Vehicle-mounted gravity dynamic error compensation method and device based on intelligent algorithm

Abstract

The invention discloses a vehicle-mounted gravity dynamic error compensation method and device based on an intelligent algorithm, and relates to the technical field of gravity dynamic measurement. The method comprises the steps that preliminary screening of minimum redundancy and maximum correlation criteria and fine selection used by combining a first sparrow search algorithm and cross validation are used for feature extraction and reconstruction of input data; a new loss function is designed by taking the domain knowledge of gravity anomaly measurement as a constraint condition so as to construct PI-LSTM for dynamic error compensation; and finally, carrying out optimization search on the model hyper-parameters of the PI-LSTM through a second sparrow search algorithm. According to the method, error compensation of vehicle-mounted gravity dynamic measurement is carried out based on the swarm intelligence algorithm and PI-LSTM, the fitting performance and generalization ability of an error compensation model are effectively improved, and then the vehicle-mounted gravity dynamic measurement precision and environmental adaptability are improved.

Description

Technical Field

[0001] This application relates to the technical field of dynamic gravity measurement, and particularly to a vehicle-mounted dynamic gravity error compensation method and device based on an intelligent algorithm. Background Art

[0002] Dynamic gravity measurement can quickly, efficiently, and accurately obtain gravity information in a region relying on a vehicle carrier, can penetrate underground structures, and has extremely important application values in aspects such as economic construction, national defense construction, basic scientific research, and social development. Vehicle-mounted dynamic gravity measurement provides a better choice for the construction of a high-precision, high-efficiency, refined, and rich earth gravity field due to its advantages such as strong signal-to-noise ratio, high spatial resolution, flexibility, and convenience in establishing reference information. However, the dynamic errors introduced by environmental factors such as road surface bumps, survey line conditions, and vehicle carrier maneuvers in vehicle-mounted gravity measurement are extremely significant, posing a huge challenge to the measurement accuracy. To improve the accuracy of dynamic gravity measurement, many scholars have made great efforts. The team of Li Xiaopeng deduced an exact compensation model for the tilt error of an inertial stable platform, reducing the influence of short-wave random errors. Huang Motao proposed a general model applicable to compensating the remaining errors of various dynamic effects to effectively eliminate the influence of high-dynamic measurement environment effects in view of the problem of the widespread existence of dynamic environment effects in current operations. Cai Shaokun proposed a general model to compensate the dynamic errors of strapdown gravity measurement through the mechanism analysis of strapdown dynamic errors, effectively suppressing the dynamic errors of gravity measurement. Pan Guowei proposed a compensation method for the non-modeled errors of strapdown gravity measurement, estimating the non-modeled errors through real navigation data and trajectory generator data, and significantly improving the internal consistency accuracy.

[0003] Most existing studies are based on the understanding of the principle of dynamic gravity measurement to build an error compensation model to fit the dynamic measurement error. However, due to the incomplete understanding of the error influencing mechanism and the complexity and variability of the measurement environment, the accuracy, reliability and versatility of the constructed model are often difficult to guarantee and cannot be universally applied to all measurement processes, especially in high dynamic environments. The error suppression effect is very limited. The Long Short Term Memory (LSTM) model is a recurrent neural network with a special structure, which has unique advantages in nonlinear time series prediction. The neural network that integrates physical information can utilize the constraints of the physical model to improve the accuracy and generalization ability of the model prediction, and has the advantages of data-driven and physical mechanism fusion. Therefore, the physical information long short-term memory model is undoubtedly the best response strategy for solving tasks that contain both physical laws and time-dependent data. It has shown excellent model performance in many application fields such as fault diagnosis, medical image analysis, stock prediction, and natural disaster warning. The compensation of dynamic error in gravity measurement belongs to the regression problem in time series signal processing. Combining LSTM with physical constraints of gravity information is used to suppress the measurement error introduced by the complex dynamic environment interference whose mathematical model is not yet clear, which plays an important role in improving the accuracy and environmental adaptability of vehicle-mounted gravity dynamic measurement. Therefore, how to effectively suppress the random complex and multi-frequency coupled noise interference in the process of vehicle-mounted dynamic measurement, construct a physical information constraint long short-term memory model (Physics-informed Long ShortTerm Memory, PI-LSTM) for dynamic error compensation of gravity measurement, and design a hyperparameter optimization selection strategy for the error compensation model are technical problems that need to be solved in the field of gravity measurement error compensation based on machine learning models. Summary of the invention

[0004] The purpose of this application is to provide a vehicle-mounted gravity dynamic error compensation method and device based on an intelligent algorithm, which can solve the problem that the accuracy, reliability and versatility of the existing error compensation model are difficult to ensure, can effectively suppress the measurement errors introduced by complex dynamic environments, and further improve the vehicle-mounted gravity dynamic measurement accuracy and environmental adaptability.

[0005] To achieve the above objectives, this application provides the following solutions.

[0006] In a first aspect, the present application provides a vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm, including: obtaining a dynamic error compensation data set; the dynamic error compensation data set includes vehicle-mounted state parameter groups and gravity reference values under different operations; using the variational mode decomposition algorithm, adaptively decomposing each dimension of the input vector in the vehicle-mounted state parameter group into multiple eigenvectors respectively to obtain a total eigenvector set; the eigenvectors are intrinsic mode functions; the set of all input vectors corresponding to the same input vector is a feature in the total eigenvector set; based on the minimum redundancy maximum correlation criterion, performing preliminary feature screening on the total eigenvector set to obtain a preliminary screened feature subset; using the first sparrow search algorithm and the cross-validation algorithm to perform fine feature screening on the preliminary screened feature subset to obtain an optimal feature subset; based on the optimal feature subset, performing data reconstruction on multiple vehicle-mounted state parameter groups in the dynamic error compensation data set respectively to obtain vehicle-mounted state eigenvectors corresponding to each vehicle-mounted state parameter group; constructing a physics-informed long short-term memory model; based on the physical laws of gravity information, constructing a total loss function of the physics-informed long short-term memory model; using the vehicle-mounted state eigenvector as the input quantity, the gravity reference value as the output quantity, and the total loss function as the fitness function value of the second sparrow search algorithm, determining the optimal hyperparameter group of the physics-informed long short-term memory model based on the search strategy to obtain a dynamic error compensation model; using the dynamic error compensation model to perform vehicle-mounted gravity dynamic error compensation.

[0007] In a second aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the above vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm.

[0008] According to the specific embodiments provided by the present application, the following technical effects are disclosed.

[0009] The present application provides a vehicle-mounted gravity dynamic error compensation method and device based on intelligent algorithms. The data feature selection scheme that initially screens based on the minimum redundancy maximum relevance criterion (mRMR) and finely selects using SSA-CV (sparrow search algorithm - cross-validation) combines computational efficiency and reliability, effectively reducing the interference of non-significantly relevant features on model training and significantly improving the accuracy of the model. By embedding the physical constraints of gravity information into the data-driven loss function to form PI-LSTM, the requirements for the scale and quality of training data by the model can be greatly reduced, and the accuracy of gravity dynamic measurement error compensation and the model generalization ability in high-noise scenarios are significantly improved. Based on the swarm intelligence optimization algorithm, the hyperparameters of the PI-LSTM model are optimized and selected, effectively avoiding the problem of model performance degradation caused by manual parameter selection and further improving the performance of the model. It has a certain degree of feasibility, stability, and robustness, and has important value both in the theoretical level of machine learning model construction and in the engineering level of dynamic gravity measurement, and can be widely applied to the measurement and error compensation of marine, underwater, and aviation dynamic gravity. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0011] Figure 1 It is a flowchart of a vehicle-mounted gravity dynamic error compensation method based on intelligent algorithms in an embodiment of the present application.

[0012] Figure 2 It is a schematic diagram of a vehicle-mounted gravity dynamic error compensation method based on intelligent algorithms in an embodiment of the present application.

[0013] Figure 3 It is a result graph of the ablation experiment on the test set in an embodiment of the present application.

[0014] Figure 4 It is a result graph before dynamic error compensation in an embodiment of the present application.

[0015] Figure 5 It is a result graph after dynamic error compensation in an embodiment of the present application.

[0016] Figure 6 It is a result graph before and after dynamic error compensation in an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0018] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0019] In an exemplary embodiment, as Figure 1 shown, a vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm is provided, including steps 101-step 109.

[0020] Step 101: Obtain a dynamic error compensation data set. The dynamic error compensation data set includes vehicle-mounted state parameter groups and gravity reference values under different operations. The operations include: acceleration and deceleration operations, start and stop operations, and turning operations. The vehicle-mounted state parameter group includes: carrier attitude, carrier speed, carrier acceleration, carrier position, and measured gravity data. The vehicle-mounted state parameter group is a 13-dimensional input vector. The carrier attitude, carrier speed, carrier acceleration, and carrier position are all 3-dimensional input vectors. The measured gravity data is a 1-dimensional input vector.

[0021] Step 102: Use the variational mode decomposition algorithm to adaptively decompose each dimension of the input vector in the vehicle-mounted state parameter group into multiple feature vectors respectively, and obtain a total feature set. The feature vector is an intrinsic mode function. The set of all input vectors corresponding to the same input vector is a feature in the total feature set.

[0022] Step 103: Based on the minimum redundancy maximum correlation criterion, perform preliminary feature screening on the total feature set to obtain a preliminary screened feature subset.

[0023] Step 104: Use the first sparrow search algorithm and the cross-validation algorithm to perform fine feature screening on the preliminary screened feature subset to obtain an optimal feature subset.

[0024] Step 105: Based on the optimal feature subset, perform data reconstruction on multiple vehicle-mounted state parameter groups in the dynamic error compensation data set respectively to obtain a vehicle-mounted state feature vector corresponding to each vehicle-mounted state parameter group.

[0025] Step 106: Construct a physics-informed long short-term memory model. The activation function of the physics-informed long short-term memory model is the Relu function. The physics-informed long short-term memory model is trained using the Adam optimizer and has a fully connected layer for model output.

[0026] Step 107: Based on the physical laws of gravity information, construct the total loss function of the physical information constrained long short-term memory model.

[0027] Step 108: Using the vehicle state feature vector as the input quantity, the gravity reference value as the output quantity, and the total loss function as the fitness function value of the second sparrow search algorithm, determine the optimal hyperparameter group of the physical information constrained long short-term memory model based on the search strategy to obtain the dynamic error compensation model.

[0028] Step 109: Use the dynamic error compensation model to perform vehicle-mounted gravity dynamic error compensation.

[0029] Step 103 specifically includes steps 103-1 to 103-8.

[0030] Step 103-1: Calculate the mutual information between each feature and the target variable respectively. The target variable is the gravity reference value. The mutual information is: .

[0031] In the formula, is the i-th feature and the target variable 's mutual information. is the feature vector. is the target variable in one vector. is the joint probability distribution. is the marginal distribution of the feature vector . is the vector 's marginal distribution.

[0032] Step 103-2: Determine the feature corresponding to the maximum mutual information as the initial feature subset.

[0033] Step 103-3: Determine the initial feature subset as the current feature subset.

[0034] Step 103-4: Determine all feature vectors in the total feature set except the current feature subset as candidate features.

[0035] Step 103-5: Based on the minimum redundancy maximum correlation criterion, determine the correlation-redundancy index difference criterion scores of each candidate feature and the current feature subset respectively.

[0036] Step 103-6: Add the candidate feature corresponding to the maximum correlation-redundancy index difference criterion score to the current feature subset to obtain the updated current feature subset. The maximum correlation-redundancy index difference criterion score is as follows.

[0037] .

[0038] In the formula, is the j-th candidate feature. is the total feature set. is the feature subset. is the candidate feature and the mutual information with the target variable . is the candidate feature and the mutual information with the feature vector .

[0039] Step 103-7: Determine the evaluation index of the updated current feature subset. The evaluation index is the difference between correlation and redundancy.

[0040] The formula for correlation is as follows.

[0041] .

[0042] In the formula, is the correlation between the feature subset and the target variable . is the mutual information between the input data and the target variable .

[0043] The formula for redundancy is as follows.

[0044] .

[0045] In the formula, is the redundancy of the feature subset . is the mutual information of the feature vector pair in the feature subset .

[0046] Step 103-8: Take the updated current feature subset as the current feature subset, and return to Step 103-4 until the number of feature vectors in the current feature subset reaches the preset number, and determine the current feature subset before update as the preliminary screening feature subset.

[0047] Step 104 specifically includes Step 104-1 to Step 104-5.

[0048] Step 104-1: Define the individual classification and quantity of the sparrow population. Take the preliminary screening feature subset as the population dimension, and generate the first initial population through binary encoding based on the basic principle of the first sparrow search algorithm. One individual in the first initial population corresponds to one feature subset of the preliminary screening feature subset.

[0049] Step 104-2: Use multiple feature subsets corresponding to the initial population to perform 5-fold cross-validation on the regression base model to obtain a fitting model. The regression base model is a Gaussian regression model with multiple inputs and a single output.

[0050] Step 104-3: Calculate the root mean square error of the validation set in the fitting model, and use the root mean square error as the fitness function value for optimization by the first sparrow search algorithm. The validation set is a subset of the dynamic error compensation data set.

[0051] Step 104-4: Perform iterative optimization based on the search strategy of the first sparrow search algorithm, and gradually find the population individual corresponding to the minimum fitness function value as the first optimal population individual.

[0052] Step 104-5: Determine the feature subset corresponding to the first optimal population individual as the best feature subset.

[0053] The formula for the total loss function is as follows.

[0054] 。

[0055] Where, 。

[0056] 。

[0057] 。

[0058] 。

[0059] 。

[0060] In the formula, is the total loss function. is the weight of the physical loss function. is the weight of the monotonicity constraint loss function. is the weight of the initial value condition constraint loss function. is the data loss function. is the physical loss function. is the monotonicity constraint loss function. is the initial value condition constraint loss function. is the data loss function value. is the model parameter during training. is the length of the test data set. is the square root calculation. is the test data set. is the model input The output value obtained after is the input data. is the input data The corresponding true value of the model. Is the physical loss function value. Is the model input The output value obtained after Is the input data at time t. Is an intermediate parameter. a 1. a 2 and a 3均 Are the coefficients of the Gaussian - Markov model. Is the model input The output value obtained after Is the input data at time t - 1. Is the model input The output value obtained after Is the input data at time t - 2. Is the model input The output value obtained after Is the input data at time t - 3. Is the monotonicity constraint loss function value. Is the monotonicity factor. Is the sign function. Is the height information of the carrier position in the input data. Is the initial value condition constraint loss function value. Is the model input The output value obtained after Is the input data at the initial moment. Is the input data at the initial moment The corresponding true value of the model.

[0061] Step 108 specifically includes steps 108 - 1 to 108 - 8.

[0062] Step 108 - 1: Define the population size and the preset number of iterations of the second sparrow search algorithm.

[0063] Step 108 - 2: Determine the hyperparameter group of the physical information - constrained long - short - term memory model.

[0064] Step 108 - 3: Based on the parameter boundaries of each variable to be optimized in the hyperparameter group and the population size, perform population initialization to obtain the second initial population. One individual in the second initial population corresponds to a hyperparameter group.

[0065] Step 108 - 4: Substitute each individual in the second initial population into the physical information - constrained long - short - term memory model respectively, and use the training set to train each physical information - constrained long - short - term memory model.

[0066] Step 108-5: Use the validation set to determine the total loss of each individual in the second initial population, and use the total loss as the fitness function value of the second sparrow search algorithm.

[0067] Step 108-6: Update the second initial population and return to Step 108-4 until the number of iterations reaches the preset number of iterations, and determine the population individual corresponding to the minimum fitness function value as the second optimal population individual.

[0068] Step 108-7: Determine the hyperparameter group corresponding to the second optimal population individual as the optimal hyperparameter group.

[0069] Step 108-8: Substitute the hyperparameter group into the physics-informed constrained long short-term memory model to obtain a dynamic error compensation model.

[0070] Step 109 specifically includes Step 109-1 to Step 109-3.

[0071] Step 109-1: Obtain the vehicle state parameter group.

[0072] Step 109-2: Based on the best feature subset, perform data reconstruction on the vehicle state parameter group to obtain the target vehicle state feature vector.

[0073] Step 109-3: Input the vehicle state feature vector into the dynamic error compensation model to obtain the dynamic error compensation result of the gravity measurement value.

[0074] As Figure 2 , in an exemplary embodiment, a vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm is provided, including the following steps.

[0075] Step 1: Determine the model architecture; construct a machine learning model for dynamic error compensation of gravity anomaly measurement based on PI-LSTM, use the Relu function as the activation function, and train it using the Adam optimizer. A fully connected layer is used for model output. Define the input and output of the model and initialize the model parameters, including the number of neurons Num, the number of training epochs Epochs, the training batch size BatchSize, the initial learning rate LR, the learning rate decay factor DecayFactor, and the weight coefficient of physical information constraint. The model input and output need to be determined according to the actual needs of dynamic error compensation of gravity measurement, and a training data set with comprehensive features is constructed. The purpose of building the training model in this paper is to obtain the internal law between the carrier's dynamic measurement conditions and the gravity anomaly measurement error. Therefore, the motion characteristics of the carrier are used as the model input, including the attitude, velocity, and acceleration in three directions in the northeast-down geodetic coordinate system. Since the gravity information is closely related to the position of the carrier, especially the carrier height, the three-dimensional positioning coordinates of the carrier are also used as the model input. In addition, the gravity anomaly measurement results are also used as training input data, and the true gravity anomaly value is used as the output data for model training. The output of the obtained model can be used as the measurement result after dynamic error compensation. In summary, the constructed PI-LSTM is a model with 13-dimensional input and 1-dimensional output. In addition, the initialization of the model parameters is set according to the practical experience of model training.

[0076] The real data collected through experiments is used for model training, which needs to include the carrier attitude, velocity, acceleration, position, and measured gravity data in three directions in the northeast-down geodetic coordinate, a total of 13-dimensional model input, and the model output is the gravity reference value obtained by static measurement and interpolation fitting along the measurement line in advance. The experimental data is divided into a training set, a validation set, and a test set according to the ratio of 4:1:1. During the training data collection process, the vehicle performs obvious acceleration, deceleration, start-stop, turning and other operations to obtain training data with comprehensive maneuvering characteristics. Construct a machine learning model for dynamic error compensation of gravity anomaly measurement based on PI-LSTM, use the Relu function as the activation function, and train it using the Adam optimizer. A fully connected layer is used for model output. The initialization of the model hyperparameters is set manually according to practical experience, including the number of neurons in LSTM, the number of training epochs Epochs, the batch size BatchSize, the initial learning rate LR, the learning rate decay factor DecayFactor, and the hyperparameters of physical information constraint a 1, a 2, a 3, , and .

[0077] Step 2: Input data modal decomposition; variational mode decomposition (VMD) is used to adaptively decompose the 13-dimensional input data. The input data of each dimension is decomposed into x i ( i ∈[1,13]) intrinsic mode functions (IMFs). Therefore, the dimension of the input data will be extended to sum( x i ).

[0078] Variational mode decomposition is an adaptive, completely non-recursive modal variational and signal processing method. This technology has the advantage of being able to determine the number of modal decompositions. Its adaptability is manifested in determining the number of modal decompositions of the given sequence according to the actual situation. During the subsequent search and solution process, it can adaptively match the best center frequency and finite bandwidth of each mode, and can achieve effective separation of the intrinsic mode components and frequency domain division of the signal, and then decompose to obtain subsequences containing multiple different frequency scales and relatively stable.

[0079] Step 3: Feature selection and data reconstruction; based on the minimum redundancy maximum correlation (mRMR) criterion, a preliminary screening of the feature subset is carried out, and then a fine screening of the feature subset is carried out through the first sparrow search algorithm (SSA1) and cross-validation (CV), and the screened data is reconstructed into the input data for model training.

[0080] Among them, the preliminary screening of the feature subset based on mRMR includes the following steps.

[0081] Step 3-1: Calculate the mutual information between all features and the target variable. For feature X i and target C , the calculation method of mutual information is as follows.

[0082] .

[0083] In the formula, x i is a feature vector in X i , c is a vector in the target, p ( x i , c ) is the joint probability distribution, p ( x i ) and p ( c ) are marginal distributions.

[0084] Step 3-2: Select the feature with the maximum mutual information as the initial feature subsetS 0。

[0085] Step 3-3: Calculate the score of the candidate feature correlation and redundancy index difference criterion according to the mRMR criterion, and its calculation method is as follows.

[0086] 。

[0087] Step 3-4: Perform incremental feature selection, and add the feature vector with the highest calculated score x to the feature subset S 。

[0088] Step 3-5: Calculate the current correlation D and redundancy R ; the correlation D is the average mutual information between the feature subset S and the target C ; the redundancy R is the average mutual information between the features within the feature subset S. The calculation formulas are as follows.

[0089] 。

[0090] 。

[0091] Step 3-6: Gradually add the optimal individuals in the candidate features to the feature subset through the greedy algorithm to maximize the evaluation index shown in the following formula.

[0092] 。

[0093] Step 3-7: Terminate the iteration when the preset feature number is reached or adding features cannot significantly improve the model performance. At this time, the feature subset S 1 is the input feature data after preliminary screening.

[0094] In addition, the fine screening of the feature subset based on SSA1 and cross-validation includes the following steps.

[0095] Step 3-8: Define the individual classification and its quantity of the sparrow population. The population dimension is S 1 dimension, and generate the initial population through binary coding based on the basic principle of the sparrow search algorithm. 1 or 0 represents whether the IMF component is selected.

[0096] Step 3-9: Select the to-be-determined feature subset S from the initial population and then perform 5-fold cross-validation. The regression base model used is a multi-input single-output Gaussian regression process to simulate and fit the unknown functional relationship.

[0097] Step 3-10: Calculate the root mean square error (RMSE) of the model on the validation set and use it as the fitness function value for optimization by SSA1.

[0098] Step 3-11: Perform iterative optimization based on the search strategy of SSA1 to gradually find the population individual corresponding to the minimum fitness value.

[0099] Step 3-12: According to the correspondence between the population individual and the feature subset selection, carefully select the globally optimal feature subset.

[0100] Step 3-13: Reconstruct the data based on the best feature subset obtained after feature selection to obtain the dataset for model training.

[0101] The two-step feature selection strategy combining mRMR and cross-validation takes into account both computational efficiency and model reliability. On the one hand, it effectively solves the computational efficiency problem by initially screening and quickly narrowing the feature range through mRMR. On the other hand, it improves the reliability of feature selection through global optimization and fine selection based on SSA1 and CV.

[0102] Step 4: Design of the PI-LSTM loss function; according to the physical laws of gravity information, design physical constraint loss, monotonicity loss, and initial value condition loss and add them to the data-driven loss function. The weighted sum of each loss term is used as the loss function of PI-LSTM; use the training data to train the model under the current parameter conditions and calculate the total loss of the validation set on the obtained model.

[0103] Among them, the design process of the total loss function is as follows.

[0104] Step 4-1: Mean square error as the data loss of the model The formula is as follows.

[0105] .

[0106] In the formula, are the model parameters during training, S test is the test dataset, N is the length of the test dataset, x is the input data, is the model output value, y is the true value of the model.

[0107] Step 4-2: The gravity anomaly satisfies the third-order Gaussian-Markov process. The formula for physical information constraint under discrete time conditions is as follows.

[0108] .

[0109] In the formula,y ( t ) is a time series, a 1, a 2, and a 3 are the coefficients of the Gaussian-Markov model, with values in the range [0, 1] and satisfying a 1 + a 2 + a 3 = 1, is Gaussian white noise, based on which the physical loss function of the dynamic error compensation model for gravity anomaly measurement can be designed The formula is as follows.

[0110] .

[0111] In the formula, x t is t the input data at time

[0112] Step 4-3: The partial derivative of gravity data with respect to the carrier height satisfies the monotonicity constraint. Based on this, the loss function of the monotonicity constraint The formula is as follows.

[0113] .

[0114] m is the monotonicity factor. If the physical monotonicity is positive, then m = 1, otherwise m = -1. h is the height information of the carrier position in the input data, sign is the sign function.

[0115] Step 4-4: Dynamic gravity measurement generally starts from a reference point. At this time, the initial gravity value is accurately known. Therefore, the gravity anomaly measurement result also satisfies the following formula of the initial value condition constraint.

[0116] .

[0117] In the formula, x 1 is the input data at the initial time.

[0118] Step 4-5: The total loss function of PI-LSTM is the weighted sum of the data loss and the physical constraint loss, as shown in the following formula.

[0119] .

[0120] In the formula, , , and are the weights corresponding to the physical constraint loss, with values in the range [0, 1] and satisfying + + <1。

[0121] Step 4 - 6: Use the training data to train the model under the current parameter conditions, and calculate the total loss of the validation set on the obtained model.

[0122] Step 5: Hyperparameter iterative optimization; use the calculated total loss as the fitness function value of the second Sparrow Search Algorithm (SSA2), and perform iterative optimization based on the SSA2 search strategy. The population individual with the smallest fitness value is the current optimal hyperparameter combination of PI - LSTM. The model obtained after training is the final dynamic error compensation model.

[0123] Among them, the specific process of hyperparameter optimization based on SSA2 includes the following steps.

[0124] Step 5 - 1: Parameter initialization, define the population size and the number of iterations of SSA2.

[0125] Step 5 - 2: Since there is a certain quantitative relationship among the physical constraint weight coefficients, select a 1, a 2, , and as well as a total of ten variables including Num, Epochs, BatchSize, LR, and DecayFactor of LSTM as the hyperparameters to be optimized.

[0126] Step 5 - 3: Give the parameter boundaries of the variables to be optimized based on the practical experience of model training, and randomly generate a feasible solution within the feasible domain to initialize the population.

[0127] Step 5 - 4: Use the training data to train PI - LSTM under this hyperparameter combination.

[0128] Step 5 - 5: Calculate the total loss of the validation set on the obtained model, and use it as the fitness function value of SSA2.

[0129] Step 5 - 6: Perform iterative optimization based on the SSA2 search strategy. After each update and iteration of the population individuals, return to Step 5 - 4 for model training.

[0130] Step 5 - 7: Find the population individual corresponding to the smallest fitness value, and select the current optimal PI - LSTM hyperparameter combination accordingly.

[0131] Step 5 - 8: The model obtained after training is the final dynamic error compensation model for gravity measurement.

[0132] The proposed dynamic error compensation method is specifically described through specific embodiments.

[0133] The test vehicle is equipped with a dual-axis laser strapdown inertial navigation system, GNSS facilities, an odometer, and a barometric altimeter. High-precision navigation data is provided through integrated navigation solution to obtain the attitude, speed, and position information of the carrier. The carrier acceleration is obtained by taking the first difference of the calculated speed. In addition, a platform gravimeter is equipped to dynamically measure gravity anomaly data, jointly constituting a 13-dimensional model input. The experiment process is selected to be carried out on a closed road. Before collecting dynamic measurement experiment data, discrete true values of gravity anomalies are statically measured every 2 km along the survey line using a CG-5 high-precision relative gravimeter, and the true values of gravity anomalies on the survey line are obtained by interpolation fitting as the model output of the training dataset.

[0134] S1: PI-LSTM uses the Relu function as the activation function, is trained using the Adam optimizer, and has a fully connected layer for model output. Among them, LSTM uses a single-layer structure with 30 neurons, the number of training epochs Epochs = 300, the batch size BatchSize = 32, the initial learning rate LR = 0.001, the learning rate decay factor DecayFactor = 0.2, and the hyperparameters of the physical information constraint a 1 = 0.6, a 2 = 0.3, a 3 = 0.1, = 0.15, = 0.17, = 0.18.

[0135] S2: The variational mode decomposition is used to adaptively decompose the input data, and each-dimensional input vector is decomposed into several intrinsic mode functions.

[0136] S3: Feature selection and reconstruction of the input data, and the specific implementation steps are as follows.

[0137] S3-1: Based on the minimum redundancy maximum correlation (mRMR) criterion, a preliminary screening of the feature subset is carried out.

[0138] S3-2: The population dimension of SSA1 is the dimension of the feature subset obtained after preliminary screening, and the population is initialized by implementing binary encoding. Among them, the population size is set to 20, and the maximum number of iterations is 50.

[0139] S3-3: Fine screening of the feature subset is carried out through SSA1 and 5-fold cross-validation.

[0140] S3-4: The data obtained after screening is reconstructed as the input data for model training.

[0141] S4: Design the loss function of PI-LSTM, embed the physical constraint conditions of gravity information into the data-driven loss function, and then perform model training.

[0142] S5: Optimize the hyperparameters of PI-LSTM based on SSA2, and the specific implementation steps are as follows.

[0143] S5-1: Determine the hyperparameters to be optimized, including Num, Epochs, BatchSize, LR, DecayFactor, and the a 1, a 2, , and There are a total of ten variables, so the population dimension is 10.

[0144] S5-2: Define the population size and the number of iterations of SSA2. The model training is included in the iterative optimization process. Due to the limitation of computing resources, the population size and the number of iterations should not be too large. Therefore, the population size is set to 10, and the number of iterations is 10.

[0145] S5-3: According to the practical experience of model training, it can be set that Num ∈ [5, 100], Epochs ∈ [50, 500], BatchSize ∈ [16, 256], LR ∈ [1e-4, 1e-1], DecayFactor ∈ [0.2, 0.5]. Among them, the first three variables are integers, so integer encoding is required. The remaining 5 hyperparameter value ranges of the variables to be optimized are all [0, 1], and satisfy a 1 + a 2 < 1, + + < 1, and a feasible solution is randomly generated within the feasible region to initialize the population.

[0146] S5-4: Train PI-LSTM using the training data under the current hyperparameter combination, calculate the total loss of the test set on the obtained model, and use it as the fitness function value of SSA2.

[0147] S5-5: Perform iterative optimization until the population individual corresponding to the minimum fitness value is found. Based on this, select the current optimal PI-LSTM hyperparameter combination. The model obtained after training is the final dynamic error compensation model for gravity measurement.

[0148] Example 1.

[0149] The experimental data was divided into a training set, a validation set, and a test set according to a ratio of 4:1:1. Based on the dynamic error compensation model obtained after training on the training set and the validation set, this embodiment examined the improvement effect of input data feature screening and reconstruction, physical information constraint embedding loss function, and model hyperparameter optimization on the performance of LSTM through the test set, and ablation experiments were conducted. The experimental results are as Figure 3 shown, and the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) were statistically calculated in Table 1, where the abbreviations "FS" and "HO" represent feature screening and hyperparameter optimization respectively, "None" means no treatment, and the meaning of percentage is the percentage of the accuracy improvement of other methods compared with training LSTM (None-LSTM) with the original data.

[0150] Table 1 Statistical table of the test result accuracy of LSTM and PI-LSTM

[0151] The experimental results show that the error metrics of the LSTM model based on feature screening show significant improvement on the test set. Its RMSE, MAE, and MAPE are reduced by 41.1%, 32.8%, and 21.5% respectively compared with the model trained with the original feature data. And for PI-LSTM, the three accuracy metrics after feature screening also decreased by 29.8%, 31.4%, and 19.5% respectively compared with those before feature screening. In addition, the verification effect of PI-LSTM on the test set has a significant improvement compared with LSTM. The three error metrics decreased by 21.9%, 12.1%, and 15.4% respectively before feature screening, and by 6.98%, 10.3%, and 13.2% respectively after feature screening. The results show that the physical information constraint not only increases the interpretability of the model, but also has advantages in improving the fitting accuracy and enhancing the generalization ability of the model. After the selection of model hyperparameters, the error metrics of the test results of LSTM and PI-LSTM can be further decreased, demonstrating that hyperparameter optimization based on SSA2 has a significant effect on the performance improvement of machine learning models. In summary, the LSTM improved by the three technical modules of feature screening, incorporating physical information, and intelligent hyperparameter optimization has the best model performance, and the RMSE, MAE, and MAPE have achieved the largest improvements of 76.7%, 74.1%, and 69.5% respectively, showing significant advantages in time series tasks under complex engineering backgrounds.

[0152] Example 2.

[0153] This embodiment is used to evaluate the actual application effect of on-vehicle gravity anomaly dynamic measurement error compensation during the smooth operation of the vehicle. After the model training is completed, a closed road with good road conditions and no traffic signal interference is re-selected to conduct an on-vehicle gravity dynamic measurement experiment, and the vehicle motion characteristics are kept as stable as possible during the experiment.

[0154] For Experiment 1, the one-way survey line maintained a stable vehicle speed of about 26 km / h for 1 hour of measurement, and a total of two round trips, 4 times of repeated measurements were carried out. Use CG-5 to statically measure the true value of gravity anomaly every 2 km on the survey line in advance. After interpolation and fitting, it is used as the gravity anomaly benchmark on the survey line for the external conformity accuracy evaluation of the dynamic measurement results. The results of the repeated survey lines are used to evaluate the internal conformity accuracy of the results. The alignment operation is carried out at the beginning of each measurement in the repeated survey line experiment, and the navigation solution is restarted to ensure that the integrated navigation has a navigation result with sufficient high accuracy.

[0155] Construct a data set with the navigation data and the dynamic measurement results of gravity anomaly on each survey line in the format of model input, perform feature extraction and reconstruction on the input data according to the method of feature screening, and then input it into the error compensation model of the SSA2 optimized PI-LSTM hyperparameters that has been trained. The output result is the gravity anomaly measurement result after dynamic error compensation, and the measurement data on each survey line is processed in turn. The results before and after the dynamic error compensation of gravity measurement are as Figure 4 and Figure 5 shown, and calculate the accuracy indexes of the results in Table 2.

[0156] The results show that the overall trend of the measurement results is in good agreement with the reference value, but the impact of inevitable dynamic interference during the measurement process on the measurement results is also very significant, especially near 18 km of Survey Line 1 and near 5 km of Survey Line 2. The internal conformity accuracies of Survey Lines 1-4 are 2.62 mGal, 3.22 mGal, 2.93 mGal and 1.88 mGal respectively, and the external conformity accuracies are 3.12 mGal, 4.03 mGal, 3.17 mGal and 1.88 mGal respectively. Generally, an internal conformity of 2.71 mGal and an external conformity of 3.14 mGal are achieved. Such an accuracy level is obviously not sufficient for practical engineering applications. The fluctuating error of the gravity anomaly measurement results after dynamic error compensation is effectively suppressed. The internal conformity accuracies of Survey Lines 1-4 are respectively improved to 0.64 mGal, 0.35 mGal, 0.42 mGal and 0.31 mGal, and the external conformity accuracies are respectively improved to 0.92 mGal, 1.03 mGal, 1.02 mGal and 0.69 mGal. The overall internal and external conformity accuracies are also improved by 83.39% and 70.70% respectively, reaching 0.45 mGal and 0.92 mGal. The results show that the accuracy of on-vehicle gravity anomaly measurement is significantly improved after dynamic error compensation.

[0157] Table 2 Statistical Table of Measurement Result Accuracy before and after Dynamic Error Compensation in Example 2

[0158] Example 3

[0159] This example is used to test the application effect of the dynamic error compensation model for on-vehicle gravity measurement under high-dynamic operating conditions of vehicles. Another survey line is selected to re-conduct the dynamic measurement of on-vehicle gravity, and there are obvious acceleration and deceleration processes during the experiment. A total of 2 hours of data is collected under high-dynamic conditions, with a total mileage of about 40 km, and 2 repeated measurements are completed for 1 round trip. The experimental equipment and other experimental conditions are the same as those in Example 2, and the internal and external consistency accuracies are also calculated to evaluate the model performance. Figure 6 Table 3 shows the comparison results of the dynamic error compensation for gravity anomaly measurement in Example 3 before and after, and the accuracy indicators are calculated in Table 3.

[0160] Table 3 Statistical Table of Measurement Result Accuracy before and after Dynamic Error Compensation in Example 3

[0161] The results show that before the dynamic error compensation, the internal and external consistency accuracies of survey lines 1 and 2 are 3.36 mGal, 5.11 mGal, 3.36 mGal, and 3.02 mGal respectively, and overall, an internal consistency of 3.86 mGal and an external consistency of 4.19 mGal are achieved. It can be seen that under high-dynamic conditions, the measurement results are more significantly affected by dynamic interference, and the need for dynamic error compensation is more urgent and more practical. After the dynamic error compensation, the compliance of the overall trend of the measurement results on the 2 repeated survey lines with the reference value has been visibly improved. The internal and external consistency accuracies have been improved to 0.87 mGal, 1.34 mGal, 0.87 mGal, and 1.31 mGal respectively, and the overall internal and external consistency accuracies have also been improved by 74.11% and 68.50% respectively, reaching 0.87 mGal and 1.32 mGal. It can be seen that the dynamic error compensation effectively enhances the environmental adaptability of on-vehicle gravity measurement, but the measurement results under high-dynamic conditions are more strongly interfered by the movement of the carrier, and the improvement amplitude of the accuracy after the dynamic error compensation is less than that in Example 2 under steady operating conditions.

[0162] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm.

[0163] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0164] In an exemplary embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0165] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0166] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0167] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0168] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0169] In this article, specific examples are used to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm, characterized in that, Including: Obtain a dynamic error compensation data set; The dynamic error compensation data set includes a set of vehicle state parameters and a gravity reference value under different operations; Using the variational mode decomposition algorithm, adaptively decompose each dimension of the input vector in the set of vehicle state parameters into multiple feature vectors respectively to obtain a total feature set; the feature vectors are intrinsic mode functions; the set of all input vectors corresponding to the same input vector is a feature in the total feature set; Based on the minimum redundancy maximum correlation criterion, perform preliminary feature screening on the total feature set to obtain a preliminary screening feature subset; Using the first sparrow search algorithm and the cross-validation algorithm, perform fine feature screening on the preliminary screening feature subset to obtain an optimal feature subset; Based on the optimal feature subset, perform data reconstruction on multiple sets of vehicle state parameters in the dynamic error compensation data set respectively to obtain a vehicle state feature vector corresponding to each set of vehicle state parameters; Construct a physics-informed long short-term memory model; Based on the physical laws of gravity information, construct a total loss function of the physics-informed long short-term memory model; Using the vehicle state feature vector as the input quantity, the gravity reference value as the output quantity, and the total loss function as the fitness function value of the second sparrow search algorithm, determine the optimal hyperparameter set of the physics-informed long short-term memory model based on the search strategy to obtain a dynamic error compensation model; Use the dynamic error compensation model to perform vehicle gravity dynamic error compensation.

2. The vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to claim 1, characterized in that The operations include: acceleration and deceleration operations, start and stop operations, and turning operations; The set of vehicle state parameters includes: carrier attitude, carrier speed, carrier acceleration, carrier position, and measured gravity data; The set of vehicle state parameters is a 13-dimensional input vector; The carrier attitude, the carrier speed, the carrier acceleration, and the carrier position are all 3-dimensional input vectors; The measured gravity data is a 1-dimensional input vector.

3. The vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to claim 1, characterized in that, Based on the minimum redundancy maximum correlation criterion, perform preliminary feature screening on the total feature set to obtain a preliminary screening feature subset, specifically including: Calculate the mutual information between each feature and the target variable respectively; the target variable is the gravity reference value; Determine the feature corresponding to the maximum mutual information as the initial feature subset; Determine the initial feature subset as the current feature subset; Determine all feature vectors in the total feature set except the current feature subset as candidate features; Based on the minimum redundancy maximum correlation criterion, determine the correlation-redundancy index difference criterion score between each candidate feature and the current feature subset respectively; Add the candidate feature corresponding to the maximum correlation-redundancy index difference criterion score to the current feature subset to obtain an updated current feature subset; Determine the evaluation index of the updated current feature subset; Take the updated current feature subset as the current feature subset, and return to the step "Determine all feature vectors in the total feature set except the current feature subset as candidate features" until the number of feature vectors in the current feature subset reaches the preset number, and determine the current feature subset before updating as the preliminary screening feature subset.

4. The method for vehicle gravity dynamic error compensation based on intelligent algorithms according to claim 3, characterized in that The mutual information is as follows: ; Wherein, is the i-th feature and the mutual information with the target variable ; is the feature vector; is the target variable in one of the vectors; is the joint probability distribution; is the marginal distribution of the feature vector ; is the vector marginal distribution; The differential criterion score of the maximum correlation-redundancy index is as follows: ； Wherein, is the j-th candidate feature; is the total feature set; is the feature subset; is the candidate feature and the target variable mutual information; is the candidate feature and the feature vector mutual information; The evaluation index is the difference between correlation and redundancy; The correlation is: ； In the formula, is the feature subset and the target variable correlation; is the input data and the target variable mutual information; The redundancy is: ； wherein, is the redundancy of the feature subset ; is the mutual information of the feature vector pairs in the feature subset .

5. The vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to claim 1, characterized in that Using the first sparrow search algorithm and the cross-validation algorithm, perform fine feature screening on the preliminary screened feature subset to obtain the best feature subset, specifically including: Define the individual classification and quantity of the sparrow population, use the preliminary screened feature subset as the population dimension, and generate the first initial population through binary coding based on the basic principle of the first sparrow search algorithm; 1 individual in the first initial population corresponds to a feature subset of the preliminary screened feature subset; Use the multiple feature subsets corresponding to the initial population to perform 5-fold cross-validation on the regression base model to obtain the fitting model; the regression base model is a multi-input single-output Gaussian regression model; Calculate the root mean square error of the validation set in the fitting model, and use the root mean square error as the fitness function value for the first sparrow search algorithm to optimize; the validation set is a subset of the dynamic error compensation data set; Based on the search strategy of the first sparrow search algorithm, perform iterative optimization, and gradually find the population individual corresponding to the minimum fitness function value as the first optimal population individual; Determine the feature subset corresponding to the first optimal population individual as the best feature subset.

6. The vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to claim 1, characterized in that The activation function of the physical information constrained long short-term memory model is the Relu function; The physical information constrained long short-term memory model is trained using the Adam optimizer and has a fully connected layer for model output.

7. The vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to claim 1, wherein The total loss function is: ； Among them, ; ； ； ； ； In the formula, is the total loss function; is the weight of the physical loss function; is the weight of the monotonicity constraint loss function; is the weight of the initial value condition constraint loss function; is the data loss function; is the physical loss function; is the monotonicity constraint loss function; is the initial value condition constraint loss function; is the value of the data loss function; are the model parameters during training; is the length of the test data set; is the square root calculation; is the test data set; is the model input and the output value obtained after; is the input data; is the input data and the corresponding true value of the model; is the value of the physical loss function; is the model input and the output value obtained after; is the input data at time t; is an intermediate parameter; a 1, a 2 and a 3均 are the coefficients of the Gaussian-Markov model; is the model input and the output value obtained after; is the input data at time t-1; is the model input and the output value obtained after; is the input data at time t-2; is the model input and the output value obtained after; is the input data at time t-3; is the value of the monotonicity constraint loss function; is the monotonicity factor; is the sign function; is the height information of the carrier position in the input data; is the value of the initial value condition constraint loss function; is the model input and the output value obtained after; is the input data at the initial time; is the input data at the initial time and the corresponding true value of the model.

8. The vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to claim 1, characterized in that Using the vehicle-mounted state feature vector as the input quantity, the gravity reference value as the output quantity, and the total loss function as the fitness function value of the second sparrow search algorithm, determine the optimal hyperparameter group of the physical information constrained long short-term memory model based on the search strategy to obtain the dynamic error compensation model, specifically including: Define the population quantity and the preset number of iterations of the second sparrow search algorithm; Determine the hyperparameter group of the physical information constrained long short-term memory model; Based on the parameter boundaries of each variable to be optimized in the hyperparameter group and the population quantity, perform population initialization to obtain the second initial population; 1 individual in the second initial population corresponds to a hyperparameter group; Substitute each individual in the second initial population into the physical information constrained long short-term memory model respectively, and use the training set to train each physical information constrained long short-term memory model respectively; Use the validation set to determine the total loss of each individual in the second initial population respectively, and use the total loss as the fitness function value of the second sparrow search algorithm; Update the second initial population and return to the step "Substitute each individual in the second initial population into the physical information constrained long short-term memory model respectively, and use the training set to train each physical information constrained long short-term memory model respectively" until the number of iterations reaches the preset number of iterations, and determine the population individual corresponding to the minimum fitness function value as the second optimal population individual; Determine the hyperparameter group corresponding to the second optimal population individual as the optimal hyperparameter group; Substitute the hyperparameter group into the physical information constrained long short-term memory model to obtain the dynamic error compensation model.

9. The vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to claim 1, characterized in that Use the dynamic error compensation model to perform vehicle-mounted gravity dynamic error compensation, specifically including: Obtain the vehicle-mounted state parameter group; Based on the optimal feature subset, perform data reconstruction on the vehicle-mounted state parameter group to obtain the target vehicle-mounted state feature vector; Input the vehicle-mounted state feature vector into the dynamic error compensation model to obtain the error compensation result of the gravity measurement value.

10. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the vehicle-mounted gravity dynamic error compensation method based on an intelligent algorithm according to any one of claims 1-9.

Images