Combined navigation type underwater operation platform high-precision positioning method

Abstract

The invention relates to the technical field of underwater navigation and positioning, and discloses a high-precision positioning method for an integrated navigation type underwater operation platform. According to the method, inertial navigation data, sonar positioning data and satellite positioning data are acquired to form a multi-dimensional navigation data source, and a navigation data warehouse is generated after data fusion processing. And further extracting a navigation feature matrix, establishing a feature link map to represent the association relationship of the multi-source data, and screening a core positioning feature sequence for positioning the control platform. According to the method, stable positioning in an underwater complex environment is realized. A weight distribution mechanism of multi-source data is optimized through introduction of a feature link graph, the screening process of a core feature sequence is adaptive to environmental changes, and the negative influence of redundant data on positioning precision is reduced. Compared with a traditional fixed weight fusion method, the method shows higher robustness in the scenes of water flow disturbance, acoustic signal attenuation and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater navigation and positioning, and in particular to a high-precision positioning method for a combined navigation underwater operating platform. Background Art

[0002] Positioning technology for underwater operating platforms has always been a key and difficult point in the fields of natural fluid engineering and resource exploration. Due to the complex and changing underwater environment and the severe attenuation of electromagnetic wave signals, traditional single navigation methods often cannot meet the high-precision and high-reliability positioning requirements. Although inertial navigation systems can provide continuous and autonomous position information, their errors accumulate over time, resulting in insufficient long-term accuracy. Sonar positioning technology can achieve relative positioning within a certain range, but it is easily affected by water conditions, multipath effects, and environmental noise, and its coverage is limited. Satellite positioning can achieve higher accuracy on the water surface, but for underwater platforms, the signal cannot penetrate the water and must rely on buoys or surface units for relaying, which limits its real-time performance and availability.

[0003] Traditional methods have weak tolerance for abnormal data and lack a systematic screening and optimization mechanism for navigation features, which limits further improvement in overall positioning performance. Therefore, it is urgent to develop a high-precision positioning method that can deeply integrate multi-source navigation information and has the ability to analyze and optimize feature correlations to meet the long-term, stable, and precise operation requirements of underwater platforms in complex underwater environments.

[0004] Existing technologies have attempted to improve positioning accuracy through multi-sensor data fusion, but these methods typically employ simple weighted fusion or Kalman filtering, failing to fully consider the interdependencies between the characteristics of different navigation data sources. For example, the dynamic characteristics of inertial navigation data lack deep coupling with the spatial geometric constraints of sonar positioning, resulting in insufficient positioning robustness in complex underwater environments. Furthermore, traditional methods employ a relatively rigid data source selection mechanism, failing to adapt to dynamic changes in the underwater environment (such as water flow disturbances and terrain obstructions), leading to prone to positioning jumps or failures.

[0005] Some studies have attempted to incorporate machine learning to optimize the data fusion process, but model training relies on extensive historical data and fails to address the spatiotemporal alignment of multi-source data. Issues such as sampling frequency discrepancies between sonar and inertial navigation data and intermittent loss of satellite signals can still lead to lags or sudden changes in the fused positioning results. Therefore, a high-precision underwater positioning method that can adaptively select core features and dynamically establish data associations is urgently needed. Summary of the Invention

[0006] The purpose of the present invention is to provide a high-precision positioning method for a combined navigation underwater working platform to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides a high-precision positioning method for an integrated navigation underwater working platform, the method comprising: Acquire a multi-dimensional navigation data source of the underwater operation platform, wherein the multi-dimensional navigation data source includes inertial navigation data, sonar positioning data, and satellite positioning data; Based on the multi-dimensional navigation data source, data fusion processing is performed to generate a navigation data warehouse; extracting a navigation feature matrix based on the navigation data warehouse; Determining a feature link graph of a multidimensional navigation data source based on the navigation feature matrix; Based on the feature linkage map, screening core positioning feature sequences; Based on the core positioning feature sequence, the positioning process of the underwater operating platform is controlled.

[0008] Preferably, said extracting the navigation feature matrix based on said navigation data warehouse comprises: separating navigation data of various dimensions from said navigation data warehouse; The navigation data of each dimension is subjected to feature extraction processing to generate a dimension feature vector; the dimension feature vectors of all dimensions are integrated to construct a navigation feature matrix.

[0009] Preferably, determining the feature link map of the multi-dimensional navigation data source based on the navigation feature matrix includes: selecting a target dimension in the navigation feature matrix and obtaining a dimensional feature vector of the target dimension; Calculating the feature correlation between the navigation features in the dimensional feature vector; generating a dimensional link graph of the target dimension based on the feature correlation; Traverse all dimensions and repeat the steps of generating dimension link diagram; Based on the dimension link graph of all dimensions, the feature link graph of the multi-dimensional navigation data source is generated by connecting the feature associations of the key navigation features.

[0010] Preferably, the screening of the core positioning feature sequence based on the feature link graph includes: extracting the dimensional entropy value corresponding to each navigation feature in the navigation feature matrix; Extracting the feature correlation degree of each navigation feature from the feature link graph; Calculate the core entropy value of each navigation feature by combining the dimension entropy value and the feature correlation degree; Based on the core entropy value, filtering out navigation features that are higher than a preset threshold as core positioning features; Aggregate all core positioning features to generate a core positioning feature sequence.

[0011] Preferably, the process of controlling the positioning of the underwater operating platform based on the core positioning feature sequence includes: analyzing common patterns in the core positioning feature sequence to extract reusable positioning parameters; Analyze the difference patterns in the core positioning feature sequence and extract scene-specific positioning parameters; Based on the reusable positioning parameters and the specific scenario positioning parameters, the positioning parameters of the underwater operation platform are adjusted.

[0012] Preferably, the method further comprises: classifying the positioning state type of the underwater working platform based on the navigation data warehouse, wherein the positioning state type includes a stable positioning state and an unstable positioning state; Based on the positioning status type, attach a status priority tag; Classifying resource consumption types based on the multi-dimensional navigation data source, the resource consumption types including computing resource consumption and communication resource consumption; Based on the resource consumption type, a resource type tag is added.

[0013] Preferably, triggering a positioning parameter adjustment signal based on the positioning state type and the resource type tag includes: generating a first positioning parameter adjustment signal when an unstable positioning state is detected; generating a second positioning parameter adjustment signal when a change in the resource consumption type is detected; generating a positioning parameter change node based on the first positioning parameter adjustment signal and the second positioning parameter adjustment signal; The positioning parameter replacement node includes a parameter adjustment node within the same stable positioning state and a parameter reset node when switching between different positioning state types.

[0014] Preferably, calculating the positioning offset parameter based on the positioning parameter change node includes: extracting parameter adjustment nodes within the same stable positioning state and marking them as same positioning parameter change nodes; Extract the parameter reset nodes when switching between different positioning state types and mark them as different positioning parameter replacement nodes; Replace the node based on the same positioning parameters, calculate the parameter offset in the stable positioning state, and record it as the first positioning offset parameter; Based on the node replacement with different positioning parameters, the parameter offset when the positioning state type is switched is calculated and recorded as the second positioning offset parameter; The first positioning offset parameter and the second positioning offset parameter are combined to generate a positioning offset parameter.

[0015] Preferably, said optimizing positioning control based on said positioning offset parameters comprises: analyzing reusable patterns in the positioning offset parameters to extract migration positioning parameters; Analyze specific patterns in positioning offset parameters and extract independent positioning parameters; Based on the migration positioning parameters, set the global positioning benchmark; Dynamically adjust parameters for specific positioning scenarios based on independent positioning parameters; Combining the global positioning reference and dynamically adjusting parameters, adaptive positioning control is achieved.

[0016] Preferably, the method further comprises: setting an early warning mechanism and an exit mechanism for the high-precision positioning state; When the early warning mechanism is triggered, the underwater operating platform enters a high-precision positioning state, and all new navigation data enters a priority sequence for processing; Optimize sequence processing order based on status priority tags and resource type tags; When the exit mechanism is triggered, the underwater operating platform exits the high-precision positioning state and the new navigation data enters the normal processing sequence.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention significantly improves the positioning accuracy of underwater work platforms by integrating multidimensional navigation data sources and constructing a feature-linked map. The short-term stability of inertial navigation data complements the spatial constraints of sonar positioning, while satellite positioning data provides absolute position calibration. These three factors work together to mitigate the inherent limitations of a single data source. The feature-linked map dynamically reflects the strength of associations between different data sources, avoiding the error amplification problem caused by fixed-weight fusion.

[0018] The core positioning feature sequence selection mechanism is adaptive to environmental changes. When sonar signals are affected by multipath interference, the system automatically downweights sonar data and enhances the short-term prediction capabilities of inertial navigation. When satellite signals are lost, positioning continuity is maintained through deep coupling of sonar and inertial data. This dynamic adjustment effectively suppresses the positioning jumps common in traditional methods.

[0019] The structured storage of the navigation data warehouse supports rapid retrieval and comparison of historical data, providing real-time reference for updating the feature matrix. During the data fusion process, a spatiotemporal alignment algorithm eliminates sampling frequency differences among multi-source data, ensuring temporal consistency in feature extraction. This method demonstrates greater adaptability in complex underwater terrain. For example, on steep slopes or in areas with dense obstacles, the geometric features of sonar data match the inertial trajectory with higher accuracy.

[0020] The construction of the feature link graph incorporates a topology optimization strategy to reduce computational resource consumption. Compared to traditional machine learning methods that rely on offline training, the present invention's online feature screening mechanism reduces reliance on historical data and is more suitable for dynamic scenarios in real-world operations. The synergy of multidimensional data sources extends the system's effective operating time, maintaining positioning accuracy even during prolonged deepwater operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a working principle diagram of the high-precision positioning method for an integrated navigation underwater operating platform according to the present invention; Figure 2 Flowchart for navigation feature matrix extraction; Figure 3 Flowchart for core positioning feature sequence screening; Figure 4 Flowchart attached for classification and labeling of location status types and resource consumption types. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figure 1 The present invention provides a high-precision positioning method for a combined navigation underwater operating platform, the method comprising: By integrating multi-dimensional navigation data sources, precise position control of the underwater operating platform is achieved, and multi-dimensional navigation data sources for the underwater operating platform are obtained, including inertial navigation data, sonar positioning data, and satellite positioning data. These data sources are collected in real time through sensors and receiving devices. Inertial navigation data comes from the platform's inertial measurement unit (IMU), providing acceleration and angular velocity information; sonar positioning data is obtained through an underwater sonar system and is used to measure distance and direction relative to a reference point; satellite positioning data is obtained through a global navigation satellite system (GNSS) receiver, providing global position coordinates when the platform approaches the water surface. All data sources are synchronized with timestamps to ensure time consistency.

[0024] Data fusion processing is performed to generate a navigation data warehouse. Data fusion uses the Kalman filter algorithm to perform weighted fusion of information from different data sources to reduce noise and errors. The navigation data warehouse is a structured database that stores historical and real-time navigation data, including position, velocity, attitude, and time information. The data warehouse is designed to support efficient query and data retrieval, facilitating subsequent feature extraction. Based on the navigation data warehouse, a navigation feature matrix is ​​extracted. The navigation feature matrix is ​​a multidimensional matrix in which each row represents a time point and each column represents a navigation feature. The feature extraction process involves calculating derived indicators from the raw data, such as position deviation, velocity change rate, attitude angle, etc. These features are used to capture the dynamic behavior of the navigation data.

[0025] Based on the navigation feature matrix, a feature link graph is determined for the multidimensional navigation data source. The feature link graph is a graph structure in which nodes represent navigation features and edges represent the degree of correlation between features. This graph is constructed by calculating the correlation coefficient or mutual information between features to reveal the inherent connections between data sources. Based on the feature link graph, a core positioning feature sequence is screened. The core positioning feature sequence is a set of high-impact features that significantly contribute to positioning accuracy. The screening process selects features with a score above a threshold based on their importance score. Based on the core positioning feature sequence, the positioning process of the underwater operating platform is controlled. The control process involves adjusting the platform's positioning parameters, such as filter settings, sensor weights, or output frequency, to optimize positioning performance. The control algorithm adaptively adjusts parameters based on the dynamic changes in the core feature sequence to ensure the platform maintains high-precision positioning in various underwater environments.

[0026] Example 1: See Figure 2A navigation feature matrix is ​​extracted from the navigation data warehouse, and a feature linkage graph of multidimensional navigation data sources is constructed. The navigation data warehouse, a centralized storage system, contains raw data records from inertial navigation units, sonar positioning devices, and satellite receivers. This data is typically stored in time series, with each data point accurately timestamped to ensure temporal consistency. The separation operation requires data segmentation based on source and type. For example, the acceleration and angular velocity data output by the inertial measurement unit (IMU) can be classified as inertial navigation data, the distance and azimuth angles measured by the sonar device can be classified as sonar positioning data, and the longitude and latitude coordinates provided by the satellite receiver can be classified as satellite positioning data. During this process, special attention must be paid to data synchronization, as different sensors may have different sampling frequencies. In this case, interpolation or sampling rate unification is required to align all data on the time axis. After data separation, the datasets for each dimension are extracted separately, forming data subsets for further processing. These data subsets contain not only the raw measurements but also some pre-processed derivative data, such as sensor readings that have undergone preliminary filtering or position information that has undergone coordinate transformation.

[0027] To complete data separation, feature extraction is performed on each dimension of navigation data. The goal of feature extraction is to extract quantitative metrics that represent the data's characteristics from the raw data. For inertial navigation data, statistical features may be calculated within a moving window, such as the mean and variance of acceleration data within a sliding time window. These statistics can reflect the stability of the motion state. Frequency domain features can also be calculated by performing a Fourier transform on the acceleration data to obtain the frequency domain energy distribution, which helps identify periodic motion patterns. For sonar positioning data, feature extraction may include calculating signal quality metrics such as the signal-to-noise ratio of the echo signal and the strength of multipath effects. These features help assess the reliability of sonar measurements. Feature extraction for satellite positioning data focuses on positioning accuracy metrics such as the calculation of horizontal and vertical dilutions of precision, as well as trends in satellite signal strength. Each feature extraction process requires selecting appropriate algorithms and parameter settings based on the specific data characteristics. All extracted feature values ​​are organized into dimensional feature vectors, each representing a set of features for that dimension at a specific point in time. The length of a dimensional feature vector depends on the number of features extracted; typically, dozens of feature metrics are extracted for each dimension.

[0028] The completed feature vectors for all dimensions need to be integrated into a navigation feature matrix. This integration process requires time alignment to ensure that feature vectors from different dimensions correspond to each other at each time point. The integrated navigation feature matrix is ​​a two-dimensional array structure, with rows representing time points and columns representing feature indicators. The number of columns in the matrix equals the sum of the number of features across all dimensions, and the number of rows equals the number of time points. After constructing the matrix, data normalization is typically required because different features may have different dimensions and numerical ranges. Normalization converts all feature values ​​to the same numerical range. Common methods include minimum-maximum normalization and standard deviation normalization. A normalized feature matrix is ​​more conducive to subsequent feature correlation analysis because it eliminates the influence of dimensional differences. Based on the navigation feature matrix, to determine the feature linkage map of a multidimensional navigation data source, a target dimension in the navigation feature matrix must be selected. The choice of target dimension is arbitrary, but processing typically begins with the dimension with the richest or most important data. Obtaining the dimensional feature vector for the target dimension involves extracting all feature columns belonging to that dimension from the feature matrix. For example, if the inertial navigation dimension is selected as the target dimension, all feature columns related to inertial navigation must be extracted from the matrix. These feature columns constitute the set of feature vectors of this dimension at all time points.

[0029] Calculating the feature correlation between navigation features in the dimensional feature vector is the next important step. Feature correlation measures the degree of correlation between different features, which can be linear or nonlinear. Correlation coefficients are usually used to calculate linear correlation. For example, the Pearson correlation coefficient can measure the linear correlation strength between two features. For nonlinear relationships, indicators such as mutual information may be used to measure the degree of dependence between features. The calculation process requires traversing all feature pairs. For The dimension of the features needs to be calculated The correlation between features. The correlation values ​​generated by these calculations are stored in a correlation matrix, which is symmetric and the elements on the diagonal are usually set to 1, indicating that the feature is completely correlated with itself.

[0030] Based on the calculated feature correlations, a dimension link graph for the target dimension can be generated. A dimension link graph is a graph structure in which nodes represent features and edges represent the relationships between features. Edge weights are determined by feature correlations. Typically, only edges with correlations exceeding a certain threshold are retained to avoid accidental correlations that may be caused by noise. This graph can be constructed using graph theory algorithms, with each feature as a node and the correlation as the edge weight, forming a weighted undirected graph. This graph structure intuitively demonstrates the closeness of the relationships between features within a dimension. Repeating the above steps across all dimensions is necessary because each dimension requires its own dimension link graph. For the sonar positioning dimension, the feature columns of that dimension are extracted, the correlations between these features are calculated, and a link graph for the sonar positioning dimension is constructed. Similarly, the same process is performed for the satellite positioning dimension. A separate dimension link graph is generated for each dimension, reflecting the relationship structure between features within that dimension.

[0031] Based on the dimensional link graphs across all dimensions, a feature link graph of multidimensional navigation data sources is generated by connecting key navigation features through their feature correlations. Key navigation features are those that appear in multiple dimensions or are of particular importance. For example, position estimation features may exist in both inertial navigation and satellite positioning dimensions, serving as a bridge connecting different dimensions. The linking process identifies these cross-dimensional features and connects the link graphs of different dimensions through these common features. The resulting feature link graph is a comprehensive graph structure that displays the relationships between all features across all dimensions, including both intra-dimensional and inter-dimensional correlations. This graph helps understand the inherent connections between different data sources and provides a foundation for subsequent feature screening and analysis. The graph is typically stored in the form of an adjacency list or adjacency matrix, while retaining metadata for all features, such as feature name and dimension.

[0032] Example 2: See Figure 3Based on the feature link graph, a core positioning feature sequence is selected and used to precisely control the positioning process of the underwater platform. Dimensional entropy is a metric used to measure the uncertainty of the eigenvalue distribution. Its calculation is based on concepts from information theory and is obtained by analyzing the statistical distribution characteristics of the eigenvalue sequence. Specifically, for each feature column in the navigation feature matrix, the system constructs a probability distribution model of the eigenvalue at different time points. This model can be implemented using methods such as histogram statistics or kernel density estimation. Based on this probability distribution, the information entropy formula is used for calculation, which quantifies the degree of chaos or randomness of the eigenvalue. For example, if a feature value remains relatively stable over time and fluctuates slightly, its entropy will be low, indicating that the feature contains relatively limited and predictable information. Conversely, if the eigenvalue fluctuates wildly and exhibits a high degree of randomness, its entropy will be high, indicating that the feature may carry more dynamic information but may also introduce more noise. In actual operation, this calculation process requires traversing the navigation feature matrix column by column, performing probability modeling and entropy calculation on the feature value sequence of each column, and storing the result as an entropy value vector corresponding to the feature column, which will be used for subsequent comprehensive evaluation.

[0033] Based on the obtained dimensional entropy values, the system further extracts the feature relevance of each navigation feature from the feature link graph. The feature link graph is a graph structure whose nodes represent individual navigation features and whose edges represent the relationships between features. The edge weights represent the numerical value of the feature relevance. Feature relevance reflects the strength of the mutual influence between different features, with higher values ​​indicating greater synergy or dependency between the features. The extraction process involves traversing all nodes in the feature link graph. For each feature corresponding to a node, the weights of all adjacent edges are collected. A representative relevance score is calculated through aggregation operations such as weighted averaging or max pooling. For example, if a feature has strong connections with multiple other features, its relevance score will be high, indicating that the feature is central to the overall data relationship. Conversely, if a feature has few connections or low connection weights, its relevance score will be low, indicating that the feature may be relatively independent or less influential. The output of this step is a relevance vector, where each element corresponds to the relevance score of a navigation feature.

[0034] The system needs to combine the dimension entropy and feature correlation to calculate the core entropy of each navigation feature. The core entropy is a comprehensive indicator that aims to balance the information content of the feature itself and its influence in the overall map. Its calculation formula is:

[0035] in: represents the core entropy value, represents the dimension entropy value, represents the degree of feature correlation, while α and β are weighting coefficients used to adjust the relative importance of entropy and correlation in the comprehensive evaluation. The weighting coefficients need to be adjusted based on the specific application scenario. For example, in scenarios that emphasize data stability, feature correlation may be given a higher weight to prioritize features that synergize well with other features. In scenarios that prioritize information richness, dimension entropy may be given a higher weight to retain features that carry more dynamic information. After calculation, each navigation feature is assigned a core entropy score, which is used in subsequent feature screening. Based on the core entropy score, the system selects navigation features that exceed a preset threshold as core positioning features. The preset threshold is typically determined based on historical data or experimental experience; for example, it can be set to the upper quartile of the core entropy sequence or a dynamically adjusted statistical quantile. The screening process iterates over all navigation features, compares their core entropy scores with the threshold, and selects those with higher scores for inclusion in the core set. These core positioning features typically represent key factors that significantly contribute to positioning accuracy, such as high-precision position estimation features or stable velocity measurement features. The system aggregates all the selected core positioning features and sorts them from high to low by core entropy value to form a core positioning feature sequence. This sequence serves as an ordered set and provides input for subsequent positioning control.

[0036] After obtaining the core positioning feature sequence, the system begins controlling the underwater platform's positioning process based on this sequence. This control process first involves analyzing common patterns within the features in the sequence to extract reusable positioning parameters. Common patterns are stable behavioral patterns that occur across multiple core features. For example, in a stable positioning state, multiple features may exhibit low variance or high autocorrelation. Pattern recognition typically uses methods such as cluster analysis or principal component analysis to classify the feature value sequence into distinct pattern categories and extract representative parameters from each category, such as mean offset or standard deviation. These reusable positioning parameters constitute a basic parameter set suitable for general positioning scenarios and provide a stable baseline configuration for the system. Simultaneously, the system analyzes differential patterns within the core positioning feature sequence to extract scenario-specific positioning parameters. Differential patterns are unique characteristic behaviors that occur under specific conditions or abnormal circumstances. For example, when the sonar signal is interfered with, the relevant features may exhibit unusual fluctuations or sudden changes.

[0037] Differential pattern detection typically employs anomaly detection algorithms or change point detection techniques to identify outliers or sudden changes in the feature sequence and extract corresponding parameter adjustments, such as temporary filter coefficients or sensor weight adjustments. These scenario-specific positioning parameters address atypical or unexpected situations and provide the system with flexible configurations for special environments. The system dynamically adjusts the underwater platform's positioning parameters based on reusable positioning parameters and scenario-specific positioning parameters. Parameter adjustments are implemented using a control algorithm, such as a proportional-integral-derivative controller or model predictive control algorithm, which calculates and outputs parameter adjustment commands based on real-time input of core positioning feature values. Adjustments may involve modifying the filter cutoff frequency, adjusting sensor data fusion weights, or changing the positioning output frequency. The entire control process forms a closed-loop feedback loop, with the system continuously monitoring changes in the core positioning feature sequence and dynamically updating parameter settings to ensure high-precision positioning of the underwater platform under various operating conditions. During implementation, computational efficiency may need to be considered, employing incremental learning or sliding window techniques to optimize real-time performance and ensure the system can rapidly respond to environmental changes.

[0038] Example 3: See Figure 4 The system classifies the underwater platform's positioning status and resource consumption patterns and triggers positioning parameter adjustment signals based on these classifications. This classification process is achieved by analyzing historical and real-time positioning data stored in the navigation data warehouse. This data includes multivariate time series information such as position coordinates, velocity vectors, and attitude angles. The system uses a sliding time window technique to perform statistical analysis on the positioning data within each time window, calculating metrics such as the variance of the position data within the window, the magnitude of velocity variation, and the fluctuation range of attitude angles. For example, for a 30-second time window, the system calculates the three-dimensional coordinate variance of all points within the window. If the variance in all three coordinate axes is below a preset threshold, the positioning state for that period is classified as stable. Conversely, if the variance in any direction exceeds the threshold, the positioning state is classified as unstable. In addition to variance analysis, the system also examines the continuity and smoothness of the data, detecting abnormal jumps by calculating the rate of position change between adjacent time points. These jumps often indicate interference or malfunction in the positioning system. Classification algorithms can use rule-based methods or machine learning classifiers, such as support vector machines or random forests, to build classification models by training historical data. Regardless of the method used, the classification results are stored as time series labels, with each time point corresponding to a location status type.

[0039] After completing the positioning status type classification, the system will attach a status priority label to each status. The status priority label is a quantitative indicator that represents the importance or urgency of the status. Its value is based on factors such as the nature of the status type itself and the duration of the status. For example, an unstable positioning state is usually assigned a high priority label because this state indicates that there may be a problem with the positioning system and requires timely attention and treatment; while a stable positioning state is assigned a normal priority label. The priority can be calculated using the formula:

[0040] Among them: Ps represents the status priority, Indicates the state type indicator (such as unstable state is 1, stable state is 0), D indicates the state duration, and is the weight coefficient. The significance of this formula lies in comprehensively considering the impact of state type and duration on priority. For example, a long-lasting unstable state will receive a higher priority score than a short-lived unstable state. The value of the weight coefficient needs to be adjusted according to the actual application scenario. For example, in applications with high security requirements, the state type indicator may be given a greater weight. The addition of priority tags enables the system to distinguish the urgency of different states, providing a basis for subsequent resource scheduling and parameter adjustments.

[0041] The system classifies resource consumption types based on multi-dimensional navigation data sources. Resource consumption types include two main categories: computing resource consumption and communication resource consumption. Computing resource consumption refers to the system's demand for computing resources for data processing, algorithmic operations, and other aspects. Quantitative metrics include CPU utilization, memory usage, and computational latency. Communication resource consumption refers to the system's demand for network resources for data transmission and communication interaction. Quantitative metrics include bandwidth utilization, data transmission volume, and communication latency. This classification process is achieved by monitoring system resource usage. The system regularly collects various resource usage metrics and classifies them into high or low consumption levels based on preset thresholds. For example, if CPU utilization exceeds 70% for a continuous period of time, the current computing resource consumption is classified as high computing resource consumption; if bandwidth utilization is less than 20%, communication resource consumption is classified as low communication resource consumption. The classification results are also stored as time series labels, with each time point labeled with a corresponding resource consumption type.

[0042] Based on the resource consumption type classification results, the system adds a resource type tag to each resource consumption type. A resource type tag is an identifier that describes resource consumption characteristics, identifying not only the type of consumption but also the level of consumption. For example, a "High Compute Resources" tag indicates high compute resource consumption, while a "Low Communication Resources" tag indicates low communication resource consumption. Adding tags enables the system to quickly identify current resource status, providing information support for subsequent resource management decisions.

[0043] Based on the positioning status type and resource type tag, the system triggers corresponding positioning parameter adjustment signals. When an unstable positioning state is detected, the system generates a first positioning parameter adjustment signal. This detection process is achieved by real-time monitoring of the positioning status tag. The system continuously checks the current status tag and immediately triggers signal generation if it detects an unstable state. The first positioning parameter adjustment signal contains information about the unstable state, such as the start time, duration, and severity. This information assists in subsequent parameter adjustment decisions. When a change in resource consumption type is detected, the system generates a second positioning parameter adjustment signal. A change in resource consumption type refers to a shift from high to low or from low to high resource consumption levels, which may affect system performance and stability. This detection mechanism is implemented by comparing resource type tags at adjacent time points. If a change in the tag is detected, a signal is generated. The second positioning parameter adjustment signal contains detailed information about the resource type change, such as the resource consumption levels before and after the change and the time of the change.

[0044] Based on the first positioning parameter adjustment signal and the second positioning parameter adjustment signal, the system will generate a positioning parameter replacement node. The positioning parameter replacement node is a specific point on the timeline that marks the moment when the positioning parameters need to be adjusted. The generation of the node is based on the time information and content information of the signal. The system will determine the timing of node generation based on the severity and urgency of the signal. There are two types of nodes: parameter adjustment nodes within the same stable positioning state and parameter reset nodes when switching between different positioning state types. Parameter adjustment nodes within the same stable positioning state are used to fine-tune and optimize parameters while maintaining the current positioning state; parameter reset nodes when switching between different positioning state types are used to make larger adjustments to parameters when the state undergoes fundamental changes. The node generation logic comprehensively considers multiple factors, such as the importance of the signal, the current load of the system, historical adjustment effects, etc., to ensure that the timing and magnitude of parameter adjustments are just right.

[0045] Example 4: Positioning offset parameters are calculated based on positioning parameter change nodes and used to optimize positioning control. Positioning parameter change nodes are specific time points or event points derived from parameter adjustment signals. These nodes identify critical moments when positioning parameter adjustments are required. The system extracts parameter adjustment nodes from the node sequence that occur within the same stable positioning state. These nodes typically correspond to moments when parameters are fine-tuned while maintaining the current positioning state. The extraction process is based on matching timestamps and state labels. The system checks the timestamp of each node to confirm whether the positioning state label corresponding to that time point is stable and whether the duration of this stable state has exceeded a minimum threshold to ensure state stability. Nodes that meet these conditions are marked as identical positioning parameter change nodes, meaning that these nodes all require parameter adjustment within the same positioning state. The system extracts parameter reset nodes that occur when switching between different positioning state types. These nodes correspond to moments when fundamental changes in the positioning state require parameter reset. The extraction process is achieved by detecting change points in the state labels. The system identifies the time points when state types transition and marks parameter reset nodes at these time points. These nodes are labeled as nodes with different positioning parameter changes, indicating that these nodes are associated with positioning state transitions and parameter reconfigurations. The entire extraction and labeling process requires precise time synchronization and state tracking to ensure that each node is correctly classified and labeled.

[0046] After extracting and marking nodes, the system begins calculating positioning offset parameters. For node changes with the same positioning parameters, the system calculates the parameter offset within the stable positioning state. This offset reflects the magnitude of parameter adjustment required to maintain a stable state. The calculation is based on parameter value changes over the node time series. The system compares parameter differences between adjacent nodes and determines the offset using differential or proportional change calculations. For example, for the cutoff frequency parameter of a positioning filter, the system calculates the absolute difference or relative rate of change between the current node value and the previous node value. This difference represents the parameter offset within the current stable state. All these offsets are recorded as the first positioning offset parameter and reflect the natural drift of the parameter or the need for gradual adjustment within the stable state. For node changes with different positioning parameters, the system calculates the parameter offset when switching between positioning states. This offset reflects the magnitude of parameter adjustment required during the state transition. The calculation is based on a comparison of parameter values ​​before and after the state switch. The system extracts the parameter value at the last time point before the state switch and the parameter value at the first time point after the state switch and calculates the difference between the two. For example, when switching from a stable to an unstable state, the system may need to significantly adjust the process noise parameters of the Kalman filter. The magnitude of this adjustment is the offset of the parameter during the state transition. All of these offsets are recorded as second positioning offset parameters, which reflect the need for sudden parameter adjustment during state transitions. The system combines the first and second positioning offset parameters to form a complete set of positioning offset parameters. This combination can be simple vector concatenation or weighted fusion, depending on the importance and reliability of the different offset parameters. The combined positioning offset parameters provide a comprehensive view of system parameter changes, including gradual changes in the stable state and sudden changes during state transitions (see Table 1).

[0047] Table 1: Parameter replacement node data is as follows.

[0048] Node ID Timestamp Node Type Positioning status Parameter Type Original parameter value New parameter value Offset N001 2025-06-1008:30:15 Same Node steady state Filter cutoff frequency 1.5Hz 1.8Hz 0.3Hz N002 2025-06-1008:35:22 Same Node steady state Data fusion weight 0.6 0.7 0.1 N003 2025-06-1008:40:18 Different nodes Stable → Unstable Process noise parameters 0.01 0.05 0.04 N004 2025-06-1008:45:30 Different nodes Unstable → Stable Measurement noise parameters 0.02 0.01 -0.01 Based on the positioning offset parameters, the system optimizes positioning control. During the optimization process, it analyzes reproducible patterns in the positioning offset parameters. These patterns are parameter variation patterns that recur across multiple offset instances. The system uses pattern recognition algorithms to analyze the offset parameter sequence and identify recurring patterns under similar conditions. For example, the system may discover that each time a transition from a stable state to an unstable state requires an increase in the process noise parameter by 0.03 to 0.05. This range of values ​​represents a reproducible pattern. From these reproducible patterns, the system extracts migration positioning parameters, which are universal parameter adjustments that can be applied to similar scenarios. The extraction of migration positioning parameters relies on statistical analysis methods, such as calculating the mean, variance, or confidence interval of the offset, to determine the most representative parameter values. The system also analyzes specific patterns in the positioning offset parameters. These patterns are unique parameter changes that occur only under specific conditions. Specific patterns are often associated with abnormal situations or special environmental conditions, such as the need for special parameter adjustments when encountering strong water flow disturbances. From these specific patterns, the system extracts independent positioning parameters, which are specialized adjustments for specific scenarios and are not universally applicable. Based on the migration positioning parameters, the system sets a global positioning benchmark, which is a set of basic parameter values ​​that provides a default configuration solution for the system.

[0049] The global positioning benchmark is set based on the statistical characteristics of the migration positioning parameters, such as using the mode or median of the offset as the benchmark value. Based on the independent positioning parameters, the system dynamically adjusts the parameters for specific positioning scenarios. This adjustment process is adaptive. The system monitors environmental conditions and operating status in real time and automatically applies the corresponding independent positioning parameters when a specific scenario is detected. The system combines the global positioning benchmark and dynamically adjusted parameters to achieve adaptive positioning control. This combination creates a hierarchical control structure, with the global benchmark providing a stable foundation and dynamic adjustment providing flexible response, enabling the system to maintain optimized positioning performance under various operating conditions. The entire optimization process is ongoing, and the system will continuously learn new offset patterns, update the migration positioning parameters and independent positioning parameters, and gradually improve the accuracy and adaptability of control.

[0050] Example 5: Set up a warning mechanism and exit mechanism for the high-precision positioning state, and optimize the data processing order based on these mechanisms. The design of the warning mechanism is based on continuous monitoring of the operating status of the navigation system. The system will analyze the data streams from multi-source sensors in real time, including the output of the inertial measurement unit, the readings of the acoustic positioning system, and the availability indicators of satellite navigation signals. After preliminary processing, these data generate a series of performance indicators, such as the eigenvalues ​​of the covariance matrix of the position estimate, the filter residual statistics, and the time-varying characteristics of the signal strength. The determination of the warning conditions depends on the comprehensive evaluation of these performance indicators. When the key indicators exceed the preset threshold range, the system triggers the warning state. The evaluation process adopts a weighted summation model:

[0051] Among them: W represents the comprehensive early warning index, The standardized value of the performance index, Indicates the weight coefficient of the corresponding indicator. The value range of is [0,1]. The larger the value, the more serious the deviation of the indicator from the normal level. The value of reflects the relative importance of different indicators in early warning judgment, which is usually determined by domain experts based on actual application scenarios. Indicates the total number of performance indicators included in the evaluation. When the value exceeds the preset warning threshold, the system determines that it is necessary to enter the high-precision positioning state. For example, when the uncertainty of the position estimate increases significantly, as manifested by the main eigenvalue of the covariance matrix continuously exceeding the normal range, or when the consistency between multi-sensor data suddenly decreases, as manifested by the statistical characteristics of the filter residual becoming abnormal, the system will determine that the current positioning accuracy is at risk. The triggering of the warning mechanism not only considers the instantaneous indicator value, but also analyzes the time series characteristics of the indicator, and uses a sliding window statistical method to identify trend changes to avoid false triggering due to short-term interference. Once the warning conditions are met, the system immediately initiates the state transition process and switches the positioning system of the underwater operating platform to high-precision positioning mode. This mode is characterized by higher computing and communication resource investment, and aims to restore and maintain positioning accuracy through enhanced data processing capabilities.

[0052] After entering the high-precision positioning state, all newly arriving navigation data is entered into a priority processing queue. This queue is managed using a priority-based scheduling strategy. Each data packet is assigned a priority weight based on the importance of its source sensor, data freshness, and expected contribution to positioning accuracy. Data freshness is determined by the difference between the acquisition timestamp and the current time; a smaller difference indicates higher freshness. The expected contribution is assessed based on the data source's historical performance record and an estimated reliability under current environmental conditions. The priority weight calculation integrates these factors, generating a priority score for each data packet. Packets with higher scores are prioritized for processing. This optimized processing order ensures that data that contributes most to improving positioning accuracy is utilized promptly, while also ensuring that lower-priority data is not completely neglected. Instead, a balanced processing efficiency and positioning performance are achieved through a reasonable scheduling strategy. Status priority tags and resource type tags play a key role in this process, providing complementary information to the priority score. The status priority tag indicates the urgency of the current positioning state. In high-urgency states, the system further reduces the processing delay of high-priority data. Resource type tags reflect the computing and communication resource requirements of different data processing tasks. The system dynamically adjusts scheduling strategies based on available resources, prioritizing resource-efficient data processing tasks when resources are scarce. Implementing this entire priority processing sequence requires efficient queue management algorithms and a real-time scheduler to maximize the performance of the positioning system within limited resource constraints.

[0053] The exit mechanism is designed to correspond to the early warning mechanism, and its triggering conditions are based on the recovery of positioning performance and resource constraints. The system continuously monitors positioning accuracy indicators. When these indicators remain stable and within the normal range, the system determines that the high-precision positioning state can be exited. The exit decision not only considers accuracy indicators but also evaluates resource usage. If maintaining high-precision state leads to excessive resource consumption for a long time, the system may trigger the exit mechanism to ensure stable operation of the entire system. The exit process adopts a gradual approach, gradually reducing data processing priority and resource allocation to avoid performance fluctuations caused by sudden state transitions. When the exit mechanism is triggered, the underwater operation platform exits the high-precision positioning state, and newly collected navigation data no longer enters the priority processing sequence, but enters the normal processing flow. The normal processing sequence uses a standard data processing strategy, no longer giving priority to specific data, but processing according to chronological order or default scheduling. This processing method is relatively resource-efficient and more suitable for normal operation. The implementation of the entire early warning and exit mechanism requires careful parameter tuning and state management to ensure that the system can respond promptly to changes in positioning performance while maintaining a good balance between resource constraints and positioning accuracy.

[0054] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0055] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A high-precision positioning method for an integrated navigation underwater working platform, characterized in that: The following steps are involved: Acquire a multi-dimensional navigation data source of the underwater operation platform, wherein the multi-dimensional navigation data source includes inertial navigation data, sonar positioning data, and satellite positioning data; Based on the multi-dimensional navigation data source, data fusion processing is performed to generate a navigation data warehouse; extracting a navigation feature matrix based on the navigation data warehouse; Determining a feature link graph of a multidimensional navigation data source based on the navigation feature matrix; Based on the feature linkage map, screening core positioning feature sequences; Based on the core positioning feature sequence, the positioning process of the underwater operating platform is controlled.

2. The high-precision positioning method for an integrated navigation underwater work platform according to claim 1, characterized in that: The extracting of the navigation feature matrix based on the navigation data warehouse includes: separating navigation data of various dimensions from the navigation data warehouse; The navigation data of each dimension is subjected to feature extraction processing to generate a dimension feature vector; the dimension feature vectors of all dimensions are integrated to construct a navigation feature matrix.

3. The high-precision positioning method for an integrated navigation underwater work platform according to claim 2, characterized in that: Determining the characteristic link map of the multi-dimensional navigation data source based on the navigation characteristic matrix includes: selecting a target dimension in the navigation characteristic matrix and obtaining a dimensional characteristic vector of the target dimension; Calculating the feature correlation between the navigation features in the dimensional feature vector; generating a dimensional link graph of the target dimension based on the feature correlation; Traverse all dimensions and repeat the steps of generating dimension link diagram; Based on the dimension link graph of all dimensions, the feature link graph of the multi-dimensional navigation data source is generated by connecting the feature associations of the key navigation features.

4. The high-precision positioning method for an integrated navigation underwater work platform according to claim 1, characterized in that: The screening of the core positioning feature sequence based on the feature link graph includes: extracting the dimensional entropy value corresponding to each navigation feature in the navigation feature matrix; Extracting the feature correlation degree of each navigation feature from the feature link graph; Calculate the core entropy value of each navigation feature by combining the dimension entropy value and the feature correlation degree; Based on the core entropy value, filtering out navigation features that are higher than a preset threshold as core positioning features; Aggregate all core positioning features to generate a core positioning feature sequence.

5. The high-precision positioning method for an integrated navigation underwater work platform according to claim 4, characterized in that: The process of controlling the positioning of the underwater operating platform based on the core positioning feature sequence includes: analyzing common patterns in the core positioning feature sequence and extracting reusable positioning parameters; Analyze the difference patterns in the core positioning feature sequence and extract scene-specific positioning parameters; Based on the reusable positioning parameters and the specific scenario positioning parameters, the positioning parameters of the underwater operation platform are adjusted.

6. The high-precision positioning method for an integrated navigation underwater work platform according to claim 1, characterized in that: Also includes: Based on the navigation data warehouse, classify the positioning state type of the underwater working platform, wherein the positioning state type includes a stable positioning state and an unstable positioning state; Based on the positioning status type, attach a status priority tag; Classifying resource consumption types based on the multi-dimensional navigation data source, the resource consumption types including computing resource consumption and communication resource consumption; Based on the resource consumption type, a resource type tag is added.

7. The high-precision positioning method for an integrated navigation underwater work platform according to claim 6, characterized in that: Based on the positioning state type and the resource type tag, triggering a positioning parameter adjustment signal includes: generating a first positioning parameter adjustment signal when an unstable positioning state is detected; generating a second positioning parameter adjustment signal when a change in the resource consumption type is detected; generating a positioning parameter change node based on the first positioning parameter adjustment signal and the second positioning parameter adjustment signal; The positioning parameter replacement node includes a parameter adjustment node within the same stable positioning state and a parameter reset node when switching between different positioning state types.

8. The high-precision positioning method for an integrated navigation underwater work platform according to claim 7, characterized in that: Calculating the positioning offset parameter based on the positioning parameter change node includes: extracting parameter adjustment nodes within the same stable positioning state and marking them as same positioning parameter change nodes; Extract the parameter reset nodes when switching between different positioning state types and mark them as different positioning parameter replacement nodes; Replace the node based on the same positioning parameters, calculate the parameter offset in the stable positioning state, and record it as the first positioning offset parameter; Based on the node replacement with different positioning parameters, the parameter offset when the positioning state type is switched is calculated and recorded as the second positioning offset parameter; The first positioning offset parameter and the second positioning offset parameter are combined to generate a positioning offset parameter.

9. The high-precision positioning method for an integrated navigation underwater work platform according to claim 8, characterized in that: The optimizing positioning control based on the positioning offset parameters includes: analyzing reusable patterns in the positioning offset parameters to extract migration positioning parameters; Analyze specific patterns in positioning offset parameters and extract independent positioning parameters; Based on the migration positioning parameters, set the global positioning benchmark; Dynamically adjust parameters for specific positioning scenarios based on independent positioning parameters; Combining the global positioning reference and dynamically adjusting parameters, adaptive positioning control is achieved.

10. The high-precision positioning method for an integrated navigation underwater working platform according to claim 1, characterized in that: Also includes: Establish early warning and exit mechanisms for high-precision positioning status; When the early warning mechanism is triggered, the underwater operating platform enters a high-precision positioning state, and all new navigation data enters a priority sequence for processing; Optimize sequence processing order based on status priority tags and resource type tags; When the exit mechanism is triggered, the underwater operating platform exits the high-precision positioning state and the new navigation data enters the normal processing sequence.

Images