Marine resource exploration positioning method and system based on inertial navigation

Abstract

The invention relates to the technical field of marine resource exploration, and discloses a marine resource exploration positioning method and system based on inertial navigation. The method comprises the following steps: acquiring a multi-axis acceleration and angular velocity sequence of movement of exploration equipment; dividing and clustering local motion intervals, and identifying abnormal motion feature subsequences; calculating a motion stability index of the characteristic sub-sequence, and extracting an acceleration peak attenuation characteristic and a time sequence interval variation characteristic; based on the difference, the mean value of the motion stability indexes and the mean value of the morphological similarity of the feature subsequences, determining a motion cycle confusion coefficient; constructing a joint trajectory matrix and performing orthogonal decomposition, and calculating a core component proportion and a component discrete index; and generating a positioning quality coefficient by combining the joint trajectory matrix data correlation, the motion period chaos coefficient, the core component proportion and the component discrete index, and calibrating the marine resource exploration position according to the positioning quality coefficient. The method improves the accuracy and stability of inertial navigation positioning in a complex marine environment.

Description

Technical Field

[0001] The present invention relates to the technical field of marine resource exploration, and in particular to a marine resource exploration and positioning method and system based on inertial navigation. Background Art

[0002] In the field of marine resource exploration, accurate positioning technology is a prerequisite for obtaining effective exploration data. Currently, commonly used marine positioning methods rely on satellite navigation or underwater acoustic positioning systems, but these technologies often face numerous limitations in complex marine environments. For example, satellite signals are easily blocked by terrain in offshore areas, resulting in a significant decrease in positioning accuracy. Furthermore, in the high-pressure environment of the deep sea, acoustic positioning systems are susceptible to interference from changes in water temperature and salinity, which not only increases positioning delays but can also lead to significant errors due to signal attenuation. Inertial navigation technology, as an autonomous positioning method that does not rely on external signals, is gaining increasing attention in marine exploration. By integrating accelerometers and gyroscopes, it can collect real-time acceleration and angular velocity information of the device's motion, thereby inferring the motion trajectory. However, traditional inertial navigation methods are susceptible to factors such as sensor drift and ocean turbulence when operating continuously for long periods of time, and the accumulated positioning error increases over time. This is especially true when exploration equipment is performing complex maneuvers, such as fixed-point sampling by deep-sea probes or trajectory planning for underwater robots. The dramatic changes in instantaneous acceleration and angular velocity can lead to nonlinear characteristics in the motion sequence. Traditional linear filtering algorithms have difficulty accurately capturing these characteristics, further reducing positioning accuracy. Existing technologies are insufficiently capable of identifying abnormal motion in marine environments. During operation, exploration equipment may experience irregular motion due to undercurrents, mechanical vibrations of the equipment, and other conditions. If these abnormal motion data are not effectively identified and processed, they will be directly mixed into the normal trajectory calculation, resulting in deviations in the positioning results. At the same time, most inertial navigation methods only analyze acceleration or angular velocity sequences separately, ignoring the inherent correlation between the two, making it difficult to fully reflect the actual motion state of the equipment, which also limits the improvement of positioning accuracy. Therefore, how to combine multi-dimensional motion information to effectively identify abnormal motion and correct positioning errors has become a key issue in improving the inertial navigation positioning performance of marine resource exploration. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for marine resource exploration and positioning based on inertial navigation to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides a method for marine resource exploration and positioning based on inertial navigation, the method comprising: Collect multi-axis acceleration sequences and angular velocity sequences generated when exploration equipment moves in the marine environment; The local motion intervals are divided according to the distribution characteristics of specific fluctuation patterns in the acceleration sequence. The motion sequence is clustered based on the spatiotemporal correlation of the local motion intervals to identify characteristic subsequences that represent abnormal motion. The motion stability index of each characteristic subsequence is calculated by combining the degree of fluctuation of the acceleration trend and the consistency level of angular velocity changes within the characteristic subsequence. Extract the attenuation characteristics of the acceleration peak and the variation characteristics of the time interval between the peaks in the characteristic subsequence, and obtain the acceleration attenuation difference and time interval difference respectively; The motion period chaos coefficient of the exploration equipment is determined based on the acceleration attenuation difference, time interval difference, the mean motion stability index of all characteristic subsequences, and the mean morphological similarity between characteristic subsequences. Construct a joint trajectory matrix of the acceleration sequence and the angular velocity sequence, perform orthogonal decomposition on the joint trajectory matrix to obtain decomposition components, and calculate the core component ratio and component dispersion index of the trajectory decomposition based on the concentration distribution, dispersion degree and variance characteristics of the decomposition components. The positioning quality coefficient of the exploration equipment is generated according to the correlation strength of each row of the joint trajectory matrix, the motion cycle chaos coefficient, the core component ratio and the component dispersion index; the marine resource exploration position is calibrated and positioned based on the positioning quality coefficient.

[0005] Preferably, identifying a characteristic subsequence representing abnormal motion includes: Obtaining the spatiotemporal coupling matrix of the acceleration gradient field and the angular velocity gradient field in the characteristic subsequence; Detecting the coordinates of critical points in the coupling matrix where the gradient direction abruptly exceeds a preset threshold; Divide the equipment motion abnormality area according to the spatial distribution density and time continuity characteristics of the critical point coordinates; Extract the extreme value of the second-order derivative of acceleration and the mutation point of the angular velocity spectrum at the boundary of the abnormal area; Based on the spatiotemporal correlation between extreme points and mutation points, a set of equipment motion mutation feature sets is generated; Features in the motion mutation feature set that have a duration exceeding a preset period and a spatial coverage greater than a preset radius are marked as persistent abnormal areas.

[0006] Preferably, the method for obtaining the motion stability index is: For the target exploration area, the acceleration mean sequence and angular velocity mean sequence of its associated characteristic subsequence are obtained; Based on the fluctuation characteristics of the acceleration trend in the characteristic subsequence of the target exploration area and the difference characteristics of the local peak and valley values ​​in the angular velocity sequence, the axial oscillation coefficient and the dynamic offset coefficient are calculated respectively; The motion stability index of the target exploration area is obtained by multiplying the axial oscillation coefficient of the target exploration area by the dynamic offset coefficient; the axial oscillation coefficient is jointly determined by the variance of the acceleration sequence and the covariance of the angular velocity sequence; the dynamic offset coefficient is calculated by the ratio of the difference between the first and last accelerations of the characteristic subsequence and the angular velocity integration result.

[0007] Preferably, the method for obtaining the dynamic offset coefficient is: Extract the acceleration sequence mean of all exploration equipment in the target exploration area as the benchmark motion parameter; The exploration equipment is sorted in the order of equipment numbers to generate a first equipment sequence; and the equipment is sorted in descending order according to the motion stability index of each equipment to generate a second equipment sequence; If the device numbers at the same serial number position in the first device sequence and the second device sequence are the same, the device is marked as a matching device, and the ratio of the number of matching devices to the total number of devices is calculated as the first matching degree; The unmatched device numbers in the first device sequence are replaced by the Euclidean distance between their acceleration sequences and the reference motion parameters to generate a first distance sequence; the unmatched device numbers in the second device sequence are replaced by the Manhattan distance between their angular velocity sequences and the reference motion parameters to generate a second distance sequence; Calculating the cosine similarity between the first distance sequence and the second distance sequence as a second matching degree; The weighted product of the first matching degree and the second matching degree is used as the equipment coordination weight; the motion stability index of the target exploration area is multiplied by the equipment coordination weight to obtain the positioning coordination coefficient of the target exploration area.

[0008] Preferably, the method for obtaining the trajectory feature segment is: Convert the multi-device motion data aligned in time dimension into a three-dimensional tensor structure; Perform low-rank sparse decomposition on the three-dimensional tensor to obtain a low-rank tensor representing normal motion patterns and a sparse tensor containing abnormal motion features; Based on the abnormal entries marked by sparse tensors, the acceleration offset and angular velocity mutation value of each device in the corresponding time window are extracted; The abnormal entries are divided into equipment failure type abnormalities and environmental interference type abnormalities through clustering algorithm. Equipment failure type abnormalities are associated with the sudden change characteristics of acceleration sequence, while environmental interference type abnormalities are associated with the continuous offset characteristics of angular velocity sequence. The sparse components corresponding to the equipment failure type anomaly are used as the first anomaly subset, and the sparse components corresponding to the environmental interference type anomaly are used as the second anomaly subset; the first anomaly subset is used to trigger the equipment calibration instruction, and the second anomaly subset is used to trigger the path replanning instruction.

[0009] Preferably, the method for generating the exploration scheduling configuration table is: Call the equipment motion state data stored in the leader node and integrate the acceleration sequence mean and angular velocity covariance matrix of all exploration equipment in the current exploration cycle; Receive the global positioning parameter set broadcast by the leader node, which includes the average motion stability index of each device in the current exploration cycle, the target positioning coordinates of the next cycle generated by the leader node based on the device motion status data, and the verification identifier of the candidate leader node; Calculate the target location coordinates to be verified based on the global positioning parameter set, and verify whether the deviation value between the target location coordinates and the next cycle target location coordinates is within the threshold range, and verify the authority identification of the candidate leader node; When both the deviation verification and the authority verification are passed, a positioning confirmation signal is sent to the leader node; After receiving the scheduling instruction issued by the leader node, the candidate leader node is set as the control node of the next exploration cycle, and the target positioning coordinates and equipment motion status data are written into the exploration scheduling configuration table.

[0010] Preferably, the method for obtaining the positioning quality coefficient is: Calculate the absolute value of the mutual information entropy between each row of data in the joint trajectory matrix and other rows of data, and take the mean of the mutual information entropy of each row as the matrix autocorrelation; The positioning quality coefficient is calculated as follows: multiply the core component ratio by the component dispersion index, add the matrix autocorrelation, and then divide by the sum of the motion cycle chaos coefficient and the preset adjustment factor; the core component ratio is determined by the ratio of the energy of the first K components to the total energy after orthogonal decomposition, and the component dispersion index is calculated by the absolute value of the difference between the main component variance and the secondary component variance.

[0011] Preferably, the calibration and positioning of the marine resource exploration position includes: Calculate the error compensation threshold of the inertial navigation system based on the positioning quality coefficient. The threshold is inversely proportional to the sampling frequency of the device motion data and proportional to the square root of the positioning quality coefficient. The angular velocity integration result is corrected using the error compensation threshold, and the calibrated three-dimensional space coordinates are generated by combining the acceleration double integration result. The real-time positioning trajectory of the exploration equipment is updated based on the calibration coordinates, and the target location coordinates for marine resource exploration are output.

[0012] Preferably, the present invention further includes a marine resource exploration and positioning system based on inertial navigation, which is used to implement the above-mentioned marine resource exploration and positioning method based on inertial navigation, and the system includes: The data acquisition module is used to collect the multi-axis acceleration sequence and angular velocity sequence generated by the exploration equipment when it moves in the marine environment; The interval division and clustering module is used to divide the local motion intervals according to the distribution characteristics of specific fluctuation patterns in the acceleration sequence, cluster the motion sequence based on the spatiotemporal correlation of the local motion intervals, and identify characteristic subsequences that represent abnormal motion; The motion stability index calculation module is used to calculate the motion stability index of each characteristic subsequence by combining the fluctuation degree of acceleration trend and the consistency level of angular velocity change within the characteristic subsequence; The difference extraction module is used to extract the attenuation characteristics of the acceleration peak and the variation characteristics of the peak time interval in the feature subsequence, and obtain the acceleration attenuation difference and time interval difference respectively; A motion period chaos coefficient determination module is used to determine the motion period chaos coefficient of the exploration equipment based on the acceleration attenuation difference, the time interval difference, the mean motion stability index of all characteristic subsequences, and the mean morphological similarity between characteristic subsequences; The trajectory decomposition analysis module is used to construct the joint trajectory matrix of the acceleration sequence and the angular velocity sequence, perform orthogonal decomposition on the joint trajectory matrix to obtain decomposition components, and calculate the core component ratio and component dispersion index of the trajectory decomposition based on the concentration distribution, dispersion degree and variance characteristics of the decomposition components; The positioning calibration module is used to generate the positioning quality coefficient of the exploration equipment based on the correlation strength of each row of data in the joint trajectory matrix, the motion cycle chaos coefficient, the core component ratio and the component discrete index; and calibrate the marine resource exploration position based on the positioning quality coefficient.

[0013] Preferably, the system also includes a data preprocessing module, which is used to filter and denoise the multi-axis acceleration sequence and angular velocity sequence collected by the data acquisition module, eliminate outliers and perform data standardization, and transmit the processed data to the interval division and clustering module.

[0014] Compared with the prior art, the present invention has the following beneficial effects: By processing motion data in multiple dimensions, the accuracy and stability of the positioning process are significantly improved. By dividing local motion intervals and performing cluster analysis, it can accurately identify characteristic subsequences that represent abnormal motion, effectively distinguishing abnormal motion data, avoiding the interference of irregular motion on normal trajectory calculations, and reducing positioning deviations caused by the mixing of abnormal data. In terms of motion feature analysis, the motion stability index is calculated by combining the fluctuations in acceleration trends with the consistency of angular velocity changes. Furthermore, the acceleration peak attenuation characteristics and time interval variation characteristics are extracted to fully capture the detailed characteristics of the device's motion. The combined use of these characteristics provides a more comprehensive description of the device's motion state, overcoming the limitations of traditional methods that rely solely on a single motion parameter and more realistically reflecting the device's motion patterns in complex marine environments. The introduction of the motion cycle chaos coefficient quantifies the regularity of device motion by integrating multiple indicators, including acceleration attenuation differences, time interval differences, mean motion stability index, and mean morphological similarity. This provides an effective basis for assessing the reliability of motion sequences. This coefficient enables quantitative analysis of the overall motion trend when processing nonlinear motion sequences, eliminating the subjectivity of human judgment and enhancing the objectivity of motion state assessment. The construction and orthogonal decomposition of the joint trajectory matrix fuses the information of the acceleration and angular velocity sequences. By calculating the core component ratio and component dispersion index, key information about the motion trajectory can be effectively extracted and redundant data can be filtered out. This multi-parameter joint analysis method fully utilizes the inherent correlation between acceleration and angular velocity, overcoming the shortcomings of traditional methods that process single-dimensional data independently. This enriches the basis for trajectory calculation and improves the ability to depict the actual motion state of the device. The positioning quality coefficient is generated based on the correlation of joint trajectory matrix data, the motion cycle chaos coefficient, the core component ratio, and the component dispersion index. This is used to calibrate the positioning results, achieving multi-factor collaborative correction. This correction method comprehensively considers the correlation and regularity of motion data and the proportion of key information. It can adjust positioning parameters in a targeted manner, maintaining high positioning accuracy in the presence of interference factors such as sensor drift and ocean turbulence, and adapting to the complex and changing marine exploration environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a working principle diagram of the marine resource exploration and positioning method based on inertial navigation according to the present invention; Figure 2 Flowchart for abnormal motion feature subsequence identification; Figure 3 Flowchart for coordination of dynamic offset coefficients and equipment; Figure 4 Flowchart for trajectory feature segmentation and anomaly classification; Figure 5 Flowchart generated for the survey scheduling configuration table. DETAILED DESCRIPTION

[0016] 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.

[0017] See also Figure 1The present invention provides a method for marine resource exploration and positioning based on inertial navigation, the method comprising: By collecting the multi-axis acceleration sequence and angular velocity sequence generated by the exploration equipment when it moves in the marine environment, high-precision positioning is achieved by combining signal processing and motion feature analysis. The specific process includes: First, the acceleration sequence is analyzed, and local motion intervals are divided according to the distribution characteristics of specific fluctuation patterns. Characteristic subsequences of abnormal motion are identified through spatiotemporal correlation clustering. The motion stability index of these characteristic subsequences is then calculated, and the motion period chaos coefficient is determined by combining the attenuation characteristics of the acceleration peak and the differences in the temporal intervals. Furthermore, a joint trajectory matrix of acceleration and angular velocity is constructed, and the core component proportions and component dispersion indices are extracted through orthogonal decomposition. Finally, a positioning quality coefficient is generated based on the matrix correlation, motion period chaos coefficient, and decomposition characteristics to complete the calibration of the exploration position.

[0018] Example 1: See Figure 2 When identifying characteristic subsequences representing abnormal motion, the collected multi-axis acceleration sequence and angular velocity sequence are first preprocessed to eliminate high-frequency noise and baseline drift. The acceleration gradient field is calculated using the central difference method. Data within a certain time window near each sampling point is taken, and the difference values ​​before and after are calculated to form a gradient sequence. The calculation method of the angular velocity gradient field is similar, but the vector characteristics of the angular velocity must be considered. The gradients of the three-axis angular velocity components are calculated separately and then synthesized. The construction of the spatiotemporal coupling matrix is ​​generated by aligning the acceleration gradient field and the angular velocity gradient field in the time dimension and then superimposing them. The rows of the matrix represent time points, and the columns represent the gradient components of different axes.

[0019] To detect sudden changes in the gradient direction within the coupling matrix, a sliding time window is used to traverse the matrix, calculating the rate of change of the gradient direction within each window. The rate of change is measured by the angle between the gradient vectors at adjacent time points. If the angle exceeds a preset threshold, the time point is recorded as a critical point. The preset threshold is determined based on the motion characteristics of the exploration equipment and is typically derived from historical data. The spatial distribution density of the critical point coordinates is calculated using kernel density estimation, generating a density distribution map on a two-dimensional time-space plane. High-density areas are identified as potential areas of anomalous motion.

[0020] Anomaly regions are delineated using a density-based clustering algorithm. After inputting the coordinates of critical points into the algorithm, clusters are formed based on the neighborhood radius and minimum number of points. The choice of neighborhood radius influences the granularity of the anomaly region. A smaller radius can detect localized mutations, while a larger radius is suitable for identifying persistent anomalies. Temporal continuity is assessed by examining the continuity of time points within a cluster. Any discrete points separated by excessively large time intervals are excluded. The extreme values ​​of the second-order derivative of acceleration at the boundaries of the anomaly region are derived by polynomial fitting and differentiation. The fitting order is adaptively selected based on the smoothness of the data. Angular velocity spectrum mutation points are detected using time-frequency analysis. Spectral energy is calculated using a short-time Fourier transform to identify frequency points with sudden energy increases.

[0021] The generation of a set of device motion mutation feature sets is based on the spatiotemporal correlation analysis of extreme points and mutation points. When constructing a graph model, nodes represent detected extreme points or mutation points, and edge weights are determined by a comprehensive function of the time difference, spatial distance, and magnitude of change between the two points. The community discovery algorithm divides the graph into several subgraphs, each corresponding to a motion mutation pattern. Persistent anomaly areas are marked based on two conditions: the duration must exceed a certain multiple of the device sampling period, and the spatial coverage must be calculated using a convex hull algorithm and compared to a preset radius.

[0022] The motion stability index is calculated for each characteristic subsequence. When calculating the axial oscillation coefficient, the variance of the acceleration sequence reflects the severity of the motion fluctuations, while the covariance of the angular velocity sequence characterizes the coordination of different axial motions. When calculating the dynamic offset coefficient, the difference in acceleration between the beginning and end of the characteristic subsequence reflects the overall trend of the motion, while the integral of the angular velocity reflects the cumulative rotation of the device. The motion stability index combines the axial oscillation coefficient and the dynamic offset coefficient; lower values ​​indicate more unstable motion corresponding to the subsequence.

[0023] In practical applications, the identification of abnormal motion feature subsequences and the calculation of the motion stability index require iteration. The initial local motion intervals may contain errors due to noise or interference. After selecting the more reliable intervals using the motion stability index, the raw data can be clustered again to improve the accuracy of anomaly detection. The construction of the spatiotemporal coupling matrix and the detection of sudden changes in gradient direction require high computing resources, which can be accelerated using parallel computing.

[0024] During the demarcation and labeling of abnormal regions, the parameter settings of the density clustering algorithm significantly impact the results. A too small neighborhood radius can fragment the abnormal region, while a too large radius can obscure local details. The assessment of temporal continuity features requires integration with the device's kinematic model to avoid misinterpreting brief pauses during normal motion as abnormalities. Detection of extreme values ​​in the second-order derivative of acceleration is sensitive to data smoothness and requires a balance between noise suppression and feature preservation. Detection of abrupt changes in the angular velocity spectrum is limited by the sampling frequency, and the spectral resolution of high-frequency motion must meet analysis requirements.

[0025] When calculating the motion stability index, the weighting of the axial oscillation coefficient and the dynamic offset coefficient can be adjusted based on the specific application scenario. For scenarios requiring high-precision attitude control, the weighting of angular velocity-related terms can be increased; for scenarios primarily focused on displacement measurement, the emphasis can be placed on acceleration characteristics. The dynamic range of the motion stability index must be normalized based on the actual data distribution to ensure comparability between different devices or over different time periods.

[0026] This method can effectively distinguish between abnormal device motion and motion deviations caused by environmental interference in marine resource exploration. By combining the spatiotemporal coupling matrix with gradient directional mutation detection, it can capture hidden anomaly patterns that are difficult to identify with traditional methods. The introduction of a motion stability index enables quantitative assessment of the degree of anomaly, providing a reliable basis for subsequent positioning calibration.

[0027] The identification of abnormal motion feature subsequences and the calculation of motion stability indices can be further expanded. For example, machine learning methods can be introduced to train historical abnormal patterns to improve the efficiency of generating mutation feature sets. The calculation of the motion stability index can be combined with the device dynamics model and include more motion parameters to improve the robustness of the index. The construction method of the spatiotemporal coupling matrix can be optimized, for example, by using wavelet transform instead of gradient calculation to enhance adaptability to non-stationary signals.

[0028] Example 2: See Figure 3 The optimization calculation process of the dynamic offset coefficient involves the collaborative analysis of multiple devices. Its core lies in improving positioning accuracy through the motion consistency evaluation between devices. First, the mean of the acceleration sequence of all exploration devices in the target exploration area is extracted as the benchmark motion parameter. The generation of the benchmark parameter adopts a time alignment strategy, and the acceleration sequence collected by each device is interpolated and aligned according to a unified time axis to ensure the time matching of the data points. The arithmetic mean of the aligned acceleration sequence is calculated dimension by dimension according to the time point to form a benchmark motion parameter sequence. This sequence reflects the overall motion trend of the device group in the exploration area and can be used as a reference standard for the subsequent motion status evaluation of individual devices.

[0029] A first device sequence is generated based on device number order. This sequence maintains the original device numbering, reflecting the initial system configuration order. Simultaneously, each device is sorted in descending order based on its motion stability index to generate a second device sequence. Devices with higher motion stability indices are positioned higher in the second sequence, indicating more stable and reliable motion. Device numbers at the same position in the two sequences are compared; if they match, the device is marked as a match. The ratio of the number of matching devices to the total number of devices is used as the first matching degree, reflecting the degree of consistency between the device number sequence and the motion stability sequence.

[0030] For devices that fail to match at the same position in the two sequences, their motion feature differences are further analyzed. The number of the unmatched device in the first device sequence is replaced with the Euclidean distance between its acceleration sequence and the benchmark motion parameters. The Euclidean distance is calculated for the entire time series, measuring the overall deviation between the acceleration of individual devices and the average acceleration of the group. At the same time, the number of the unmatched device in the second device sequence is replaced with the Manhattan distance between its angular velocity sequence and the benchmark motion parameters. The Manhattan distance is calculated using a point-by-point absolute value accumulation method to highlight the local difference characteristics of the angular velocity sequence. Through this replacement operation, the device number sequence is converted into a distance feature sequence, which facilitates subsequent similarity analysis.

[0031] The cosine similarity between the first and second distance sequences is calculated as the second matching degree. Cosine similarity measures the directional consistency of two distance sequences in vector space. Values ​​closer to 1 indicate more similar trends between the sequences. During distance sequence construction, the Euclidean distance reflects the overall acceleration offset, while the Manhattan distance captures the local fluctuations in angular velocity. The combination of the two comprehensively assesses the difference between the device's motion state and the group benchmark. The calculation of the second matching degree provides a complementary dimension for device coordination assessment, forming a multi-angle evaluation system alongside the first matching degree.

[0032] The first and second matching degrees are weighted and combined to generate a device coordination weight. The weight distribution coefficient can be adjusted based on specific application requirements; generally, the contributions of the two matching degrees are equal. The weighted product comprehensively reflects the degree of consistency between the device numbering sequence and the motion stability sequence, as well as its coordination with the group's motion pattern. A higher device coordination weight indicates that the device's motion performance in the group more closely matches the expected pattern, and its positioning data is more reliable.

[0033] The positioning coordination coefficient for the target exploration area is calculated by multiplying the motion stability index and the device coordination weight. The physical significance of this coefficient lies in its consideration of both the stability of individual devices and the coordination consistency among the group of devices. In collaborative group operations, the positioning coordination coefficient effectively identifies devices that maintain their own stability while coordinating with the group's motion patterns. The positioning data from these devices is therefore more valuable as a reference. The coefficient calculation process utilizes normalization, ensuring comparability between results across different exploration areas or time periods.

[0034] In practical applications, optimizing the dynamic offset coefficient requires addressing the spatiotemporal synchronization of multi-source data. The sampling clocks of individual devices may exhibit slight deviations, necessitating data alignment through timestamp correction techniques. The generation of baseline motion parameters must account for the dynamic changes in the number of devices. When new devices join or old ones exit, the baseline parameters must be rapidly updated. The matching calculation is sensitive to outliers, and clearly unreasonable data points must be removed during preprocessing to avoid interference with statistical results.

[0035] The strategy for allocating device coordination weights can be flexibly adjusted based on task requirements. For applications that prioritize group consistency, the weight of the first degree of matching can be appropriately increased; for applications that focus on individual motion characteristics, the influence of the second degree of matching can be emphasized. The application of the positioning coordination coefficient is not limited to positioning calibration; it can also be used to monitor device health. Devices with consistently low coefficients may have sensor failures or mechanical problems and require timely inspection and maintenance.

[0036] The implementation of collaborative analysis across multiple devices relies on reliable data communication mechanisms. In a distributed exploration system, motion data from each device must be transmitted in real time to a central node for processing. Communication delays can reduce the timeliness of data analysis, necessitating the design of a suitable buffering mechanism to balance real-time performance with computational accuracy. For large-scale device clusters, a layered processing architecture can be employed, performing local analysis first and then global integration to reduce computational complexity.

[0037] The optimization process for the dynamic offset coefficient is adaptive. As exploration operations progress, the system continuously accumulates historical data, allowing for gradual optimization of matching calculation parameters. For example, the detection threshold for sudden changes in gradient direction can be adjusted based on long-term statistical results, ensuring that anomaly identification is more consistent with the actual operating environment. The calculation of equipment coordination weights can also incorporate a sliding time window mechanism to reflect the latest trends in equipment status.

[0038] Example 3: See Figure 4The extraction of trajectory feature segments is based on the tensor representation and analysis of multi-device motion data. The multi-device motion data after time dimension alignment is converted into a three-dimensional tensor structure. The three dimensions of the tensor correspond to the device number, time sampling point and motion parameters, respectively. The motion parameters include three-axis acceleration and three-axis angular velocity, forming a six-dimensional parameter space. In the process of tensor construction, the raw data of each device is first synchronized in time, and the interpolation method is used to ensure that all devices have corresponding sampling data at the same time point. For devices with short communication interruptions, the linear prediction method is used to fill the missing data segments to maintain the continuity of the time series.

[0039] The constructed three-dimensional tensor undergoes a low-rank sparse decomposition, breaking it down into low-rank components representing normal motion patterns and sparse components containing abnormal motion characteristics. The low-rank components are extracted using robust principal component analysis, and the rank of the tensor is minimized through iterative thresholding. Appropriate regularization parameters are set during the calculation of the sparse components to balance the sensitivity and false alarm rate of anomaly detection. The decomposed low-rank tensor reflects the common motion patterns of the exploration equipment group and can be used to establish a background model of environmental interference. The sparse tensor captures the abnormal motion characteristics of individual equipment, providing data support for fault diagnosis.

[0040] Abnormal entries are detected based on the distribution of non-zero elements in a sparse tensor. A threshold is set to filter out significant abnormal entries, and the threshold value is dynamically adjusted based on the statistical characteristics of historical data. For each detected abnormal entry, the device motion parameters within the corresponding time window are extracted. The acceleration offset is calculated by taking the difference between the mean within the window and the long-term trend value, which is obtained through low-pass filtering. The angular velocity mutation value is calculated based on the range within the window, reflecting the intensity of the instantaneous angular velocity fluctuation.

[0041] Anomaly entries are classified using an improved density clustering algorithm, which adaptively determines the neighborhood radius to accommodate anomaly distributions of varying densities. Equipment failure anomalies are characterized by discontinuous step changes in the acceleration sequence, accompanied by random fluctuations in angular velocity. These anomalies are typically caused by sensor failure or mechanical damage and exhibit localized bursts. Environmental interference anomalies are characterized by persistent shifts in the angular velocity sequence, while acceleration remains relatively smooth. These anomalies are often caused by changes in ocean currents or sudden changes in equipment drag and exhibit temporal continuity.

[0042] The generation of the first anomaly subset is for equipment failure type anomalies, and includes all sparse component entries classified as equipment failures. When this subset triggers the device calibration instruction, the system automatically selects the appropriate calibration strategy: for acceleration step mutations, zero bias calibration is performed; for random angular velocity fluctuations, gyroscope stability detection is performed. The adjustment amplitude of the calibration parameters is proportional to the sparse component amplitude of the anomaly entry, achieving a graded response. The generation of the second anomaly subset is for environmental interference type anomalies, and includes all sparse component entries classified as environmental interference. The path replanning instruction triggered by this subset takes into account the duration and spatial distribution range of the anomaly, and adopts the following strategy for motion trajectory optimization: in, represents the path correction angle, is the number of abnormal entries, For the The angular velocity change of the abnormal entry, For the The correction angle is used to adjust the overall movement direction of the device group to avoid the area where the interference persists.

[0043] The choice of tensor decomposition parameters has a significant impact on the analysis results. The rank of the low-rank component determines the accuracy of the representation of normal motion patterns, and the optimal value is determined through cross-validation. The regularization parameter of the sparse component controls the sensitivity of anomaly detection. Too large a parameter can miss subtle anomalies, while too small a parameter can lead to an increase in false positives. In practice, a gradual adjustment strategy is used, with the initial value set based on the signal-to-noise ratio of the device motion data and subsequently dynamically optimized based on feedback from detection results.

[0044] The division of the time window for abnormal entries must take into account the inertial characteristics of device motion. A window that is too short may fragment continuous abnormal processes, while a window that is too long may dilute the abnormal characteristics. The window length setting is related to the device's motion speed. A shorter window is used for fast motion, while a longer window is appropriately extended for slow motion. Acceleration offset and angular velocity mutation values ​​are calculated using an overlapping sliding window method to ensure accurate positioning of the abnormal event boundary.

[0045] The feature extraction process for anomaly classification integrates multi-dimensional information. In addition to basic motion parameter statistical features, frequency domain features and joint time-frequency features are also considered. Frequency domain analysis uses fast Fourier transforms to extract dominant frequency components, while time-frequency analysis uses wavelet transforms to capture transient features. Classifier training utilizes a semi-supervised learning approach, leveraging both labeled anomaly samples from historical data and unlabeled samples from the current test to construct a classification model.

[0046] Device calibration instructions are executed using a hierarchical processing mechanism. Minor anomalies require only software parameter adjustments; severe anomalies trigger a hardware self-check and prompt manual intervention. The calibration process retains a complete record of raw motion data and calibration parameters, providing a long-term record of device health. Path replanning instructions are generated to ensure the overall coordination of the device group, preventing sudden changes in individual device movements that could disrupt the group's movement.

[0047] The analysis results of trajectory feature segments form an effective linkage with other modules. Identified device failure anomalies can trigger the device coordination weight update in Example 2, reducing the contribution of the faulty device to group positioning. The distribution characteristics of environmental interference anomalies can provide a regional risk map for scheduling decisions in Example 4, guiding task allocation for the device group. The confidence index of anomaly classification can be integrated into the positioning quality assessment in Example 5 to form a more comprehensive reliability assessment.

[0048] In actual deployment, operational convenience and interpretability of results must be considered. Anomaly detection results are visualized, highlighting faulty devices and interference areas. Calibration and replanning recommendations are accompanied by detailed rationale to assist operators in decision-making. The system provides a multi-level alert mechanism, with varying response strategies tailored to anomaly severity, balancing the boundaries between automation and manual control.

[0049] The advantage of the three-dimensional tensor representation method is that it fully preserves the spatiotemporal correlation of device motion. Compared to traditional independent analysis of individual devices, it can more effectively identify relevant patterns and abnormal characteristics in group motion. The theoretical framework of low-rank sparse decomposition provides a mathematical foundation for anomaly detection, ensuring the rigor of the analysis process. The diverse design of classification features enhances the ability to identify complex abnormal patterns, reducing false positives and missed detections.

[0050] The implementation process for trajectory feature segment extraction emphasizes systematicity and automation. A complete closed-loop system, from data preprocessing to anomaly response, reduces manual intervention. Parameter selection and threshold setting establish an adaptive mechanism, reducing maintenance costs. Standardized output facilitates integration with other subsystems. These design features ensure the method's practical engineering value and ensure sustained and stable operation in actual exploration missions.

[0051] Example 4: See Figure 5 The generation process of the exploration scheduling configuration table involves collaborative decision-making and data integration among distributed nodes. Its core lies in ensuring the continuity and reliability of exploration missions in dynamic ocean environments through interactive verification between the leader node and candidate nodes. Taking an actual ocean exploration mission as an example, the system includes six underwater exploration vehicles, numbered AUV-01 to AUV-06, coordinated by the master control platform (leader node).

[0052] At the beginning of the exploration cycle, the leader node collects motion status data from each device to form a global view of the current cycle. The acceleration sequence mean is calculated using a sliding window averaging method with a 10-second window length, covering the entire wave motion cycle. The angular velocity covariance matrix is ​​constructed by considering the correlation of the three-axis motion and reflecting the coordinated motion patterns between devices. The following table shows a fragment of device motion status data collected at a specific time.

[0053] Table 1: Shows a fragment of the device motion status data collected at a certain moment:

[0054]

[0055] After analyzing this data, the leader node calculated the mean motion stability index for each device during the current cycle. AUV-04 was identified as a potential anomalous node due to its significantly higher mean acceleration and angular velocity covariance than other devices. Based on this, the leader node generated a set of target positioning coordinates for the next cycle, with AUV-04's target coordinates set to a conservative position that maintained a greater distance from other devices. Simultaneously, the system selected AUV-03, which had the most stable motion, as a candidate leader node and generated a verification identifier for it, including a timestamp and digital signature.

[0056] The global positioning parameter set is broadcast to all nodes via the underwater acoustic communication network. This set uses a compact binary encoding format and consists of three main components: a quantized value of the mean kinematic stability index (6 bytes), the UTM encoding of the target positioning coordinates (12 bytes per device), and a cryptographic hash value of the candidate node verification identifier (16 bytes). Forward error correction is used during transmission to ensure data integrity despite ocean channel attenuation.

[0057] After receiving the broadcast data, each device first verifies the message's digital signature to confirm the authenticity of the source. The AUV-01, acting as the node to be verified, calculates the expected target coordinates based on locally stored historical motion data. This calculation takes into account the combined effects of the device's current velocity vector and the water flow prediction model, generating three sets of possible target position hypotheses. By comparing these with the corresponding values ​​in the broadcast coordinate set, the AUV-01 confirms that its target coordinates have a deviation of 1.2 meters, which falls within the preset 2-meter threshold.

[0058] The authority verification of candidate leader node AUV-03 utilizes a double-check mechanism. First, the digital signature chain in the verification token is verified to confirm that it is signed by both the current leader node and the leader node of the previous cycle. Second, the node's reputation score, recorded in the distributed ledger, is checked. AUV-03's historical mission completion rate is 98%, meeting the minimum requirement for candidate nodes. After both verifications pass, AUV-01 transmits a positioning confirmation signal to the leader node. This signal contains its device ID, a verification timestamp, and a compressed summary of its motion status.

[0059] After receiving confirmation signals from more than half of the devices, the leader node initiates the control transfer process. First, the configuration parameters for the current cycle are frozen, and the dynamic scheduling data in memory is persisted. A new exploration scheduling configuration table is then generated. Key fields include the new leader node ID (AUV-03), the target positioning coordinates of each device, and the SHA-256 checksum of the motion status data. The configuration table is transmitted to the AUV-03 via a secure channel, and a control change notification is simultaneously sent to the entire network.

[0060] In actual operations, AUV-03 immediately executed the initial dispatch decision after taking over leadership. Based on AUV-04's unusual movement characteristics, it issued a deceleration command, forcing it to leave the core area of ​​the formation. Simultaneously, the track spacing between AUV-01 and AUV-02 was adjusted to compensate for the coverage gap caused by AUV-04's displacement. The new dispatch configuration table was updated every 30 seconds to dynamically respond to position drift caused by changing ocean currents.

[0061] The system employs a redundant storage strategy to ensure the reliability of configuration tables. In addition to the master copy stored on the leader node, two other devices store encrypted backups. Each time a configuration is updated, consistency is verified by comparing the checksums of the three copies. When a storage anomaly is detected, a majority-based data recovery mechanism is triggered to prevent single-point failures from causing scheduling information loss.

[0062] The exploration scheduling configuration table's structure is designed to support incremental updates. Fields that remain unchanged between successive cycles are coded with a "same as previous cycle" tag to reduce data transmission. Modified fields use differential encoding, transmitting only the differences from the previous configuration. This optimization keeps the size of a typical update message under 200 bytes, adapting to the low-bandwidth nature of underwater acoustic communications.

[0063] Exception handling mechanisms play a crucial role in the scheduling process. When AUV-05 failed to respond to a configuration update in the third cycle, the system automatically marked it as "pending repair." The leader node first attempted to resend the configuration data via a backup communication channel while simultaneously instructing the neighboring AUV-06 to establish a close-range optical link. After three unsuccessful retries, the system removed the device from the current task list and recalculated the coverage pattern for the remaining devices.

[0064] Configuration table version management utilizes a hybrid logical clock scheme. Each version number combines the leader node ID sequence and a logical counter to ensure correct ordering even in clock-out scenarios. When a device receives a configuration update, it compares the version numbers in lexicographic order to determine whether to apply the update, preventing configuration rollbacks caused by network latency. Historical version data retains a complete record of the last five cycles, supporting fault diagnosis and status rollback.

[0065] Example 5: The calculation process of the positioning quality coefficient is based on the fusion of multi-dimensional motion features, and the dynamic evaluation of positioning accuracy is achieved by quantitatively analyzing the intrinsic characteristics of the device's motion trajectory. The method first processes the data structure of the joint trajectory matrix, where the rows correspond to different time sampling points and the columns contain the measured values ​​of multi-axis acceleration and angular velocity and their derived features. A sliding window mechanism is used in the matrix filling process, and the window width is adaptively adjusted according to the device's motion speed. A narrower window is used during fast motion to capture detailed changes, and an appropriately widened window is used during slow motion to smooth out the effects of noise.

[0066] Matrix autocorrelation is calculated using mutual information entropy as the core metric. This method differs from traditional linear correlation analysis in that it captures the nonlinear dependencies between acceleration and angular velocity sequences. A kernel density estimation model is constructed for each row of data during the calculation, and the bandwidth parameter is automatically adjusted based on the overall matrix range. The absolute value of the mutual information entropy is calculated to avoid information loss due to directional bias. The average value of the mutual information entropy across all rows is ultimately used as the matrix autocorrelation. Higher values ​​indicate greater temporal consistency in the device's motion pattern, and generally improve the reliability of the positioning results.

[0067] The orthogonal decomposition process uses an improved singular value decomposition algorithm, optimizing numerical stability processing for marine exploration scenarios. The joint trajectory matrix is ​​standardized and preprocessed before decomposition to eliminate scale differences caused by different physical dimensions. The proportion of core components is determined using an adaptive energy threshold method, which automatically selects the number of principal components by analyzing the inflection points of the cumulative contribution rate curve of singular values. This method avoids the information loss or redundancy that may be caused by a fixed threshold and can adapt to changes in motion characteristics under different sea conditions. The calculation of the component dispersion index focuses on the distribution differences between the main components and the secondary components, and reflects the concentration of the motion pattern through variance comparison.

[0068] The calculation of the motion period chaos coefficient integrates motion characteristic indicators from multiple sources. Acceleration attenuation differences are analyzed using envelope detection to identify the degree of regularity in energy dissipation during motion. Time interval differences are calculated based on a peak detection algorithm to assess the stability of periodic motion. Morphological similarity between characteristic subsequences is measured using a dynamic time warping algorithm, which can align motion segments of varying lengths for comparison. A weighted fusion of these indicators yields the motion period chaos coefficient, with larger values ​​indicating greater environmental interference.

[0069] The final synthesis of the positioning quality coefficients uses a normalized weighting method. The product of the core component ratio and the component dispersion index reflects the degree of structure of the motion trajectory, and the matrix autocorrelation represents the stability of the time dimension. The two are added together to form the numerator. The denominator is composed of the motion cycle chaos coefficient and the adjustment factor. The adjustment factor is dynamically configured based on the equipment type and water depth to balance the evaluation scale in different operating environments. The calculated coefficient is mapped to a range of 0 to 1. The closer the value is to 1, the higher the credibility of the current positioning data.

[0070] The calculation process for the error compensation threshold incorporates a dynamic adjustment mechanism for the sampling frequency. The threshold amplitude is lowered at high sampling frequencies to avoid overcompensation, while the threshold is appropriately raised at low sampling frequencies to suppress the accumulation of integral errors. The square root relationship between the threshold and the positioning quality factor preserves the nonlinear adjustment characteristics, enabling the system to distinguish between different levels of positioning reliability. In practical applications, this threshold serves as a boundary condition for the angular velocity integral correction, limiting the maximum adjustment range of the compensation algorithm.

[0071] Angular velocity integral correction utilizes a variable-step-size Runge-Kutta method, with the step size adjusted in conjunction with the error compensation threshold. When the positioning quality factor is high, a smaller step size is used for refined integration; as the factor decreases, a larger step size is automatically switched to ensure computational efficiency. Kinematic constraints are introduced during the integration process, triggering a data validity check when angular velocity changes exceed the physically possible range. Acceleration dual integration utilizes velocity damping, with the damping coefficient inversely proportional to the positioning quality factor, enhancing filtering effectiveness when data reliability is low.

[0072] The calibration process for 3D coordinates takes into account the continuity constraints of the device's posture. Quaternion interpolation is used to smooth rotation changes, avoiding gimbal lock issues associated with Euler angle representation. The calibrated coordinate transformation uses the WGS84 ellipsoid model, and elevation calculations integrate pressure sensor data as an auxiliary reference. Real-time positioning trajectories are generated using B-spline curve fitting. Node vectors are dynamically distributed based on the positioning quality coefficient. Sparse nodes are used in reliable data intervals, while node density is increased in uncertain intervals to improve fitting accuracy.

[0073] The application of the positioning quality coefficient extends to system resource allocation strategies. When the coefficient consistently falls below a warning value, sensor calibration procedures are automatically triggered or a backup measurement unit is switched. In collaborative group scenarios, the coefficient serves as a weighting factor in collective decision-making, with observations from highly reliable devices receiving greater weight. Historical coefficient trend analysis is used to predict positioning performance degradation, enabling proactive maintenance procedures to avoid mission interruptions.

[0074] Positioning quality assessment results are visualized using a multi-layer interactive design. The base layer displays device trajectories and coefficient curves, while auxiliary layers provide detailed feature information. A color-coding scheme intuitively distinguishes different reliability levels, and abnormal sections are automatically labeled with timestamps and possible causes. This design allows operators to quickly understand system status and make effective decisions.

[0075] Another key feature of this method is its scalability. By adjusting the feature extraction strategy and weight distribution, it can be adapted to meet the positioning and assessment requirements of various underwater vehicles. A fusion interface with external observation data supports integration with combined navigation systems, forming an enhanced positioning solution based on multi-source information complementarity. These characteristics give it broad application prospects in marine research, resource exploration, underwater engineering, and other fields.

[0076] 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.

[0077] 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 method for marine resource exploration and positioning based on inertial navigation, characterized in that: include: Collect multi-axis acceleration sequences and angular velocity sequences generated when exploration equipment moves in the marine environment; The local motion intervals are divided according to the distribution characteristics of specific fluctuation patterns in the acceleration sequence. The motion sequence is clustered based on the spatiotemporal correlation of the local motion intervals to identify characteristic subsequences that represent abnormal motion. The motion stability index of each characteristic subsequence is calculated by combining the degree of fluctuation of the acceleration trend and the consistency level of angular velocity changes within the characteristic subsequence. Extract the attenuation characteristics of the acceleration peak and the variation characteristics of the time interval between the peaks in the characteristic subsequence, and obtain the acceleration attenuation difference and time interval difference respectively; The motion period chaos coefficient of the exploration equipment is determined based on the acceleration attenuation difference, time interval difference, the mean motion stability index of all characteristic subsequences, and the mean morphological similarity between characteristic subsequences. Construct a joint trajectory matrix of the acceleration sequence and the angular velocity sequence, perform orthogonal decomposition on the joint trajectory matrix to obtain decomposition components, and calculate the core component ratio and component dispersion index of the trajectory decomposition based on the concentration distribution, dispersion degree and variance characteristics of the decomposition components. The positioning quality coefficient of the exploration equipment is generated based on the correlation strength, motion cycle chaos coefficient, core component ratio and component dispersion index of each row of the joint trajectory matrix; Calibrate and locate the marine resource exploration position based on the positioning quality coefficient.

2. The method for marine resource exploration and positioning based on inertial navigation according to claim 1, characterized in that: The identifying of a characteristic subsequence representing abnormal motion includes: Obtaining the spatiotemporal coupling matrix of the acceleration gradient field and the angular velocity gradient field in the characteristic subsequence; Detecting the coordinates of critical points in the coupling matrix where the gradient direction abruptly exceeds a preset threshold; Divide the equipment motion abnormality area according to the spatial distribution density and time continuity characteristics of the critical point coordinates; Extract the extreme value of the second-order derivative of acceleration and the mutation point of the angular velocity spectrum at the boundary of the abnormal area; Based on the spatiotemporal correlation between extreme points and mutation points, a set of equipment motion mutation feature sets is generated; Features in the motion mutation feature set that have a duration exceeding a preset period and a spatial coverage greater than a preset radius are marked as persistent abnormal areas.

3. The method for marine resource exploration and positioning based on inertial navigation according to claim 1, characterized in that: The method for obtaining the motion stability index is: For the target exploration area, the acceleration mean sequence and angular velocity mean sequence of its associated characteristic subsequence are obtained; Based on the fluctuation characteristics of the acceleration trend in the characteristic subsequence of the target exploration area and the difference characteristics of the local peak and valley values ​​in the angular velocity sequence, the axial oscillation coefficient and the dynamic offset coefficient are calculated respectively; The motion stability index of the target exploration area is obtained by multiplying the axial oscillation coefficient of the target exploration area by the dynamic offset coefficient; the axial oscillation coefficient is jointly determined by the variance of the acceleration sequence and the covariance of the angular velocity sequence; the dynamic offset coefficient is calculated by the ratio of the difference between the first and last accelerations of the characteristic subsequence and the angular velocity integration result.

4. The method for marine resource exploration and positioning based on inertial navigation according to claim 3, characterized in that: The method for obtaining the dynamic offset coefficient is: Extract the acceleration sequence mean of all exploration equipment in the target exploration area as the benchmark motion parameter; The exploration equipment is sorted in the order of equipment numbers to generate a first equipment sequence; and the equipment is sorted in descending order according to the motion stability index of each equipment to generate a second equipment sequence; If the device numbers at the same serial number position in the first device sequence and the second device sequence are the same, the device is marked as a matching device, and the ratio of the number of matching devices to the total number of devices is calculated as the first matching degree; The unmatched device numbers in the first device sequence are replaced by the Euclidean distance between their acceleration sequences and the reference motion parameters to generate a first distance sequence; the unmatched device numbers in the second device sequence are replaced by the Manhattan distance between their angular velocity sequences and the reference motion parameters to generate a second distance sequence; Calculating the cosine similarity between the first distance sequence and the second distance sequence as a second matching degree; The weighted product of the first matching degree and the second matching degree is used as the equipment coordination weight; the motion stability index of the target exploration area is multiplied by the equipment coordination weight to obtain the positioning coordination coefficient of the target exploration area.

5. The method for marine resource exploration and positioning based on inertial navigation according to claim 1, characterized in that: The method for obtaining the trajectory feature segment is: Convert the multi-device motion data aligned in time dimension into a three-dimensional tensor structure; Perform low-rank sparse decomposition on the three-dimensional tensor to obtain a low-rank tensor representing normal motion patterns and a sparse tensor containing abnormal motion features; Based on the abnormal entries marked by sparse tensors, the acceleration offset and angular velocity mutation value of each device in the corresponding time window are extracted; The abnormal entries are divided into equipment failure type abnormalities and environmental interference type abnormalities through clustering algorithm. Equipment failure type abnormalities are associated with the sudden change characteristics of acceleration sequence, while environmental interference type abnormalities are associated with the continuous offset characteristics of angular velocity sequence. The sparse components corresponding to the equipment failure type anomaly are used as the first anomaly subset, and the sparse components corresponding to the environmental interference type anomaly are used as the second anomaly subset; the first anomaly subset is used to trigger the equipment calibration instruction, and the second anomaly subset is used to trigger the path replanning instruction.

6. The method for marine resource exploration and positioning based on inertial navigation according to claim 5, characterized in that: The method for generating the exploration scheduling configuration table is: Call the equipment motion state data stored in the leader node and integrate the acceleration sequence mean and angular velocity covariance matrix of all exploration equipment in the current exploration cycle; Receive the global positioning parameter set broadcast by the leader node, which includes the average motion stability index of each device in the current exploration cycle, the target positioning coordinates of the next cycle generated by the leader node based on the device motion status data, and the verification identifier of the candidate leader node; Calculate the target location coordinates to be verified based on the global positioning parameter set, and verify whether the deviation value between the target location coordinates and the next cycle target location coordinates is within the threshold range, and verify the authority identification of the candidate leader node; When both the deviation verification and the authority verification are passed, a positioning confirmation signal is sent to the leader node; After receiving the scheduling instruction issued by the leader node, the candidate leader node is set as the control node of the next exploration cycle, and the target positioning coordinates and equipment motion status data are written into the exploration scheduling configuration table.

7. The method for marine resource exploration and positioning based on inertial navigation according to claim 1, characterized in that: The method for obtaining the positioning quality coefficient is: Calculate the absolute value of the mutual information entropy between each row of data in the joint trajectory matrix and other rows of data, and take the mean of the mutual information entropy of each row as the matrix autocorrelation; The positioning quality coefficient is calculated as follows: multiply the core component ratio by the component dispersion index, add the matrix autocorrelation, and then divide by the sum of the motion cycle chaos coefficient and the preset adjustment factor; the core component ratio is determined by the ratio of the energy of the first K components to the total energy after orthogonal decomposition, and the component dispersion index is calculated by the absolute value of the difference between the main component variance and the secondary component variance.

8. The method for marine resource exploration and positioning based on inertial navigation according to claim 7, characterized in that: The calibration and positioning of the marine resource exploration position includes: Calculate the error compensation threshold of the inertial navigation system based on the positioning quality coefficient. The threshold is inversely proportional to the sampling frequency of the device motion data and proportional to the square root of the positioning quality coefficient. The angular velocity integration result is corrected using the error compensation threshold, and the calibrated three-dimensional space coordinates are generated by combining the acceleration double integration result. The real-time positioning trajectory of the exploration equipment is updated based on the calibration coordinates, and the target location coordinates for marine resource exploration are output.

9. A marine resource exploration and positioning system based on inertial navigation, used to implement the marine resource exploration and positioning method based on inertial navigation according to any one of claims 1 to 8, characterized in that: include: The data acquisition module is used to collect the multi-axis acceleration sequence and angular velocity sequence generated by the exploration equipment when it moves in the marine environment; The interval division and clustering module is used to divide the local motion intervals according to the distribution characteristics of specific fluctuation patterns in the acceleration sequence, cluster the motion sequence based on the spatiotemporal correlation of the local motion intervals, and identify characteristic subsequences that represent abnormal motion; The motion stability index calculation module is used to calculate the motion stability index of each characteristic subsequence by combining the fluctuation degree of acceleration trend and the consistency level of angular velocity change within the characteristic subsequence; The difference extraction module is used to extract the attenuation characteristics of the acceleration peak and the variation characteristics of the peak time interval in the feature subsequence, and obtain the acceleration attenuation difference and time interval difference respectively; A motion period chaos coefficient determination module is used to determine the motion period chaos coefficient of the exploration equipment based on the acceleration attenuation difference, the time interval difference, the mean motion stability index of all characteristic subsequences, and the mean morphological similarity between characteristic subsequences; The trajectory decomposition analysis module is used to construct the joint trajectory matrix of the acceleration sequence and the angular velocity sequence, perform orthogonal decomposition on the joint trajectory matrix to obtain decomposition components, and calculate the core component ratio and component dispersion index of the trajectory decomposition based on the concentration distribution, dispersion degree and variance characteristics of the decomposition components; Positioning calibration module, used to generate the positioning quality coefficient of the exploration equipment based on the correlation strength of each row of the joint trajectory matrix, the motion cycle chaos coefficient, the core component ratio and the component dispersion index; Calibrate and locate the marine resource exploration position based on the positioning quality coefficient.

10. The marine resource exploration and positioning system based on inertial navigation according to claim 9, characterized in that: It also includes a data preprocessing module, which is used to filter and denoise the multi-axis acceleration sequence and angular velocity sequence collected by the data acquisition module, eliminate outliers and perform data standardization, and transmit the processed data to the interval division and clustering module.

Images