode and timing of the first coupling feature.

[0093] The first coupling rule unit is constructed, and the first coupling clause is formulated based on the association patterns of feature pairs whose weights conform to the core standard. The first coupling clause clarifies the interaction mode and timing of the first coupling features. The first coupling rule unit includes a rule formulation module and a rule storage module. The association patterns are extracted from the feature interaction graph, and the first coupling clause clarifies the interaction type, interaction order, and interaction time interval of the feature pairs.

[0094] Step S143: Construct a second coupling rule unit, formulate a second coupling clause based on the association rules of feature pairs whose weights conform to the auxiliary criteria, and supplement the detailed requirements of the first coupling clause.

[0095] A second coupling rule unit is constructed, and second coupling clauses are formulated based on the association patterns of feature pairs whose weights conform to the auxiliary criteria. The second coupling clauses supplement the detailed requirements of the first coupling. The structure of the second coupling rule unit is consistent with that of the first coupling rule unit. The second coupling clauses supplement the details not covered by the first coupling clauses, such as the accuracy requirements of feature interactions and anomaly handling methods.

[0096] Step S144: Build a rule fusion unit, merge the first coupled rule unit and the second coupled rule unit to eliminate rule conflicts and form an initial rule set.

[0097] A rule fusion unit is constructed to merge the first coupled rule unit and the second coupled rule unit, eliminating rule conflicts and forming an initial rule set. The rule fusion unit includes a conflict detection module and a rule merging module. The conflict detection module compares the clauses in the two rule units to identify conflicting content; the rule merging module merges rules according to a preset conflict resolution strategy (such as prioritizing core clauses) to form the initial rule set.

[0098] Step S145: Input the optimized coupling link parameters to calibrate the initial rule set, adjust the parameter thresholds and interaction logic in the initial rule set, and generate a dynamic coupling rule set.

[0099] The initial rule set is calibrated by inputting optimized coupling link parameters, adjusting the parameter thresholds and interaction logic in the initial rule set, and generating a dynamically coupled rule set. The initial rule set calibration is achieved through a parameter calibration algorithm, which adjusts the feature association strength thresholds in the rule set based on the parameter association thresholds in the optimized coupling link parameters, and adjusts the feature interaction logic in the rule set based on the parameter interaction rules, thus generating the dynamically coupled rule set.

[0100] Step S146: Build a rule execution engine. The rule execution engine loads the dynamically coupled rule set and receives real-time BeiDou high-precision positioning signals and real-time inertial navigation raw data.

[0101] A rule execution engine is built, which loads a dynamically coupled rule set and receives real-time BeiDou high-precision positioning signals and real-time inertial navigation raw data. The rule execution engine includes a rule loading module, a data receiving module, and an execution control module. The rule loading module loads the dynamically coupled rule set into the engine; the data receiving module adapts to the output format of the BeiDou receiving module and the inertial measurement module to receive real-time data; and the execution control module controls the rule execution process.

[0102] Step S1461: Construct the core processing module of the rule execution engine. The core processing module of the rule execution engine includes a rule parsing submodule, a feature matching submodule, and an interaction processing submodule.

[0103] The core processing module of the rule execution engine comprises a rule parsing submodule, a feature matching submodule, and an interaction processing submodule. The rule parsing submodule breaks down the dynamically coupled rule set into executable rule instructions; the feature matching submodule extracts and pairs features from the input data; and the interaction processing submodule executes feature interaction operations according to the rule instructions.

[0104] Step S1462: The rule parsing submodule loads the dynamically coupled rule set and splits it into executable rule instructions, with each rule instruction corresponding to a specific coupling operation.

[0105] The rule parsing submodule loads the dynamically coupled rule set and breaks it down into executable rule instructions, each corresponding to a specific coupled operation. Rule parsing is achieved through an instruction splitting algorithm. The split rule instructions contain information such as operation type, operation object, and operation parameters, such as "weighted fusion of positioning coordinate features and motion trajectory features, with the weight ratio being the corresponding value in the feature association weight sequence".

[0106] Step S1463: Construct a signal and data receiving interface that is compatible with the output formats of the BeiDou receiving module and the inertial measurement module to achieve seamless access to real-time data.

[0107] A signal and data receiving interface is constructed, which is compatible with the output formats of the BeiDou receiver module and the inertial measurement module, enabling seamless real-time data access. The signal receiving interface supports the serial or Ethernet output format of the BeiDou receiver module, while the data receiving interface supports the CAN bus or Ethernet output format of the inertial measurement module. The interface includes a format conversion unit to convert the input data into a format that the engine can recognize.

[0108] Step S1464: The real-time BeiDou high-precision positioning signal is entered into the signal preprocessing submodule for signal noise reduction and feature extraction, and standardized BeiDou feature data is output.

[0109] The incoming real-time BeiDou high-precision positioning signal enters the signal preprocessing submodule for signal denoising and feature extraction, outputting standardized BeiDou feature data. Signal denoising uses an adaptive filtering algorithm to remove noise interference from the signal; feature extraction uses a feature extraction algorithm to extract core features such as positioning coordinates and signal stability; standardization processing converts the features into a unified format, outputting standardized BeiDou feature data.

[0110] Step S1465: The incoming real-time inertial navigation raw data enters the data preprocessing submodule, where data filtering and feature extraction are performed, and standardized inertial feature data is output.

[0111] The incoming real-time inertial navigation raw data enters the data preprocessing submodule for data filtering and feature extraction, outputting standardized inertial feature data. Data filtering removes outlier data points, and feature extraction uses feature extraction algorithms to extract core features such as motion trajectory and motion attitude. Standardization processing converts the features into a unified format, outputting standardized inertial feature data.

[0112] Step S1466: The feature matching submodule receives standardized BeiDou feature data and standardized inertial feature data, and performs feature pairing and association matching based on rule instructions.

[0113] The feature matching submodule receives standardized BeiDou feature data and standardized inertial feature data, and performs feature pairing and association matching based on rule instructions. Feature pairing is performed according to the feature pair requirements in the rule instructions, and association matching is achieved by calculating the similarity between features. Feature pairs with a similarity higher than a preset threshold are considered to be successfully matched.

[0114] Step S1467: The matched feature data is transmitted to the interaction processing submodule. The interaction processing submodule executes the feature interaction operation according to the requirements of the rule instructions to generate preliminary coupled data.

[0115] The matched feature data is transmitted to the interaction processing submodule, which executes feature interaction operations according to the rule instructions to generate preliminary coupled data. Feature interaction operations include weighted fusion, error compensation, and time synchronization. The operation parameters are extracted from the rule instructions, and the generated preliminary coupled data contains the fused feature information.

[0116] Step S1468: Initially couple the data and feed it back to the rule parsing submodule. Compare it with the rule instruction requirements and adjust the interaction processing parameters of the interaction processing submodule to make the coupling process conform to the rule instruction requirements.

[0117] The initial coupled data is fed back to the rule parsing submodule, where it is compared with the rule instructions. The interaction processing parameters of the interaction processing submodule are then adjusted to ensure the coupling process conforms to the rule instructions. The comparison is achieved through a data verification algorithm, which compares the initial coupled data with the expected results in the rule instructions. If discrepancies are found, the interaction processing parameters (such as fusion weights and compensation coefficients) are adjusted until the coupling process meets the requirements.

[0118] Step S147: The rule execution engine performs feature matching and interactive processing on the input BeiDou high-precision positioning signal and inertial navigation raw data in the order of priority of the first coupling clause and supplementation of the second coupling clause.

[0119] The rule execution engine performs feature matching and interactive processing on the input BeiDou high-precision positioning signal and inertial navigation raw data, prioritizing the first coupling clause and supplementing it with the second coupling clause. Feature matching pairs the features in the BeiDou high-precision positioning signal and the inertial navigation raw data according to the feature pairing requirements in the rule set; interactive processing performs data fusion, error compensation, and other operations on the matched features according to the interaction methods and timing in the clauses.

[0120] Step S148: In the process of feature matching and interaction processing, the feature association weight sequence is called in real time to dynamically adjust the weight allocation of feature interaction, complete the tight coupling process and generate coupling intermediate data.

[0121] During feature matching and interaction processing, the feature association weight sequence is invoked in real time to dynamically adjust the weight allocation of feature interactions, completing tight coupling processing and generating coupled intermediate data. The feature association weight sequence is invoked through a weight query module, which adjusts the interaction weights based on the feature changes in real-time data. The weight allocation affects the proportion of data fusion. After tight coupling processing is completed, coupled intermediate data containing fused features is generated.

[0122] Step S150: Output tightly coupled fused positioning data through the artificial intelligence coupling verification module. The tightly coupled fused positioning data carries the accuracy characteristics of Beidou high-precision positioning and the continuous characteristics of inertial navigation.

[0123] The AI-coupled verification module outputs tightly coupled fused positioning data, which combines the precision characteristics of BeiDou high-precision positioning with the continuous characteristics of inertial navigation. The AI-coupled verification module uses a pre-trained verification model to evaluate the quality of the intermediate coupled data. Once the evaluation is successful, it outputs tightly coupled fused positioning data containing high-precision positioning coordinates and continuous motion trajectories.

[0124] Step S151: The artificial intelligence coupling verification module loads the pre-trained verification model and receives the coupling intermediate data generated by the tight coupling processing.

[0125] The AI-coupled verification module loads the pre-trained verification model and receives coupled intermediate data generated by tight coupling processing. The verification model adopts a deep neural network structure. During the pre-training process, a large amount of qualified coupled intermediate data and corresponding standard localization data are used to learn data quality assessment rules. The coupled intermediate data contains fused feature information and is input into the verification model for quality assessment.

[0126] Step S152: Verify that the model performs feature decomposition on the coupled intermediate data to obtain the precision feature part originating from the BeiDou high-precision positioning signal and the continuous feature part originating from the original inertial navigation data.

[0127] The validation model performs feature decomposition on the coupled intermediate data to obtain a precision feature component originating from the BeiDou high-precision positioning signal and a continuous feature component originating from the original inertial navigation data. Feature decomposition is achieved through a feature separation algorithm. The precision feature component includes the precision information of the positioning coordinates and the signal stability information; the continuous feature component includes the continuity information of the motion trajectory and the stability information of the motion attitude.

[0128] Step S153: Perform feature integrity analysis on the accuracy feature part and extract the key indicators of the accuracy feature. The key indicators reflect the core accuracy attributes of Beidou high-precision positioning.

[0129] Feature integrity analysis is performed on the accuracy features to extract key indicators, which reflect the core accuracy attributes of BeiDou high-precision positioning. Feature integrity analysis is achieved through an integrity detection algorithm to check whether the accuracy features contain all core accuracy attributes. The extracted key indicators include the error range of positioning coordinates and the average signal-to-noise ratio, among others.

[0130] Step S154: Perform feature continuity analysis on the continuous feature part and extract the key indicators of the continuous features. The key indicators reflect the core continuous attributes of inertial navigation.

[0131] Feature continuity analysis is performed on continuous feature portions to extract key indicators that reflect the core continuity attributes of inertial navigation. Feature continuity analysis is implemented through a continuity detection algorithm to detect whether the motion trajectory of continuous feature portions is continuous and whether the motion attitude is stable. The extracted key indicators include the number of interruptions in the motion trajectory and the magnitude of changes in motion attitude.

[0132] Step S155: Construct a coupled quality assessment system based on the key indicators of accuracy features and the key indicators of continuous features, assess the fusion quality of the coupled intermediate data, and generate assessment results.

[0133] A coupling quality assessment system is constructed based on key indicators of precision and key indicators of continuity to evaluate the fusion quality of coupled intermediate data and generate assessment results. The coupling quality assessment system includes steps such as indicator weight allocation and assessment model construction. The assessment results include fusion quality level and indicator deviation description.

[0134] For example, step S1551: extract the precision feature key indicators and the continuous feature key indicators, and label the feature dimensions and expression forms of the precision feature key indicators and the continuous feature key indicators.

[0135] This process involves extracting key indicators for both precision and continuous features, and then labeling the feature dimensions and representations of these indicators. Key indicator extraction is achieved through an indicator extraction algorithm. The labeled feature dimensions include the physical dimensions of the indicator (such as length and angle), and the representation includes information such as the indicator's numerical range and units.

[0136] Step S1552: Construct an evaluation index weight allocation unit, and assign evaluation weights to the key indicators of accuracy features and the key indicators of continuous features based on the needs of the positioning application scenario.

[0137] An evaluation index weight allocation unit is constructed to assign evaluation weights to key indicators of accuracy and continuity based on the needs of positioning application scenarios. The evaluation index weight allocation unit includes a scenario analysis module and a weight calculation module. The scenario analysis module determines the degree of requirement for accuracy and continuity in the positioning application; the weight calculation module assigns evaluation weights to each key indicator according to the degree of requirement, with higher-demand indicators receiving greater weights.

[0138] Step S1553: Build the evaluation model framework. The evaluation model framework includes an indicator input layer, a weight fusion layer, and a result output layer. The indicator input layer, weight fusion layer, and result output layer are connected through a pre-trained mapping relationship.

[0139] An evaluation model framework is constructed, comprising an indicator input layer, a weight fusion layer, and a result output layer, which are connected through pre-trained mapping relationships. The indicator input layer receives key indicator data; the weight fusion layer fuses the key indicators with their corresponding evaluation weights; and the result output layer outputs the fused evaluation result. The pre-trained mapping relationships are learned through a large number of evaluation samples.

[0140] Step S1554: Input the key indicators of precision features and key indicators of continuous features into the indicator input layer of the evaluation model framework, and perform indicator standardization processing.

[0141] The key indicators of precision and continuous features are input into the indicator input layer of the evaluation model framework for indicator standardization. Indicator standardization is achieved through a standardization algorithm, converting key indicators of different dimensions and ranges into standardized values ​​within a unified range, facilitating subsequent weight fusion.

[0142] Step S1555: The standardized precision feature key indicators and continuous feature key indicators enter the weight fusion layer of the evaluation model framework, and are weighted and fused according to the assigned evaluation weights to generate a fused evaluation value.

[0143] The standardized precision feature key indicators and continuous feature key indicators enter the weight fusion layer of the evaluation model framework. They are then weighted and fused according to the assigned evaluation weights to generate a fused evaluation value. The weighted fusion is achieved through weight multiplication and addition operations. Each standardized indicator value is multiplied by its corresponding evaluation weight, and all products are added together to obtain the fused evaluation value.

[0144] Step S1556: The fused evaluation value is transmitted to the result output layer of the evaluation model framework. The result output layer maps the fused evaluation value to the pre-trained evaluation criteria to generate a graded evaluation result.

[0145] The fused evaluation values ​​are transmitted to the output layer of the evaluation model framework. The output layer maps the fused evaluation values ​​to pre-trained evaluation criteria, generating graded evaluation results. The evaluation criteria include multiple ranges of fused evaluation values ​​corresponding to different evaluation levels. The output layer determines the evaluation level based on the range of the fused evaluation values, generating graded evaluation results.

[0146] Step S1557: The graded evaluation results include the fusion quality level and index deviation description. The index deviation description marks the direction of the difference between the key indicators of accuracy features and the key indicators of continuous features and the standard values.

[0147] The tiered evaluation results include a fusion quality level and an indicator deviation description. The indicator deviation description indicates the direction of difference between the key indicators of accuracy characteristics and continuous key indicators and the standard values. The fusion quality level includes Excellent, Good, Pass, and Fail; the indicator deviation description, by comparing the key indicators with the standard values, indicates whether the indicators are too high or too low.

[0148] Step S1558: Feed back the hierarchical evaluation results to the tightly coupled processing stage. If there is a deviation, provide a reference for adjustment direction. If it meets the standard, trigger the data integration process.

[0149] The hierarchical evaluation results are fed back to the tightly coupled processing stage. If deviations exist, adjustment directions are provided as a reference; if the standards are met, the data integration process is triggered. The deviation adjustment direction reference is determined based on the indicator deviation description. For example, if the positioning coordinate error is too large, the fusion weight is adjusted. When the standards are met, the data integration process is triggered, and the coupled intermediate data is transmitted to the data integration module.

[0150] Step S156: When the evaluation result meets the preset standard, the intermediate data is coupled into the data integration module. The data integration module integrates the accuracy feature part and the continuous feature part according to the positioning application requirements.

[0151] When the evaluation results meet the preset standards, the intermediate data is coupled into the data integration module. The data integration module integrates the precision feature portion and the continuous feature portion according to the positioning application requirements. Data integration is achieved through a data fusion algorithm, which adjusts the fusion ratio of precision features and continuous features according to the positioning application's requirements for accuracy and continuity, generating integrated positioning data.

[0152] Step S157: Convert the format of the integrated feature data to generate standardized tightly coupled fused positioning data. The tightly coupled fused positioning data format is adapted to the reception requirements of mainstream positioning terminals.

[0153] The integrated feature data undergoes format conversion to generate standardized, tightly coupled fused positioning data. This tightly coupled fused positioning data format is compatible with the reception requirements of mainstream positioning terminals. Format conversion is achieved through a format conversion algorithm, converting the integrated positioning data into NMEA0183 or other mainstream positioning data formats to ensure that positioning terminals can recognize and receive it.

[0154] Step S158: Transmit tightly coupled fused positioning data through the data output interface. Record data transmission logs during the output process. The data transmission logs include output time and data integrity information.

[0155] Tightly coupled and fused positioning data is transmitted through the data output interface. During the output process, a data transmission log is recorded, containing output time and data integrity information. The data output interface supports serial port, Ethernet port, and other transmission methods. The transmission log is implemented through a log recording module, recording information such as the time of each output, data length, and amount of lost data.

[0156] It is worth noting that in the construction of the AI ​​coupling link parameter iteration module involved in step S120, this module adopts a three-layer fully connected neural network structure. The number of nodes in the input layer is consistent with the feature dimension of the initial coupling link parameters. The hidden layer contains two layers, each with 1.5 times the number of nodes in the input layer. The number of nodes in the output layer is consistent with the feature dimension of the optimized coupling link parameters. The layers are connected by linear activation functions, and nonlinear activation functions are added to the hidden layers to enhance the model's feature extraction capability. During model training, a training dataset containing BeiDou high-precision positioning signal samples, inertial navigation raw data samples, and corresponding standard coupling link parameters is first collected. The dataset sample size must cover at least three typical positioning scenarios, with each scenario accounting for no less than 20% of the total sample size. The training dataset is divided into a training subset and a validation subset in a 7:3 ratio. The training subset is used for model parameter updates, and the validation subset is used for evaluating the model's generalization ability. The training process uses a mini-batch gradient descent algorithm, with the batch size set to match the number of nodes in the input layer. The initial learning rate is set to a common base learning rate, and the learning rate decays by a fixed proportion after each preset training round. During training, initial coupling link parameters are input to the model's input layer, passed through to the first hidden layer, where linear transformation and nonlinear activation are applied to the input features. The processed features are then passed to the second hidden layer for secondary feature extraction. The output of the second hidden layer is then passed to the output layer to generate predicted coupling link parameters. The mean squared error between the predicted and standard coupling link parameters is calculated as the loss function. Backpropagation is used to update the weights and biases of each layer sequentially from the output to the input layer. After each training subset is completed, the validation loss is calculated using the validation subset. If the validation loss does not decrease after three consecutive training subsets, training stops, and the current model parameters are saved as pre-trained model parameters. In the construction of the AI ​​coupling verification module involved in step S150, this module adopts a structure combining convolutional neural networks and fully connected neural networks. The input layer receives coupling intermediate data, followed by two convolutional layers. The kernel size of each convolutional layer is set to a common small-sized kernel, and the number of kernels doubles layer by layer. After the convolutional layers, pooling layers are connected to reduce the feature dimension. After the pooling layers, two fully connected layers are connected, and the output layer outputs the fusion quality evaluation result. During model training, a training dataset containing coupling intermediate data samples and corresponding fusion quality evaluation results is collected. The dataset samples need to cover different fusion quality levels, and the number of samples for each level is balanced. The training process adopts the same dataset partitioning method and optimization algorithm as the AI ​​coupling link parameter iteration module. The loss function is the cross-entropy loss function. Training stops when the evaluation accuracy on the validation set reaches a preset threshold.In specific application scenarios, the input to the AI ​​coupling link parameter iteration module is the standardized initial coupling link parameters generated in step S110. The input parameters must be consistent with the input feature dimensions during model training. If the parameter dimensions do not match, the parameter dimensions are adjusted using a feature mapping algorithm before being input into the model. The optimized coupling link parameters output by the model are directly used for feature interaction map construction in step S130. The input to the AI ​​coupling verification module is the coupling intermediate data generated in step S140. The input data must be consistent with the input data format during model training. The fusion quality evaluation result output by the model is directly used to determine whether to trigger the data integration process. If the evaluation result is unqualified, it is fed back to step S140 to adjust the dynamic coupling rule set and re-perform tight coupling processing. Through the detailed settings of model construction, training, and scenario application described above, those skilled in the art can implement the solution of this invention according to the recorded content.

[0157] Figure 2 The illustration shows exemplary hardware and software components of a tightly coupled fusion positioning system 100 applying BeiDou high-precision positioning and inertial navigation, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the tightly coupled fusion positioning system 100 applying BeiDou high-precision positioning and inertial navigation and to perform the functions in this application.

[0158] The tightly coupled fusion positioning system 100 applying BeiDou high-precision positioning and inertial navigation can be a general-purpose server or a special-purpose server; both can be used to implement the tightly coupled fusion positioning method of BeiDou high-precision positioning and inertial navigation described in this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0159] For example, a tightly coupled fusion positioning system 100 applying BeiDou high-precision positioning and inertial navigation may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the tightly coupled fusion positioning system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The tightly coupled fusion positioning system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0160] For ease of explanation, only one processor is described in the tightly coupled fusion positioning system 100 applying BeiDou high-precision positioning and inertial navigation. However, it should be noted that the tightly coupled fusion positioning system 100 applying BeiDou high-precision positioning and inertial navigation in this application may also include multiple processors. Therefore, the steps executed by one processor described in this application may also be executed jointly by multiple processors or individually. For example, if the processor of the tightly coupled fusion positioning system 100 applying BeiDou high-precision positioning and inertial navigation executes steps A and B, it should be understood that steps A and B may also be executed jointly by two different processors or individually by one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0161] Furthermore, this embodiment of the invention also provides a readable storage medium, which has computer-executable instructions pre-set in it. When the processor executes the computer-executable instructions, the above-mentioned tightly coupled fusion positioning method of applying BeiDou high-precision positioning and inertial navigation is realized.

[0162] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A tightly coupled fusion positioning method applying BeiDou high-precision positioning and inertial navigation, characterized in that, The method includes: By accessing BeiDou high-precision positioning signals and raw inertial navigation data, a parallel transmission link for BeiDou high-precision positioning signals and raw inertial navigation data is established, and initial coupled link parameters are generated through link adaptation processing. The initial coupled link parameters are input into the pre-trained artificial intelligence coupled link parameter iteration module, which outputs optimized coupled link parameters based on the inherent correlation between BeiDou high-precision positioning signals and inertial navigation raw data. Based on the optimized coupling link parameters, a feature interaction map of BeiDou high-precision positioning signal and inertial navigation raw data is constructed, and a feature association weight sequence is generated by parsing the feature interaction map; A dynamic coupling rule set is constructed based on the feature association weight sequence. The Beidou high-precision positioning signal and the raw inertial navigation data are input into the dynamic coupling rule set for tight coupling processing. The tightly coupled fusion positioning data is output through the artificial intelligence coupling verification module. The tightly coupled fusion positioning data carries the accuracy characteristics of Beidou high-precision positioning and the continuous characteristics of inertial navigation.

2. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 1, characterized in that, The process involves accessing BeiDou high-precision positioning signals and raw inertial navigation data, establishing a parallel transmission link for these two data sets, and generating initial coupled link parameters through link adaptation processing. This includes: The system receives high-precision BeiDou positioning signals via a BeiDou receiver module and raw inertial navigation data via an inertial measurement module. Establish a dual-path parallel transmission link, configure the transmission protocol and data interaction timing of the dual-path parallel transmission link, and realize the synchronous transmission of Beidou high-precision positioning signals and inertial navigation raw data; Extract the link attribute information of the dual-path parallel transmission link. The link attribute information includes transmission delay characteristics, bandwidth occupancy characteristics, and signal attenuation characteristics. Input the link attribute information into the link adaptation processing module, and the link adaptation processing module outputs the link adaptation adjustment scheme, which covers transmission timing calibration and bandwidth allocation ratio. The dual-path parallel transmission link was optimized according to the link adaptation and adjustment scheme. The transmission parameters of the optimized dual-path parallel transmission link were collected. The transmission parameters include synchronization error characteristics and data integrity characteristics. An initial coupled link parameter set is constructed based on transmission parameters. The initial coupled link parameter set integrates the correlation information between the link adaptation and adjustment scheme and the transmission parameters. The initial coupled link parameter set is processed by parameter encoding to generate standardized initial coupled link parameters. The standardization process adopts the general encoding specification for link parameters.

3. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 1, characterized in that, The initial coupling link parameters are input into a pre-trained artificial intelligence coupling link parameter iteration module. This module outputs optimized coupling link parameters based on the inherent correlation between BeiDou high-precision positioning signals and raw inertial navigation data, including: The input layer of the AI ​​coupling link parameter iteration module receives the initial coupling link parameters, performs feature parsing on the initial coupling link parameters, and decomposes them to obtain the basic link parameters and associated adaptation parameters. The parsed link basic parameters and associated adaptation parameters enter the association mining layer of the artificial intelligence coupled link parameter iteration module. The association mining layer calls the pre-trained association feature extraction sub-module to mine the inherent association rules between Beidou high-precision positioning signals and inertial navigation raw data. A parameter association model is constructed based on the inherent correlation rules. The basic parameters of the link and the associated adaptation parameters are input into the parameter association model to generate a parameter association matrix. The parameter correlation matrix is ​​passed to the iterative adjustment layer of the parameter iteration module of the artificial intelligence coupling link. The iterative adjustment layer starts the parameter iteration process. The first iteration calculates the initial parameter deviation value based on the parameter correlation matrix. The first round of parameter adjustment scheme is generated based on the initial parameter deviation value, the initial coupling link parameters are adjusted and the second round of iteration parameters are generated; Repeat the process of parameter deviation calculation, adjustment scheme generation and parameter update until the deviation value of the iterative parameter meets the preset iteration termination condition of the artificial intelligence coupling link parameter iteration module; The parameters obtained from the final iteration are used as the optimized coupling link parameters and output through the output layer of the artificial intelligence coupling link parameter iteration module. Before output, the optimized coupling link parameters are formatted.

4. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 3, characterized in that, The correlation mining layer invokes a pre-trained correlation feature extraction submodule to mine the inherent correlation patterns between BeiDou high-precision positioning signals and raw inertial navigation data, including: The associated feature extraction submodule loads pre-trained feature extraction weights and receives BeiDou high-precision positioning signal samples and inertial navigation raw data samples; Signal feature decomposition was performed on the BeiDou high-precision positioning signal samples to obtain positioning coordinate features, signal matching degree features, and signal stability features; Data feature decomposition was performed on the raw inertial navigation data samples to obtain motion attitude features, motion trajectory features, and motion acceleration features; Construct feature pairing units to pair positioning coordinate features with motion trajectory features, and signal stability features with motion posture features, to form a feature pairing set; The correlation analysis is performed on the feature pairing set to generate a feature correlation sequence. The correlation analysis adopts a correlation evaluation algorithm based on sample training. Extract feature pairs from the feature correlation sequence that are correlated to a preset matching degree, and construct the basic framework of the correlation model based on the changing pattern of the feature pairs that are correlated to a preset matching degree; By inputting more BeiDou high-precision positioning signal samples and inertial navigation raw data samples, the basic framework of the correlation model is optimized, and indirect correlation information of feature pairs that do not reach the preset matching degree is supplemented, ultimately forming a description of the inherent correlation rules.

5. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 1, characterized in that, The process involves constructing a feature interaction map of BeiDou high-precision positioning signals and raw inertial navigation data based on optimized coupling link parameters, and generating a feature association weight sequence through feature interaction map parsing, including: Analyze and optimize the coupled link parameters, and extract the feature interaction rules and parameter association thresholds; A graph node generation unit is built to convert the core features of BeiDou high-precision positioning signals and the core features of inertial navigation raw data into feature interaction graph nodes; The feature interaction graph node connection relationship is constructed based on feature interaction rules. Feature interaction graph nodes with the same association type use the same type of connection lines, while nodes with different association types use different lines. The association matching degree information in the optimized coupling link parameters is assigned to the connection lines of the feature interaction graph nodes to form a weighted feature interaction graph, and the weight value directly maps to the association matching degree. A graph parsing module is built, which loads a graph parsing algorithm to traverse and analyze the connection relationships and weight distribution of feature interaction graph nodes in the feature interaction graph. Extract the incoming and outgoing edge weights of each node in the feature interaction graph, and calculate the comprehensive association weight of the feature interaction graph nodes. The comprehensive association weight reflects the importance of the feature interaction graph nodes in the feature interaction graph. The comprehensive association weights are classified and organized according to the node type of the feature interaction graph, generating a subsequence of feature association weights for BeiDou high-precision positioning signals and a subsequence of feature association weights for inertial navigation raw data. The feature association weight subsequence of BeiDou high-precision positioning signal features is concatenated with the feature association weight subsequence of inertial navigation raw data according to the feature interaction order to generate the feature association weight sequence.

6. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 5, characterized in that, The feature interaction graph node connection relationship is constructed based on feature interaction rules. Feature interaction graph nodes with the same association type use the same type of connection lines, while nodes with different association types use different lines, including: Analyze feature interaction rules and classify feature association types, which include direct association, indirect association, and feedback association. Different feature association types are assigned their own unique connection line types: direct association types use solid lines, indirect association types use dashed lines, and feedback association types use solid lines with arrows. Traverse all BeiDou high-precision positioning signal feature interaction map nodes and inertial navigation raw data feature interaction map nodes, determine the feature association type between any two feature interaction map nodes, and match the corresponding connection line type; For the connection of feature interaction graph nodes of the direct association type, the direct association basis is marked on the connection line. The direct association basis comes from the direct association clause in the optimized coupling link parameters. For indirect association type feature interaction graph node connections, trace the intermediate association feature interaction graph nodes, mark the names of the intermediate association feature interaction graph nodes next to the connecting lines, and clarify the transmission path of indirect association; For the feature interaction graph nodes of the feedback association type, the feedback direction is marked with the arrow direction to clarify the interaction direction between the Beidou high-precision positioning signal features and the inertial navigation raw data features; After completing the connection of all feature interaction graph nodes, perform a redundancy check on the connecting lines, delete duplicate connecting lines, and retain the connection relationships that best reflect the essence of the association.

7. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 1, characterized in that, The dynamic coupling rule set constructed based on feature association weight sequences involves inputting BeiDou high-precision positioning signals and raw inertial navigation data into the dynamic coupling rule set for tight coupling processing, including: Analyze the feature association weight sequence, extract feature pairs whose weights conform to the core criteria and feature pairs whose weights conform to the auxiliary criteria, and divide them into first coupled features and second coupled features. Construct the first coupling rule unit, and formulate the first coupling clause based on the association rules of feature pairs whose weights conform to the core standard. The first coupling clause clarifies the interaction mode and timing of the first coupling features. Construct a second coupling rule unit, and formulate second coupling clauses based on the association rules of feature pairs whose weights conform to auxiliary standards. The second coupling clauses supplement the detailed requirements of the first coupling. A rule fusion unit is constructed to fuse the first coupled rule unit and the second coupled rule unit to eliminate rule conflicts and form an initial rule set. The initial rule set is calibrated by inputting optimized coupling link parameters, adjusting the parameter thresholds and interaction logic in the initial rule set, and generating a dynamic coupling rule set; A rule execution engine is built, which loads a dynamically coupled rule set and receives real-time BeiDou high-precision positioning signals and real-time inertial navigation raw data. The rule execution engine performs feature matching and interactive processing on the input BeiDou high-precision positioning signal and inertial navigation raw data in the order of priority of the first coupling clause and supplementation of the second coupling clause; During feature matching and interaction processing, the feature association weight sequence is invoked in real time to dynamically adjust the weight allocation of feature interaction, complete tight coupling processing, and generate coupled intermediate data.

8. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 7, characterized in that, The rule execution engine is constructed, which loads a dynamically coupled rule set and receives real-time BeiDou high-precision positioning signals and real-time inertial navigation raw data, including: The core processing module of the rule execution engine is constructed, which includes a rule parsing submodule, a feature matching submodule, and an interaction processing submodule. The rule parsing submodule loads the dynamically coupled rule set, breaks it down into executable rule instructions, and each rule instruction corresponds to a specific coupling operation; Construct a signal and data receiving interface that is compatible with the output formats of the BeiDou receiving module and the inertial measurement module to achieve seamless access to real-time data; The real-time BeiDou high-precision positioning signal is fed into the signal preprocessing submodule for signal noise reduction and feature extraction, and outputs standardized BeiDou feature data. The incoming real-time inertial navigation raw data enters the data preprocessing submodule, where data filtering and feature extraction are performed, and standardized inertial feature data is output. The feature matching submodule receives standardized BeiDou feature data and standardized inertial feature data, and performs feature pairing and association matching based on rule instructions; The matched feature data is transmitted to the interaction processing submodule, which executes the feature interaction operation according to the rule instructions to generate preliminary coupled data. The initial coupled data is fed back to the rule parsing submodule, compared with the rule instruction requirements, and the interaction processing parameters of the interaction processing submodule are adjusted to ensure that the coupling process conforms to the rule instruction requirements.

9. The tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in claim 1, characterized in that, The tightly coupled fused positioning data output by the artificial intelligence coupling verification module carries the accuracy characteristics of BeiDou high-precision positioning and the continuous characteristics of inertial navigation, including: The AI-coupled verification module loads the pre-trained verification model and receives the coupled intermediate data generated by tight coupling processing. The verification model performs feature decomposition on the coupled intermediate data to obtain the precision feature part originating from the Beidou high-precision positioning signal and the continuous feature part originating from the original inertial navigation data; A feature integrity analysis is performed on the accuracy feature part to extract the key indicators of the accuracy feature. The key indicators reflect the core accuracy attributes of BeiDou high-precision positioning. Continuity analysis is performed on continuous features to extract key indicators of continuous features. These key indicators reflect the core continuous attributes of inertial navigation. A coupling quality assessment system is constructed based on key indicators of precision features and key indicators of continuous features to evaluate the fusion quality of coupling intermediate data and generate assessment results. When the evaluation results meet the preset standards, the intermediate data is coupled into the data integration module, which integrates the accuracy feature part and the continuous feature part according to the positioning application requirements. The integrated feature data is converted to generate standardized tightly coupled fused positioning data. The tightly coupled fused positioning data format is adapted to the reception requirements of mainstream positioning terminals. Tightly coupled fused positioning data is transmitted through the data output interface. During the output process, a data transmission log is recorded, which includes output time and data integrity information.

10. A tightly coupled fusion positioning system applying BeiDou high-precision positioning and inertial navigation, characterized in that, The tightly coupled fusion positioning system using BeiDou high-precision positioning and inertial navigation includes a processor and a memory. The memory and the processor are connected. The memory is used to store programs, instructions, or code. The processor is used to execute the programs, instructions, or code in the memory to implement the tightly coupled fusion positioning method using BeiDou high-precision positioning and inertial navigation as described in any one of claims 1-9.

Images