Method for calculating standard deviation of positioning position, electronic equipment and medium

Abstract

The invention provides a positioning position standard deviation calculation method, electronic equipment and a medium, and relates to the technical field of automobile control, and the positioning position standard deviation calculation method comprises the steps: obtaining a current multi-dimensional observation characteristic value corresponding to a current GNSS epoch; the current multi-dimensional observation characteristic value is input into a regression model obtained through machine learning training, and the regression model learns the mapping relation between the multi-dimensional GNSS observation characteristic value and the real positioning error in the training process; and determining the standard deviation of the current positioning position based on the output of the regression model. According to the technical scheme, the consistency of the predicted value of the standard deviation of the positioning position and the real positioning error can be improved.

Description

Technical Field

[0001] This application relates to the field of navigation and positioning technology, specifically to a method for calculating the standard deviation of positioning position, an electronic device, and a medium. Background Technology

[0002] Calculating the standard deviation of positioning (STD) is crucial for evaluating the reliability of GNSS positioning results and achieving high-performance integrated navigation. However, existing methods for calculating STD have shortcomings, leading to discrepancies between the predicted and actual positioning errors in complex and variable environments. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method, electronic device, and medium for calculating the standard deviation of positioning position, which can improve the consistency between the predicted value of the standard deviation of positioning position and the actual positioning error.

[0004] In a first aspect, embodiments of this application provide a method for calculating the standard deviation of a positioning location, comprising: obtaining the current multi-dimensional observation feature value corresponding to the current GNSS epoch; inputting the current multi-dimensional observation feature value into a regression model trained by machine learning, wherein the regression model learns the mapping relationship between the multi-dimensional GNSS observation feature value and the actual positioning error during the training process; and determining the standard deviation of the current positioning location based on the output of the regression model.

[0005] This application provides a method for calculating the standard deviation of a positioning location. The method inputs the current multi-dimensional observation feature values ​​into a regression model trained by machine learning. During the training process, the regression model learns the mapping relationship between the multi-dimensional GNSS observation feature values ​​and the actual positioning error. Based on the output of the regression model, the standard deviation of the current positioning location is determined. This directly establishes a dynamic mapping from rich, real-time multi-dimensional features to the actual positioning error at the model level. This overcomes the problems of poor consistency between the standard deviation of the current positioning location calculated based on traditional geometric or empirical models and the inability to match environmental changes in real time. It achieves high-precision real-time estimation of positioning errors in complex scenarios and can calculate the standard deviation of the positioning location in real time.

[0006] In one embodiment, the regression model is a regression decision tree model.

[0007] In one embodiment, the regression decision tree model runs in an embedded platform, wherein the binary tree structure of the regression decision tree model is converted into nested conditional statements that are executed in the embedded platform.

[0008] In one embodiment, the output of the regression model includes an estimate of the standard deviation of the current positioning position corresponding to the current GNSS epoch. Determining the standard deviation of the current positioning position based on the output of the regression model includes smoothing the estimate of the standard deviation of the current positioning position based on the historical standard deviation estimates of the positioning position from multiple GNSS epochs output by the regression model to obtain the standard deviation of the current positioning position.

[0009] In one embodiment, the current multidimensional observation features include at least two of the following features: geometric precision factor, receiver environment, receiver state, number of phase observations, number of pseudorange observations, number of satellites tracked by the receiver, number of satellites participating in the positioning solution, average signal-to-noise ratio, number of observations with a signal-to-noise ratio greater than a preset signal-to-noise ratio, pseudorange pre-approval residual, pseudorange post-approval residual, observation degrees of freedom, and multipath size.

[0010] In one embodiment, the regression model is configured in the GNSS of an embedded integrated navigation system, the integrated navigation system including a data fusion module, and the method further includes: if the standard deviation of the current positioning position is greater than or equal to a preset deviation value, determining that the positioning result of the GNSS is inaccurate, and using the data fusion module to reduce the weight of the positioning result in the data fusion module according to the standard deviation of the current positioning position.

[0011] In one embodiment, the depth of the regression decision tree model is less than or equal to 9, and the number of leaf node samples in the regression decision tree model is less than or equal to 7.

[0012] In one embodiment, the method further includes: obtaining multi-dimensional sample observation feature values ​​corresponding to GNSS epochs and the true positioning error; using the multi-dimensional sample observation feature values ​​as input to the regression model to be trained, and using the true positioning error as training label to train the regression model to be trained, so as to obtain the trained regression model.

[0013] Secondly, embodiments of this application provide a device for determining the accuracy of a positioning result, comprising: an acquisition module for acquiring current multi-dimensional observation feature values ​​corresponding to the current GNSS epoch; an input module for inputting the current multi-dimensional observation feature values ​​into a regression model trained by machine learning, wherein the regression model learns the mapping relationship between the multi-dimensional GNSS observation feature values ​​and the actual positioning error during the training process; and a determination module for determining the standard deviation of the current positioning position based on the output of the regression model.

[0014] Thirdly, embodiments of this application provide an electronic device, including: a processor; and a memory for storing processor-executable instructions, wherein the processor is used to execute the positioning standard deviation calculation method described in the first aspect above.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program for performing the method for calculating the standard deviation of the positioning position described in the first aspect.

[0016] Fifthly, embodiments of this application provide a computer program product, which includes a computer program. When the computer program is executed by the processor of a computer device, it enables the computer device to perform the method for calculating the standard deviation of the positioning position described in the first aspect.

[0017] In a sixth aspect, embodiments of this application provide a chip, including: a processor; and a memory for storing processor-executable instructions, wherein the processor is used to execute the positioning standard deviation calculation method described in the first aspect above.

[0018] This application provides a method, electronic device, and medium for calculating the standard deviation of a positioning location. The method inputs current multi-dimensional observation feature values ​​into a regression model trained through machine learning. During training, the regression model learns the mapping relationship between multi-dimensional GNSS observation feature values ​​and the actual positioning error. Based on the output of the regression model, the current positioning standard deviation is determined. This directly establishes a dynamic mapping from rich, real-time multi-dimensional features to the actual positioning error at the model level, overcoming the problem of poor consistency between the predicted value of the positioning standard deviation calculated based on traditional geometric or empirical models and the actual positioning error. This achieves high-precision estimation of the positioning standard deviation in complex scenarios. Furthermore, the technical solution provided in this application can adapt to changes in the environment and calculate the positioning standard deviation in real time. Attached Figure Description

[0019] Figure 1 The diagram shown is a schematic representation of the system architecture for determining the accuracy of positioning results provided in an exemplary embodiment of this application.

[0020] Figure 2 The diagram shown is a flowchart illustrating a method for calculating the standard deviation of a positioning position provided in an exemplary embodiment of this application.

[0021] Figure 3 The diagram shown is a flowchart illustrating a method for calculating the standard deviation of a positioning position according to another exemplary embodiment of this application.

[0022] Figure 4 The diagram shown is a structural schematic of a device for determining the accuracy of positioning results provided in an exemplary embodiment of this application.

[0023] Figure 5 The diagram shown is a block diagram of an electronic device for performing a method for calculating the standard deviation of a positioning position, provided in an exemplary embodiment of this application. Detailed Implementation

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

[0025] Exemplary System Figure 1 The diagram shown is a schematic representation of the system architecture of a combined navigation system 100 provided in an exemplary embodiment of this application. Figure 1 As shown, the integrated navigation system 100 may include a GNSS (Global Navigation Satellite System) 110, an INS (Inertial Navigation System) 120, and a data fusion module 130. The GNSS 110 is equipped with a trained regression decision tree model, which is converted into nested conditional statements. The INS 120 is equipped with an inertial measurement unit (IMU). The GNSS 110 may include a system that utilizes one or more navigation satellites for navigation, positioning, and timing. For example, the GNSS may include China's BDS (BeiDou Navigation Satellite System), the United States' GPS (Global Positioning System), and the European Union's Galileo system. INS (Inertial Navigation System) 120 can include a navigation system that operates autonomously without relying on external signals. The core principle of INS is to use the gyroscope and accelerometer of the inertial measurement unit (IMU) to measure the angular velocity and linear acceleration of the carrier in three-dimensional space in real time. By performing calculations such as time integration on these measurements, the attitude, velocity and position information of the carrier (such as vehicles, airplanes, ships, etc.) can be deduced.

[0026] In one example, GNSS (Global Navigation Satellite System) 110 can acquire satellite observation data at the current GNSS epoch, perform data processing on the satellite observation data at the current GNSS epoch, obtain the processing result including the GNSS positioning result, and extract features from the processing result and the satellite observation data at the current GNSS epoch to obtain the current multi-dimensional observation feature value. The GNSS positioning result is then sent to the data fusion module 130, and the current multi-dimensional observation feature value is input into the regression decision tree model. INS (Inertial Navigation System) 120 can use the inertial measurement unit (IMU) to acquire raw measurement values ​​(such as the angular velocity and linear acceleration of the vehicle), calculate the INS positioning result based on the raw measurement values, and send the INS positioning result to the data fusion module.

[0027] Furthermore, GNSS (Global Navigation Satellite System) 110 can run a regression decision tree model to obtain an estimate of the standard deviation of the current positioning position output by the regression decision tree model.

[0028] Furthermore, the GNSS (Global Navigation Satellite System) 110 can smooth the current positioning standard deviation estimate based on the historical positioning standard deviation estimates from multiple GNSS epochs output by the regression model, to obtain the current positioning standard deviation, and then send the current positioning standard deviation to the data fusion module 130. The historical positioning standard deviation estimates from multiple GNSS epochs constitute a time series of data composed of positioning standard deviation estimates output by the regression decision tree model over several epochs prior to the current epoch.

[0029] Furthermore, the data fusion module 130 can compare the standard deviation of the current positioning position with a preset deviation value. If the standard deviation of the current positioning position is greater than or equal to the first preset deviation value, it determines that the GNSS positioning result is inaccurate, reduces the weight of the GNSS positioning result in the data fusion module, and fuses the GNSS positioning result with the INS positioning result to obtain the final positioning result of the integrated navigation system 100. If the standard deviation of the current positioning position is greater than or equal to the second preset deviation value, it rejects the GNSS positioning result and uses the INS positioning result sent by the INS (Inertial Navigation System) 120 as the final positioning result of the integrated navigation system 100. Furthermore, the first preset deviation value and the second preset deviation value can be determined based on actual conditions, and the second preset deviation value is greater than the first preset deviation value.

[0030] It should be understood that the above application scenario examples are only shown to facilitate understanding of the spirit and principles of this application, and the embodiments of this application are not limited thereto. Rather, the embodiments of this application can be applied to any applicable scenario.

[0031] Exemplary methods Figure 2 The diagram shown is a flowchart illustrating a method for calculating the standard deviation of a positioning location according to an exemplary embodiment of this application. Figure 2 As shown, the method for calculating the standard deviation of the positioning position may include the following.

[0032] 210: Obtain the current multidimensional observation feature value corresponding to the current GNSS epoch.

[0033] In one embodiment, during GNSS positioning, the receiver acquires observation data from all visible satellites at a fixed processing cycle (e.g., 1Hz to 10Hz), each such processing cycle being called a GNSS epoch. Furthermore, the size of an epoch can be determined based on actual conditions; for example, an epoch may include 1Hz, 2Hz, 3Hz, etc. Further, the current GNSS epoch may include the GNSS epoch corresponding to when the receiver receives the latest observation data from visible satellites.

[0034] In one embodiment, multi-dimensional observation feature values ​​may include feature values ​​reflecting multiple dimensions of positioning accuracy. For example, multi-dimensional observation feature values ​​may include feature values ​​reflecting multiple dimensions such as satellite geometry, observed signal strength, observation consistency, and receiver status. Further, the current multi-dimensional observation feature values ​​are obtained by feature extraction from satellite observation data at the current GNSS epoch and the solution results of the satellite observation data at the current GNSS epoch. Data solution may include a series of transformations, calculations, and statistical processing on the satellite observation data at the current GNSS epoch to generate solution results such as GNSS positioning results, receiver clock errors, and observation residuals. Further, satellite observation data may include the raw observation values, measurement auxiliary data, and basic information obtained from satellite signal demodulation output by the GNSS receiver after signal tracking and measurement of satellites within its line of sight. For example, satellite observation data may include pseudorange, carrier phase, and Doppler shift observation values. Further still, multi-dimensional observation feature values ​​may include geometric accuracy factors, the number of pseudorange observation values, and the degrees of freedom of the observation values.

[0035] 220: Input the current multi-dimensional observation feature values ​​into the regression model trained by machine learning, whereby the regression model learns the mapping relationship between the multi-dimensional GNSS observation feature values ​​and the actual positioning error during the training process.

[0036] In one embodiment, the true positioning error (PE) may include a scalar measure of the error between the positioning result and the actual position of the vehicle. The true positioning error is typically calculated as the Euclidean norm (modulus) of the positioning error vector, with units of length (e.g., meters).

[0037] 230: Determine the standard deviation of the current location based on the output of the regression model.

[0038] In one embodiment, a regression model may include a mathematical model constructed using machine learning algorithms for predicting continuous numerical values, the core function of which is to learn the complex relationship between input variables (features) and output variables (target values). For example, regression models may include linear regression models, support vector regression models, neural network models, etc.

[0039] In one embodiment, the standard deviation of the positioning location (STD) can be an estimate of the standard deviation of the true positioning error (PE), i.e. (When PE follows a normal distribution centered at 0), it is a statistical indicator used to measure the accuracy of GNSS positioning results. When the true positioning error (PE) is unknown, the closer the true positioning error (PE) is to the standard deviation of the positioning location (STD), the more representative the STD output by the regression model is of the true positioning error (PE). The standard deviation of the positioning location is usually expressed in units of length (such as meters); a larger value indicates higher uncertainty and lower accuracy in the positioning result.

[0040] This application provides a method for calculating the standard deviation of a positioning location. The method inputs current multi-dimensional observation feature values ​​into a regression model trained through machine learning. During training, the regression model learns the mapping relationship between multi-dimensional GNSS observation feature values ​​and the actual positioning error. Based on the output of the regression model, the current positioning standard deviation is determined. This directly establishes a dynamic mapping from rich, real-time multi-dimensional features to the actual positioning error at the model level, overcoming the problem of poor consistency between the predicted value of the positioning standard deviation calculated based on traditional geometric or empirical models and the actual positioning error. This achieves high-precision estimation of the positioning standard deviation in complex scenarios. Furthermore, the technical solution provided in this application can adapt to changes in the environment and calculate the positioning standard deviation in real time.

[0041] According to one embodiment of this application, the regression model is a regression decision tree model.

[0042] In one embodiment, a regression decision tree model may include a tree-structured model trained through machine learning for solving regression prediction problems. The core objective of a regression decision tree model is to progressively divide the features of the input data and ultimately output a continuous predicted value (rather than a discrete category in a classification task). Its logic closely resembles the human decision-making process, making it highly interpretable and one of the commonly used fundamental models in regression problems. The essence of decision tree regression is to decompose the complex regression problem into a series of simple binary judgments (or a few multivariate judgments). By progressively dividing the data, the originally chaotic dataset is segmented into multiple subsets with similar features, and a unified predicted value is output for each subset (e.g., the average of the true labels of all samples within that subset).

[0043] In one embodiment, the structure of a regression decision tree model consists of three parts: a root node, internal nodes, and leaf nodes. The root node, located at the top of the tree, represents the entire dataset and is the starting point for the initial split, selecting the optimal feature for the first split. Internal nodes include those below the root node and above the leaf nodes; each internal node corresponds to a feature judgment condition used for further data segmentation. Leaf nodes, located at the bottom of the tree, represent subsets of similar samples obtained after multiple splits, and are used to output the predicted values ​​(regression results) for these subsets.

[0044] In one embodiment, mean squared error (MSE) can be used to partition multi-dimensional GNSS observation features and a partitioning threshold. MSE is the average of the squared differences between the true values ​​of samples and the mean of the subset (the provisional predicted value of that subset) within the two partitioned subsets. A smaller MSE indicates higher homogeneity among the partitioned subsets. Furthermore, the regression decision tree model traverses all features, tries different partitioning thresholds for each feature, calculates the MSE for each combination, and finally selects the feature and partitioning threshold that minimizes the error for partitioning. For example, the number of samples in the current node can be N, and the number of samples in the left child node after partitioning can be N. L The average value of the labels is The number of samples in the right node is N. R The label mean is ,but: Here, MSEL is the MSE of the left child node and MSER is the MSE of the right child node. All possible splitting thresholds are traversed to minimize the total MSE (weighted sum, with the weight being the proportion of child node samples) after the split, and then this feature and threshold are taken as the splitting combination.

[0045] In this embodiment, the regression model is a regression decision tree model. The regression decision tree model divides the input features layer by layer using a series of judgment rules based on feature thresholds, and finally outputs continuous standard deviation prediction values ​​at the leaf nodes. The regression decision tree model not only enhances the interpretability of the mapping process from complex features to error estimation, but its deterministic forward inference path also ensures the efficiency and stability of the calculation process, thus providing a stable and efficient model foundation for calculating the standard deviation of the positioning location.

[0046] According to one embodiment of this application, a regression decision tree model runs in an embedded platform, wherein the binary tree structure of the regression decision tree model is converted into nested conditional statements and executed in the embedded platform.

[0047] In one embodiment, the embedded platform may include a dedicated computing device or microcontroller system, designed to be integrated into a larger system, device, or apparatus to perform specific control, monitoring, or data processing tasks. Further, the embedded platform may include a hardware computing core running GNSS receiver firmware and integrated navigation algorithms; for example, the embedded platform may include a microcontroller unit (MCU) in an in-vehicle navigation system, a core processing chip inside a high-precision positioning module (such as an RTK receiver), or other suitable platforms.

[0048] In one embodiment, a conditional statement may include a basic program control statement, typically in the form of if(condition){...}else{...}, used to determine which subsequent code to execute based on the Boolean value (true or false) of a logical expression. Furthermore, nested conditional statements may include a hierarchical program structure where one or more conditional statements are placed within the execution branch of another conditional statement.

[0049] In one embodiment, the binary tree structure of a regression decision tree model can be converted into nested conditional statements using the following method: The node splitting logic of the decision tree is transformed into nested conditional statements, with the root node corresponding to the outermost if-else statement, inner nodes corresponding to middle-level if-else statements, and leaf nodes corresponding to the final predicted value. The regression decision tree is traversed, mapping each non-leaf node to an if statement, with its true / false branches corresponding to the left / right child nodes respectively. Code is recursively generated starting from the root node. If a branch points to a non-leaf node, the next level of if-else statements is recursively generated within its code block; if it points to a leaf node, the predicted value is generated. Finally, all statements are combined into a complete, sequentially executed code. Furthermore, the splitting logic of the regression decision tree model (such as feature indices, thresholds, leaf node predicted values, etc.) can be presented using visualization tools. For example, visualization tools can include Python's Graphviz library, the `plot_tree` function from the Scikit-learn library, or the Matplotlib library.

[0050] In one embodiment, the nested conditional statements converted from the binary tree structure of the regression decision tree model may include the following parameters: node_count (meaning the total number of nodes in the decision tree, i.e., the total number of internal nodes and leaf nodes), children_left[i] (meaning the index of the left child node of the i-th node, -1 indicates that the node is a leaf node), children_right[i] (meaning the index of the right child node of the i-th node, -1 indicates that the node is a leaf node), feature[i] (meaning the split feature index of the i-th internal node, such as 0=pseudorange error, 1=satellite elevation angle, 2=signal-to-noise ratio), threshold[i] (meaning the split threshold of the i-th internal node, such as when pseudorange error > 0.4, the right child node is selected), and value[i][0][0] (meaning the predicted value of the i-th leaf node, the output of the regression leaf node is the label mean of the sample of that node).

[0051] In this embodiment, the binary tree structure of the regression decision tree model is converted into nested conditional statements for execution in the embedded platform. The structural space complexity of the if-else conditional statement is only O(1), and the time complexity is only O(n) (where n is the depth of the binary tree). The interpretive traversal process of the tree structure is transformed into a completely simple and logically clear sequential execution code. Thus, with extremely low computational and memory overhead, the prediction accuracy can be completely consistent with the binary tree structure of the regression decision tree model, providing a practical solution for achieving real-time and accurate error estimation in the embedded platform.

[0052] According to one embodiment of this application, the output of the regression model includes an estimated standard deviation of the current positioning position corresponding to the current GNSS epoch. The determination of the standard deviation of the current positioning position based on the output of the regression model includes: smoothing the estimated standard deviation of the current positioning position based on the estimated standard deviation of the historical positioning positions of multiple GNSS epochs output by the regression model to obtain the standard deviation of the current positioning position.

[0053] In one embodiment, the current location standard deviation estimate may include a prediction directly output by a regression model for the input features of the current GNSS epoch, without further processing.

[0054] In one embodiment, the historical positioning standard deviation estimates of multiple GNSS epochs can be included as a time series data consisting of positioning standard deviation estimates output by a regression model from several epochs prior to the current epoch.

[0055] In one embodiment, smoothing may include a data post-processing method aimed at suppressing random fluctuations or noise that may exist in single-point estimation by combining statistical information from current and historical observations, thereby obtaining a more stable and reliable final estimate. Further, the current positioning standard deviation estimate can be smoothed using the following method: calculating the arithmetic mean of the current positioning standard deviation estimate and the historical positioning standard deviation estimates from multiple GNSS epochs, as the current positioning standard deviation.

[0056] In this embodiment, the standard deviation estimate of the current positioning position is smoothed based on the historical positioning position standard deviation estimates of multiple GNSS epochs output by the regression model to obtain the current positioning position standard deviation. By combining the historical positioning position standard deviation estimates in the time series to smooth the current positioning position standard deviation estimate, random fluctuations or instantaneous anomalies that may occur in single-point model prediction can be significantly suppressed, making the final output current positioning position standard deviation smoother and more reliable.

[0057] According to an embodiment of this application, the current multi-dimensional observation feature values ​​include at least two of the following feature values: geometric precision factor, receiver environment, receiver state, number of phase observations, number of pseudorange observations, number of satellites tracked by the receiver, number of satellites participating in the positioning solution, average signal-to-noise ratio, number of observations with a signal-to-noise ratio greater than a preset signal-to-noise ratio, pseudorange pre-approval residual, pseudorange post-approval residual, observation degrees of freedom, and multipath size.

[0058] In one embodiment, the geometric precision factor may include a scalar value used to quantitatively describe the impact of the spatial geometry of currently visible satellites on positioning accuracy. The geometric precision factor is a function of the geometry formed by the satellites and receiver currently being calculated; common types include the position precision factor (PDOP), the horizontal precision factor (HDOP), and the vertical precision factor (VDOP). A smaller geometric precision factor indicates a more favorable satellite geometry and, theoretically, a higher potential positioning accuracy.

[0059] In one embodiment, the receiver's environment may include a feature for classifying GNSS signal reception conditions. For example, the receiver's environment may include: open sky, city streets, under an overpass, dense forest, indoors, or inside a tunnel.

[0060] In one embodiment, the receiver's state may include a feature describing the motion characteristics of the carrier of the receiver. For example, the receiver's state may include: static, low-speed travel, high-speed travel, etc.

[0061] In one embodiment, the number of phase observations may include the number of satellite channels for which the receiver has successfully locked onto and acquired carrier phase observations.

[0062] In one embodiment, the number of pseudorange observations may include the number of pseudorange observations that can be used for positioning solutions.

[0063] In one embodiment, the number of satellites tracked by the receiver may include the total number of all satellites that the receiver's radio frequency front-end and signal processing channel are continuously tracking.

[0064] In one embodiment, the number of satellites participating in the positioning calculation may include the number of valid satellite observations actually used in the current epoch positioning calculation after data quality checks (such as signal-to-noise ratio screening and residual verification).

[0065] In one embodiment, the average signal-to-noise ratio may include the arithmetic or weighted average of the ratio of carrier power to noise power density (C / N0) of all observations (pseudorange and / or phase) involved in the positioning solution.

[0066] In one embodiment, the number of observations with a signal-to-noise ratio (SNR) greater than a preset SNR can include the number of observations with an SNR (C / N0) exceeding the preset SNR among all observations. Furthermore, the preset SNR can be determined based on actual conditions; for example, the preset SNR may include 40 dBHZ, 50 dBHZ, 60 dBHZ, etc.

[0067] In one embodiment, the pseudorange pre-reality residual may include the difference obtained by subtracting the geometric distance calculated based on the approximate receiver position and satellite ephemeris from the original pseudorange observation value before the positioning solution.

[0068] In one embodiment, the pseudorange post-verification residual may include the theoretical geometric distance recalculated using the optimal position estimate obtained from the current solution after completing the positioning solution, and the difference obtained by subtracting the original pseudorange observation value from the theoretical geometric distance.

[0069] In one embodiment, the degrees of freedom of the observations may include, in the localization solution, the difference between the total number of valid observations and the total number of basic state variables that need to be solved in the localization solution model.

[0070] In one embodiment, the multipath size may include an estimate of the multipath error level. Multipath error is caused by the superposition of satellite signals reflected from buildings, the ground, etc., with direct signals. The multipath size can be estimated using specific algorithms (such as signal-to-noise ratio-based observations, combinations of dual-frequency observations, or filtering based on geometric models).

[0071] In one embodiment, the current multidimensional observation feature value may also include at least one of the following: the number of high elevation angle observations, the number of cycle slip satellites, and the variance of the positioning location.

[0072] In this embodiment, the current multi-dimensional observation features include at least two of the following features: geometric precision factor, receiver environment, receiver state, number of phase observations, number of pseudorange observations, number of satellites tracked by the receiver, number of satellites participating in the positioning solution, average signal-to-noise ratio (SNR), number of observations with an SNR greater than a preset SNR, pseudorange pre-approval residual, pseudorange post-approval residual, observation degrees of freedom, and multipath size. This diverse set of features enables the regression decision tree model to simultaneously perceive multiple factors such as satellite spatial configuration, signal propagation environment, receiver dynamics, the quantity and quality of observations, and internal consistency of the solution, thereby improving the accuracy of the standard deviation of the output current positioning position.

[0073] According to one embodiment of this application, a regression model is configured in the GNSS of an embedded integrated navigation system. The integrated navigation system includes a data fusion module. The method further includes: if the standard deviation of the current positioning position is greater than or equal to a preset deviation value, determining that the positioning result of the GNSS is inaccurate, and using the data fusion module to reduce the weight of the positioning result in the data fusion module according to the standard deviation of the current positioning position.

[0074] In one embodiment, an embedded integrated navigation system may include a system integrated on an embedded hardware platform that fuses data from at least two different types of navigation sensors (such as GNSS, inertial measurement unit, odometry, etc.) using algorithms to obtain a better and more robust positioning result than a single sensor. For example, an embedded integrated navigation system may include: a vehicle-mounted GNSS / INS (inertial navigation system) integrated navigator, a GNSS and inertial sensor fusion unit in a UAV flight control system, and a GNSS and sensor fusion algorithm module in a smartphone for positioning services, etc.

[0075] In one embodiment, the data fusion module may include a core software or hardware logic unit in the integrated navigation system responsible for executing multi-source information fusion algorithms. The data fusion module typically employs optimal estimation theory (such as Kalman filtering and its variants) as a mathematical framework, calculating the fused positioning result by assigning dynamic weights to different sensors.

[0076] In one embodiment, the preset deviation value can be determined based on actual needs. For example, the preset deviation value may include 0.5 meters, 1 meter, 10 meters, etc.

[0077] In one embodiment, the GNSS positioning result may include the spatial location information of the vehicle (such as a car, airplane, ship, etc.) calculated by the receiver based on satellite observation data. For example, the GNSS positioning result can be represented by longitude, latitude, and elevation.

[0078] In one embodiment, the weight of the positioning result in the data fusion module may include a parameter in the data fusion module used to quantitatively characterize the reliability of the GNSS positioning result or the contribution of the GNSS positioning result to the final positioning result output by the embedded integrated navigation system.

[0079] In one embodiment, the weight of the GNSS positioning result in the data fusion module can be reduced based on the standard deviation of the current positioning location, using the following method: determining whether the standard deviation of the current positioning location is greater than or equal to a first preset deviation value and less than a second preset deviation value. If the standard deviation of the current positioning location is greater than or equal to the first preset deviation value and less than the second preset deviation value, the weight of the GNSS positioning result in the data fusion module is reduced; and determining whether the standard deviation of the current positioning location is greater than or equal to the second preset deviation value. If the standard deviation of the current positioning location is greater than or equal to the second preset deviation value, the GNSS positioning result is rejected, and calculation is performed solely using other sensors such as inertial navigation until the GNSS positioning standard deviation recovers to a reliable level. Further, the first preset deviation value may include a threshold for triggering a reduction in the weight of the GNSS positioning result, and the second preset deviation value may include a threshold for triggering the rejection of the GNSS positioning result. The first preset deviation value and the second preset deviation value can be determined based on actual conditions, and the first preset deviation value is less than the second preset deviation value.

[0080] In this embodiment, the regression model is configured in the GNSS of the embedded integrated navigation system. The integrated navigation system includes a data fusion module. If the standard deviation of the current positioning position is greater than or equal to a preset deviation value, it is determined that the GNSS positioning result is inaccurate. The data fusion module then reduces the weight of the positioning result in the data fusion module based on the standard deviation of the current positioning position. This allows the integrated navigation system to adaptively adjust the fusion weights of the outputs of multiple positioning systems in the integrated navigation system based on a quantitative judgment of the reliability of the standard deviation of the current positioning position. When the positioning accuracy of the GNSS system deteriorates, the system can immediately reduce its dependence on the GNSS system, preventing unreliable positioning results from contaminating the final output, and effectively improving the accuracy of the positioning results of the integrated navigation system in complex and variable signal environments.

[0081] According to one embodiment of this application, the depth of the regression decision tree model is less than or equal to 9, and the number of leaf node samples in the regression decision tree model is less than or equal to 7.

[0082] In one embodiment, the depth of a regression decision tree model may include the maximum number of decision levels traversed from the root node of the tree to the farthest leaf node.

[0083] In one embodiment, leaf nodes include terminating nodes in the regression decision tree model that no longer undergo further splitting. Furthermore, the number of leaf node samples in the regression decision tree model can include the minimum number of training samples belonging to each leaf node after the regression decision tree model has been trained.

[0084] In this embodiment, the depth of the regression decision tree model is less than or equal to 9, and the number of leaf node samples is less than or equal to 7. Limiting the model depth ensures that the maximum number of judgments required for a single prediction is controllable, allowing the inference time to meet the real-time requirements of the embedded system. Limiting the minimum number of leaf node samples improves the model's generalization ability, avoids overfitting the regression decision tree model to noise in the training data, and ensures the accuracy of the standard deviation of the current location output by the regression decision tree model under different environments.

[0085] According to one embodiment of this application, the method further includes: obtaining multi-dimensional sample observation feature values ​​corresponding to GNSS epochs and the actual positioning error; using the multi-dimensional sample observation feature values ​​as input to the regression model to be trained, using the actual positioning error as training label, and training the regression model to be trained to obtain a trained regression model.

[0086] In one embodiment, the training label may be included in the model training phase of supervised machine learning, providing a real and known reference value (e.g., the real positioning error provided in this application) for each training sample (e.g., the multi-dimensional sample observation feature value provided in this application) to guide the model to learn the mapping relationship between the multi-dimensional sample observation feature value and the real positioning error. The goal of model training is to make its output value (e.g., the standard deviation of the current positioning location STD provided in this application) as consistent as possible with the training label (e.g., the real positioning error provided in this application).

[0087] In one embodiment, the regression model to be trained can be trained based on the following method to obtain a trained regression model: construct a training sample set, where the input features are multi-dimensional sample observation feature values ​​and the sample labels are the true localization errors of the corresponding epochs. Then, initialize the model and set structural constraint parameters, such as maximum depth ≤ 9 and minimum number of samples in leaf nodes ≤ 7, to ensure that the model is lightweight. Subsequently, a recursive binary search strategy is used for training. Starting from the root node, the optimal features and segmentation thresholds are selected according to the minimum mean square error criterion to split the nodes until the stopping conditions such as depth or number of samples are met, and finally the leaf nodes are formed. Furthermore, during model training, the parameters of the regression model can be updated using the following method: For the standard deviation (STD) of the current location output at each epoch, calculate the absolute value of the difference between the current location standard deviation (STD) and the true location error (PE) of the corresponding epoch, and use this as the loss for the current epoch. Arrange the losses for all epochs in ascending order, and calculate the first loss corresponding to the first percentile, the second loss corresponding to the second percentile, and the third loss corresponding to the third percentile. If the value of the first loss is less than or equal to the first preset threshold, the value of the second loss is less than or equal to the second preset threshold, and the value of the third loss is less than or equal to the third preset threshold, the parameters of the regression model are confirmed to be reasonable; otherwise, the parameters of the regression model are confirmed to be unreasonable, and the parameters of the regression model are updated. Furthermore, the first, second, and third percentages can be determined based on actual conditions. For example, the first percentage could be 68%, the second percentage could be 99.5%, and the third percentage could be 99.9%, or, for another example, the first percentage could be 50%, the second percentage could be 95%, and the third percentage could be 99%. Furthermore, the first, second, and third preset thresholds can be determined based on actual conditions. For example, the first preset threshold could be 0.5 meters, the second preset threshold could be 1 meter, and the third preset threshold could be 1.5 meters. Alternatively, the first preset threshold could be 0.3 meters, the second preset threshold could be 0.8 meters, and the third preset threshold could be 1.2 meters. The positioning error is estimated based on the standard deviation (STD) of the current positioning location output by the model, and the estimated positioning error is compared with the actual positioning error. The parameters of the regression model are updated based on the difference between the positioning error and the actual positioning error to ensure that the standard deviation (STD) of the current positioning location is as consistent as possible with the actual positioning error.

[0088] In one embodiment, the regression decision tree model is generated using a PC-based regression decision tree model generation tool. This tool comprises a data layer, an algorithm layer, a functional layer, and an interaction layer. Specifically, the data layer handles dataset import, preprocessing, and format adaptation; the algorithm layer implements the core logic of the regression decision tree (feature selection, node splitting, pruning optimization, etc.), primarily determining the tree's depth and pruning strategies; the functional layer provides model training, evaluation, export, and visualization functions, mainly for testing regression performance; and the interaction layer receives user commands via a command-line interface (CLI) and outputs the model training results.

[0089] In this embodiment, multi-dimensional sample observation feature values ​​corresponding to GNSS epochs and the true positioning error are obtained. The multi-dimensional sample observation feature values ​​are used as input to the regression model to be trained, and the true positioning error is used as the training label to train the regression model to obtain a trained regression model. Using the true positioning error as the training label allows the standard deviation of the current positioning location (STD) output by the trained regression decision tree model to more accurately represent the true positioning error, thereby improving the accuracy of the STD output by the model.

[0090] By employing supervised learning, the regression model learns the inherent statistical regularities and complex mapping relationships between multi-dimensional sample observation features and sample location positions from a large amount of real-world data covering different scenarios. This training process based on real-world positioning errors ensures the objectivity and effectiveness of the mapping relationships learned by the regression model, and improves the consistency between the standard deviation of the current positioning position output by the regression model and the actual positioning error.

[0091] Figure 3 The diagram shown is a flowchart illustrating a method for calculating the standard deviation of a positioning position according to another exemplary embodiment of this application. Figure 3 The example is Figure 2 Examples of the embodiments are provided below; to avoid repetition, the similarities can be referred to the descriptions in the above embodiments, and will not be repeated here. For example... Figure 3 As shown, the method for calculating the standard deviation of the positioning position may include the following.

[0092] 310: Obtain satellite observation data for the current GNSS epoch, and obtain the estimated standard deviation of historical positioning positions for multiple GNSS epochs.

[0093] Specifically, the relevant information regarding the current GNSS epoch, satellite observation data, and the estimated standard deviation of historical positioning positions for multiple GNSS epochs can be found in the descriptions in the above embodiments. To avoid repetition, these details will not be repeated here.

[0094] 320: Based on the satellite observation data of the current GNSS epoch, the solution results including the GNSS positioning results are obtained, and the features of the solution results and the satellite observation data of the current GNSS epoch are extracted to obtain the current multi-dimensional observation feature values.

[0095] Specifically, the relevant content of the satellite observation data, solution results, current multi-dimensional observation feature values, and GNSS positioning results for the current GNSS epoch can be found in the relevant descriptions in the above embodiments. To avoid repetition, they will not be repeated here.

[0096] 330: Input the current multi-dimensional observation feature values ​​into the regression decision tree model, run the regression decision tree model, and obtain the current positioning standard deviation estimate output by the regression decision tree model.

[0097] Specifically, the relevant content of the current regression decision tree model and the estimated standard deviation of the current location can be found in the relevant descriptions in the above embodiments. To avoid repetition, they will not be repeated here.

[0098] 340: Based on the historical standard deviation estimates of the positioning positions from multiple GNSS epochs, the current positioning position standard deviation estimate is smoothed to obtain the current positioning position standard deviation.

[0099] Specifically, the estimated standard deviation of the current positioning position and the smoothing process for the estimated standard deviation of the current positioning position can be found in the relevant descriptions in the above embodiments. To avoid repetition, they will not be repeated here.

[0100] 350: Determine whether the standard deviation of the current positioning position is greater than or equal to the first preset deviation value. If the standard deviation of the current positioning position is less than the second preset deviation value, proceed to step 360. If the standard deviation of the current positioning position is greater than or equal to the preset deviation value, proceed to step 370.

[0101] Specifically, the relevant content of the first preset deviation value and the second preset deviation value can be found in the relevant descriptions in the above embodiments. To avoid repetition, they will not be repeated here.

[0102] 360: Maintain the weight of GNSS positioning results in the data fusion module.

[0103] Specifically, the relevant content regarding GNSS positioning results, the data fusion module, and the weight of GNSS positioning results in the data fusion module can be found in the relevant descriptions in the above embodiments. To avoid repetition, they will not be repeated here.

[0104] 370: Determine whether the standard deviation of the current positioning position is greater than or equal to the second preset deviation value. If the standard deviation of the current positioning position is greater than or equal to the second preset deviation value, proceed to step 380. If the standard deviation of the current positioning position is less than the second preset deviation value, proceed to step 390.

[0105] Specifically, the relevant details of the second preset deviation value can be found in the descriptions in the above embodiments, and will not be repeated here to avoid repetition.

[0106] 380: Reject this GNSS positioning result.

[0107] 390: Reduce the weight of GNSS positioning results in the data fusion module.

[0108] Exemplary device Figure 4 The diagram shown is a schematic representation of a positioning result accuracy determination device provided in an exemplary embodiment of this application. This positioning result accuracy determination device can be applied to electronic devices. Figure 4 As shown, the positioning result accuracy determination device 400 includes: an acquisition module 410, an input module 420, and a determination module 430.

[0109] The acquisition module 410 is used to acquire the current multi-dimensional observation feature values ​​corresponding to the current GNSS epoch. The input module 420 is used to input the current multi-dimensional observation feature values ​​into a regression model trained through machine learning, wherein the regression model learns the mapping relationship between the multi-dimensional GNSS observation feature values ​​and the actual positioning error during training. The determination module 430 is used to determine the standard deviation of the current positioning position based on the output of the regression model.

[0110] This application provides a device for determining the accuracy of positioning results. It inputs current multi-dimensional observation feature values ​​into a regression model trained through machine learning. During training, the regression model learns the mapping relationship between multi-dimensional GNSS observation feature values ​​and the actual positioning error. Based on the output of the regression model, it determines the current positioning standard deviation, thereby directly establishing a dynamic mapping from rich, real-time multi-dimensional features to the actual positioning error at the model level. This overcomes the problem of poor consistency between the predicted value of the positioning standard deviation calculated based on traditional geometric or empirical models and the actual positioning error, achieving high-precision estimation of the positioning standard deviation in complex scenarios. Furthermore, the technical solution provided in this application can adapt to changes in the environment and calculate the positioning standard deviation in real time.

[0111] According to one embodiment of this application, the regression model is a regression decision tree model.

[0112] According to one embodiment of this application, a regression decision tree model runs in an embedded platform, wherein the binary tree structure of the regression decision tree model is converted into nested conditional statements and executed in the embedded platform.

[0113] According to an embodiment of this application, the output of the regression model includes the estimated standard deviation of the current positioning position corresponding to the current GNSS epoch. The determination module 430 is used to smooth the estimated standard deviation of the current positioning position based on the estimated standard deviation of the historical positioning position of multiple GNSS epochs output by the regression model, so as to obtain the standard deviation of the current positioning position.

[0114] According to an embodiment of this application, the current multi-dimensional observation feature values ​​include at least two of the following feature values: geometric precision factor, receiver environment, receiver state, number of phase observations, number of pseudorange observations, number of satellites tracked by the receiver, number of satellites participating in the positioning solution, average signal-to-noise ratio, number of observations with a signal-to-noise ratio greater than a preset signal-to-noise ratio, pseudorange pre-approval residual, pseudorange post-approval residual, observation degrees of freedom, and multipath size.

[0115] According to one embodiment of this application, a regression model is configured in the GNSS of an embedded integrated navigation system. The integrated navigation system includes a data fusion module. The determination module 430 is further configured to determine that the positioning result of the GNSS is inaccurate if the standard deviation of the current positioning position is greater than or equal to a preset deviation value, and to use the data fusion module to reduce the weight of the positioning result in the data fusion module according to the standard deviation of the current positioning position.

[0116] According to one embodiment of this application, the depth of the regression decision tree model is less than or equal to 9, and the number of leaf node samples in the regression decision tree model is less than or equal to 7.

[0117] According to an embodiment of this application, the positioning result accuracy determination device 400 further includes a training module 440, which is used to acquire multi-dimensional sample observation feature values ​​corresponding to GNSS epochs and the actual positioning error; use the multi-dimensional sample observation feature values ​​as input to the regression model to be trained, use the actual positioning error as training label, and train the regression model to be trained to obtain a trained regression model.

[0118] It should be understood that the operation and function of the acquisition module 410, input module 420, determination module 430, and training module 440 in the above embodiments can be referred to the above. Figure 2 or Figure 3 The description of the method for calculating the standard deviation of the positioning position provided in the embodiments will not be repeated here to avoid repetition.

[0119] Figure 5The diagram shown is a block diagram of an electronic device 500 for performing a method for calculating the standard deviation of a positioning position, provided in an exemplary embodiment of this application. Specifically, the electronic device 500 may be a server, a mobile device, a control device for a mobile device, a server interacting with a mobile device, a controller, or other devices.

[0120] Reference Figure 5 The electronic device 500 includes a processing component 510, which further includes one or more processors, and memory resources represented by a memory 520 for storing instructions executable by the processing component 510, such as application programs. The application programs stored in the memory 520 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 510 is configured to execute instructions to perform the aforementioned method for calculating the standard deviation of the positioning position.

[0121] Electronic device 500 may also include a power supply component configured to perform power management of electronic device 500, a wired or wireless network interface configured to connect electronic device 500 to a network, and an input / output (I / O) interface. Electronic device 500 can be operated based on an operating system stored in memory 520, such as Windows Server. TM Mac OSX TM Unix TM Linux TM FreeBSD TM Or similar.

[0122] This application also provides an electronic device, including: an electronic device configured to execute the method for calculating the standard deviation of positioning position provided in any of the above embodiments. It should be understood that the electronic device and its operation and functions can be referred to the above description. Figure 2 or Figure 3 The description of the method for calculating the standard deviation of the positioning position provided in the embodiments will not be repeated here to avoid repetition. For example, the electronic device may be as described above. Figure 5 500 electronic devices in the system.

[0123] A non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by the processor of the aforementioned electronic device 500, enables the electronic device 500 to perform a method for calculating the standard deviation of a positioning position.

[0124] A computer program product includes a computer program that, when executed by a processor of a computer device, enables the computer device to perform the method for calculating the standard deviation of the positioning position provided in any of the above embodiments.

[0125] All of the above-mentioned optional technical solutions can be combined in any way to form optional embodiments of this application, and will not be described in detail here.

[0126] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0127] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0129] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0130] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0131] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program verification codes, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0132] It should be noted that in the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0133] 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 used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0134] The above are merely preferred embodiments of this application and are not intended to limit this application. Any modifications or equivalent substitutions made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method of calculating a positioning position standard deviation, characterized by, The method includes: Obtain the current multidimensional observation feature value corresponding to the current GNSS epoch; The current multi-dimensional observation feature values ​​are input into a regression model trained by machine learning, wherein the regression model learns the mapping relationship between the multi-dimensional GNSS observation feature values ​​and the actual positioning error during the training process; Based on the output of the regression model, the standard deviation of the current location is determined.

2. The method of claim 1, wherein, The regression model is a regression decision tree model.

3. The method according to claim 2, characterized in that, The regression decision tree model runs in an embedded platform, wherein the binary tree structure of the regression decision tree model is converted into nested conditional statements and executed in the embedded platform.

4. The method according to claim 1, characterized in that, The output of the regression model includes an estimated standard deviation of the current positioning location corresponding to the current GNSS epoch, wherein determining the standard deviation of the current positioning location based on the output of the regression model includes: The standard deviation estimate of the current positioning position is smoothed based on the historical positioning position standard deviation estimates of multiple GNSS epochs output by the regression model to obtain the standard deviation of the current positioning position.

5. The method according to claim 1, characterized in that, The current multidimensional observation features include at least two of the following features: geometric precision factor, receiver environment, receiver state, number of phase observations, number of pseudorange observations, number of satellites tracked by the receiver, number of satellites participating in the positioning solution, average signal-to-noise ratio, number of observations with a signal-to-noise ratio greater than a preset signal-to-noise ratio, pseudorange pre-approval residual, pseudorange post-approval residual, observation degrees of freedom, and multipath size.

6. The method according to claim 1, characterized in that, The regression model is configured in an embedded integrated navigation system (GNSS), the integrated navigation system includes a data fusion module, and the method further includes: If the standard deviation of the current positioning position is greater than or equal to a preset deviation value, the GNSS positioning result is determined to be inaccurate, and the weight of the positioning result in the data fusion module is reduced based on the standard deviation of the current positioning position.

7. The method according to claim 2, characterized in that, The depth of the regression decision tree model is less than or equal to 9, and the number of leaf node samples in the regression decision tree model is less than or equal to 7.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: Obtain multi-dimensional sample observation feature values ​​and actual positioning errors corresponding to GNSS epochs; The multi-dimensional sample observation feature values ​​are used as input to the regression model to be trained, and the true positioning error is used as the training label to train the regression model to obtain a trained regression model.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions. The processor is used to execute the method for determining the accuracy of the positioning result as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program for executing the method for determining the accuracy of the positioning result according to any one of claims 1 to 8.

11. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by the processor of a computer device, enables the computer device to perform the method for determining the accuracy of the positioning result as described in any one of claims 1 to 8.

Images