A positioning coordinate error correction method based on machine learning
By collecting and processing GNSS observation data at the user end, a machine learning model is built to predict positioning errors, which solves the problem of insufficient accuracy of traditional positioning methods in complex environments. It achieves high-precision positioning correction without additional hardware and is suitable for complex environments and remote areas.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUBEI LUOJIA LAB
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-09
AI Technical Summary
Traditional single-point positioning methods lack positioning accuracy in complex environments, especially in urban canyons, forested areas, or densely populated areas with severe signal obstruction and interference. Existing technologies struggle to achieve high-precision real-time error correction and rely on additional hardware or base stations, making them difficult to apply universally.
By collecting GNSS observation data from the user terminal, preprocessing and feature extraction are performed to construct epoch feature vectors. A machine learning error prediction model is used to predict and correct positioning errors. By combining observation data from multiple GNSS systems, multidimensional feature vectors are constructed and nonlinear models are used for error prediction.
It achieves real-time positioning coordinate correction without additional hardware in complex environments, significantly improving positioning accuracy and stability. It is suitable for remote and network-limited areas and has broad applicability and cross-system unified error modeling capabilities.
Smart Images
Figure CN122172237A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of navigation and positioning technology, and in particular to a positioning coordinate error correction method based on machine learning. Background Technology
[0002] Traditional Standard Point Positioning (SPP) methods are widely used in mobile devices or portable GNSS receivers, but due to limitations such as multipath effects, signal blockage, and non-ideal satellite geometry, they typically only provide meter-level positioning accuracy. In complex environments, such as urban canyons, forested areas, or densely populated areas, signal blockage and interference further increase positioning errors.
[0003] In related technologies, positioning results can be corrected using filters or differential methods. However, these methods typically rely on additional hardware, reference stations, or terrestrial networks, making them difficult to apply universally in global or remote areas. Furthermore, these technologies cannot fully utilize observational information such as satellite signal quality, pseudorange correction, and Doppler, limiting the ability to accurately model and predict errors and preventing real-time error correction for single-point positioning results.
[0004] Therefore, there is currently a lack of a method to perform high-precision correction of GNSS positioning coordinates in complex environments or remote environments lacking terrestrial network support. Summary of the Invention
[0005] This application provides a machine learning-based method for correcting positioning coordinate errors to address the shortcomings of the aforementioned related technologies. The technical solution is as follows: Firstly, this application provides a machine learning-based method for correcting positioning coordinate errors, including: GNSS observation data is collected by the user terminal, and the GNSS observation data is preprocessed to obtain preprocessed GNSS observation data. Feature extraction is performed based on the preprocessed GNSS observation data to obtain the epoch feature vector for each epoch. The epoch feature vector is input into the trained error prediction model, and the positioning error of the corresponding epoch is predicted by the error prediction model. The initial positioning coordinates obtained from the GNSS observation data are corrected based on the positioning error to obtain the corrected positioning coordinates.
[0006] In one alternative embodiment of the first aspect, the preprocessing of the GNSS observation data to obtain preprocessed GNSS observation data includes: Convert the observation time of the GNSS observation data into epoch time to unify the time format; The observation data from different GNSS systems were processed with unified column names, and missing values were filled for each type of observation. Preprocessed GNSS observation data are obtained.
[0007] In one alternative embodiment of the first aspect, the step of extracting features based on the preprocessed GNSS observation data to obtain the epoch feature vector corresponding to the epoch includes: For each epoch, the set of observed satellites for that epoch is extracted based on the preprocessed GNSS observation data; Satellite signal strength characteristics are obtained based on the statistical results of the signal strength of each satellite in the observed satellite set; The satellite quantity characteristics are obtained based on the number of satellites in each type of GNSS system within the observed satellite set; The satellite position parameter characteristics are obtained based on the statistical results of the elevation angle and azimuth angle of each satellite in the observed satellite set; The satellite observation characteristics are obtained based on the statistical results of the corrected pseudorange and Doppler shift of each satellite in the observation satellite set; Based on the satellite signal strength characteristics, the satellite quantity characteristics, the satellite position parameter characteristics, and the satellite observation value characteristics, an epoch feature vector for the corresponding epoch is constructed.
[0008] In one alternative of the first aspect, the satellite signal strength characteristics include the average, standard deviation, maximum, and minimum signal strength of all satellites within the observed satellite set; The satellite position parameter features include the average and standard deviation of the elevation angles of all satellites in the observation satellite set, as well as the signal strength-weighted average elevation angle, and also the average and standard deviation of the azimuth angles of all satellites in the observation satellite set. The satellite observation characteristics include the average and standard deviation of the corrected pseudorange of all satellites in the observation satellite set, as well as the average and standard deviation of the Doppler frequency shift of all satellites in the observation satellite set.
[0009] In one alternative embodiment of the first aspect, the method further includes: Sample GNSS observation data for historical periods were collected using GNSS receivers of the same type deployed in the same area as the user terminals. Based on the sample GNSS observation data, sample feature vectors for multiple epochs are constructed, and the single-point positioning coordinates of the GNSS receiver in each epoch are determined based on the sample GNSS observation data. Obtain the true coordinates of the GNSS receiver at each epoch, and obtain the error label based on the difference between the true coordinates and the single-point positioning coordinates of the corresponding epoch; Based on the sample feature vector of the GNSS receiver in each epoch and the single-point positioning coordinates in the corresponding epoch as sample inputs, and the error label as the supervised learning label, a sample set for training the error prediction model is constructed.
[0010] In one alternative embodiment of the first aspect, the training process of the error prediction model includes: The sample set is used as input, and the sample feature vectors are processed by the nonlinear mapping function of the error prediction model in the horizontal, vertical and vertical directions respectively, and the error value of the corresponding direction axis is output. Based on the supervised learning labels, an error label is determined for each direction axis. An optimization objective function is constructed based on the difference between the error label and the predicted error value for the corresponding direction axis. The model parameters of the error prediction model are adjusted based on the optimization objective function until the model converges, thus obtaining the trained error prediction model.
[0011] In one alternative embodiment of the first aspect, the optimization objective function is expressed as: ; ; ; Where N represents the total number of epochs, and t represents the epoch number. Error labels indicating the horizontal axis direction. Error labels indicating the vertical axis direction. Error labels indicating the vertical axis direction. This represents a nonlinear mapping function along the horizontal axis. This represents a nonlinear mapping function along the vertical axis. This represents a nonlinear mapping function along the vertical axis. This represents the eigenvector of epoch t.
[0012] Secondly, this application also provides a positioning coordinate error correction device based on machine learning, comprising: The data acquisition unit is used to acquire GNSS observation data through the user terminal, preprocess the GNSS observation data, and obtain preprocessed GNSS observation data. The feature extraction unit is used to extract features based on the preprocessed GNSS observation data to obtain the epoch feature vector for each epoch. An error calculation unit is used to input the epoch feature vector into a trained error prediction model, and to predict the positioning error of the corresponding epoch through the error prediction model. An error correction unit is used to correct the initial positioning coordinates obtained from the GNSS observation data based on the positioning error, so as to obtain the corrected positioning coordinates.
[0013] Thirdly, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method provided by the first aspect of this application or any implementation thereof.
[0014] Fourthly, this application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided by the first aspect of this application or any implementation thereof.
[0015] This application provides a machine learning-based positioning coordinate error correction method. Compared with existing technologies, it directly collects GNSS observation data from the user terminal and utilizes epoch-level feature extraction and error prediction models to achieve real-time coordinate correction without additional reference stations or differential signals. This significantly reduces deployment costs and communication dependence, and is especially suitable for remote and network-restricted areas.
[0016] Furthermore, in complex environments, this method constructs a multidimensional feature vector containing satellite signal strength, corrected pseudorange, and Doppler information, and combines it with a nonlinear model for error prediction. This effectively captures the error patterns caused by multipath effects and signal blockage, significantly improving the positioning accuracy and stability in urban canyons, densely populated high-rise areas, and forest-covered areas.
[0017] Meanwhile, this method fully integrates observation data from multiple GNSS systems, including GPS, Galileo, and BeiDou, during the feature construction stage, ensuring compatibility with different GNSS receiver models and possessing broad applicability and cross-system unified error modeling capabilities. Employing enhanced models such as XGBoost, and utilizing deep tree structures and regularization strategies, it not only accurately characterizes the nonlinear relationship between observation features and errors but also ensures the model's robustness and generalization ability across different environments and devices. The final high-precision corrected coordinates can be directly used in navigation, autonomous driving, robotics, and geographic information acquisition, providing reliable data support for precision surveying and emergency positioning. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a flowchart illustrating a machine learning-based positioning coordinate error correction method provided in an embodiment of this application. Figure 2 This is a schematic diagram of a positioning coordinate error correction device based on machine learning provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions 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, 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.
[0021] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or modules is not limited to the steps or modules listed, but may optionally include steps or modules not listed, or may optionally include other steps or modules inherent to such process, method, product, or apparatus.
[0022] It should be noted that the terms "first" and "second" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects. It is understood that "first" and "second" can be interchanged in a specific order or sequence where permitted. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in an order other than those described or illustrated herein.
[0023] The present application will now be described in detail with reference to specific embodiments.
[0024] Next, combine Figure 1 This paper introduces a machine learning-based positioning coordinate error correction method provided in an embodiment of this application. For details, please refer to... Figure 1 , Figure 1 This illustration shows a flowchart of a machine learning-based positioning coordinate error correction method provided in an embodiment of this application. Figure 1 As shown, the method includes the following steps: S101, GNSS observation data for each epoch is collected through the client, and the GNSS observation data is preprocessed to obtain preprocessed GNSS observation data; S102, Based on the preprocessed GNSS observation data, feature extraction is performed to obtain the epoch feature vector of the corresponding epoch; S103, input the epoch feature vector into the trained error prediction model, and predict the positioning error of the corresponding epoch through the error prediction model; S104, Based on the positioning error, the initial positioning coordinates obtained from the GNSS observation data are corrected to obtain the corrected positioning coordinates.
[0025] Specifically, in S101, the client can receive signals from GNSS satellites, identify the observed GNSS satellites, and thus collect GNSS observation data for each epoch, including the satellite signal strength of each observed satellite. Altitude angle Azimuth pseudo-distance , corrected pseudorange Doppler frequency shift Observational data, etc.
[0026] Specifically, in S101, the GNSS observation data can be preprocessed to obtain preprocessed GNSS observation data, including: The observation time of the GNSS observation data is converted into epoch time, and the time format is standardized, for example, the observation time (year, month, day + hour, minute, second) is converted into epoch time. The unified time format is This facilitates epoch-level data aggregation.
[0027] In addition, the observation data from different GNSS systems can be uniformly named and missing values can be filled for each type of observation to ensure the integrity of the feature matrix. This leads to the preprocessed GNSS observation data.
[0028] It should be noted that GNSS systems include GPS, BeiDou, and Galileo systems. When observation data from different GNSS systems are merged into a single dataset, the naming of the same type of observation data often differs between systems. It is necessary to unify the naming of each type of observation data, that is, to unify the same column naming rules for observation data of the same type, so as to facilitate processing by subsequent algorithms (such as machine learning or positioning algorithms).
[0029] In some embodiments, in S102, feature extraction can be performed based on the preprocessed GNSS observation data to obtain the epoch feature vector corresponding to the epoch, including: For each epoch, the set of observed satellites for that epoch is extracted based on the preprocessed GNSS observation data, denoted as: ; in, This represents the set of observed satellites corresponding to epoch t. Represents the i-th observation satellite, This represents the total number of observed satellites corresponding to epoch t.
[0030] Specifically, satellite signal strength characteristics can be obtained based on the statistical results of the signal strength of each satellite within the observed satellite set, including: Based on the statistical results, the average, standard deviation, maximum, and minimum signal strength of all satellites in the observed satellite set can be obtained.
[0031] Among them, the average signal strength of all satellites within the observation satellite ensemble Represented as: ; Standard deviation of signal strength of all satellites in the observation satellite ensemble Represented as: ; The maximum signal strength of all satellites within the observed satellite ensemble is expressed as: The minimum value is expressed as .
[0032] The above statistical results reflect the signal quality distribution of all satellites within the observed satellite set.
[0033] Specifically, the satellite quantity characteristics can be obtained based on the number of satellites for each type of GNSS system within the observed satellite set, including: ; in, The cardinality of a set is the number of elements in the set. This represents the total number of observation satellites of all GNSS systems within epoch t.
[0034] Specifically, the satellite quantity characteristic also includes the number of observation satellites for each type of GNSS system, for example... This represents the number of observable GPS satellites at epoch t. This represents the number of Galileo satellites observable at epoch t. This represents the number of observable BeiDou satellites in epoch t.
[0035] Optionally, the calculation may include other satellite systems, such as R representing the GLONASS satellite system and J representing the Quasi-Zenith Satellite System (QZSS). This application does not limit this.
[0036] For example, at a certain epoch, the set of observed satellites can be represented as: ; in, These represent the GPS satellites with serial numbers 03, 08, and 15, respectively. This refers to the Galileo satellite with serial number 11. These represent the BeiDou satellites with serial numbers 19 and 24, respectively.
[0037] Specifically, satellite position parameter characteristics can be obtained based on the statistical results of the elevation angle and azimuth angle of each satellite in the observed satellite set, including: Based on the statistical results, the average elevation angle, standard deviation, and signal strength-weighted average elevation angle of all satellites in the observed satellite set can be obtained.
[0038] Among them, the average elevation angle of all satellites in the observation satellite set. Represented as: ; Standard deviation of the elevation angle of all satellites in the observation satellite ensemble Represented as: ; To give satellites with higher signal quality a greater weight in the statistical results, a method based on signal strength is introduced. Average elevation angle as a weight , represented as: ; in, Represents the epoch No. satellite The elevation angle represents the angle between the satellite and the ground plane of the observation station, and its value ranges from 0°. 90°.
[0039] Similarly, statistical analysis of the azimuth angles can be performed to obtain the average and standard deviation of the azimuth angles of all satellites within the observed satellite ensemble, including: The average azimuth angle of all satellites in the observation satellite ensemble , represented as: ; Standard deviation of azimuth angle of all satellites in the observation satellite ensemble , represented as: ; in, Indicates satellite The azimuth angle is the direction angle of a satellite on the horizontal plane, and its value ranges from 0° to 360°.
[0040] The epoch can be characterized by satellite position parameter features. The corresponding spatial distribution characteristics of the observed satellites in the sky are analyzed. Among them, the elevation angle statistics mainly describe the distribution of satellites in the vertical direction, while the azimuth angle statistics reflect the distribution of satellites in the horizontal direction. These geometric features can effectively characterize the satellite geometry and are of great significance for assessing positioning geometry and signal environment.
[0041] Specifically, satellite observation characteristics can be obtained based on the statistical results of the corrected pseudorange and Doppler frequency shift of each satellite in the observation satellite set, including: Based on the statistical results, the average and standard deviation of the corrected pseudoranges of all satellites in the observed satellite set can be obtained.
[0042] Among them, the average corrected pseudorange of all satellites in the observation satellite set. Represented as: ; Standard deviation of the corrected pseudorange of all satellites in the observation satellite ensemble Represented as: ; Similarly, the average and standard deviation of the Doppler shift of all satellites in the observed satellite set can be obtained.
[0043] Among them, the average Doppler frequency shift of all satellites in the observation satellite set. Represented as: ; The standard deviation of the Doppler shift of all satellites in the observation satellite ensemble Represented as: ; in, Indicates satellite Corrected pseudorange, Indicates satellite The Doppler shift, based on the characteristics of satellite observations, can represent the degree of dispersion of observations within each epoch, thereby measuring the stability of the signal.
[0044] Based on the satellite signal strength features, satellite quantity features, satellite position parameter features, and satellite observation features extracted above, the epoch feature vector for the corresponding epoch can be represented as:
[0045]
[0046]
[0047] Wherein, the epoch feature vector corresponding to epoch t , The feature dimension is represented in the current embodiment. .
[0048] Optionally, other features can be added to the epoch feature vector according to actual needs, such as a specific satellite list, geometric features such as PDOP and HDOP, and the corresponding feature dimensions will also change accordingly. This application does not limit this.
[0049] In some embodiments, the error prediction model in S103 can be trained based on a pre-constructed sample set. The step of constructing the sample set includes: S201 collects sample GNSS observation data for historical periods using GNSS receivers of the same type deployed in the same area as the user terminal.
[0050] It should be noted that the user end is a GNSS receiver. In order to ensure that the model can cover the errors of the device itself and that the model can correct the errors of the static reference point, a sample set for training can be constructed using sample GNSS observation data from GNSS receivers of the same type and located in the same region as the user end.
[0051] S202, based on the sample GNSS observation data, construct sample feature vectors for multiple epochs, and determine the single-point positioning coordinates of the GNSS receiver in each epoch based on the sample GNSS observation data.
[0052] It should be noted that the process of constructing the sample feature vector is the same as in S102. Please refer to the description in S102 for details, which will not be repeated here.
[0053] Specifically, the coordinates of a single point in a given epoch can be obtained based on traditional single-point positioning methods. ,in, These represent the single-point positioning coordinates of epoch t on the horizontal axis ( (axis), vertical axis ( (axis), vertical axis ( The components of the axis.
[0054] S203: Obtain the true coordinates of the GNSS receiver at each epoch, and obtain the error label based on the difference between the true coordinates and the single-point positioning coordinates of the corresponding epoch.
[0055] Specifically, the real coordinates can be represented as Real coordinates Single-point positioning coordinates Divide the results and obtain the following: axis, axis, The error value in the axial direction is expressed as: ; in, Indicates error label, This indicates the component of the error label along the horizontal axis. This indicates the component of the error label along the vertical axis. This indicates the component of the error label in the vertical direction.
[0056] S204. Based on the sample feature vector of the GNSS receiver in each epoch and the single-point positioning coordinates in the corresponding epoch as sample inputs, and using the error label as the supervised learning label, a sample set for training the error prediction model is constructed.
[0057] Next, the error prediction model can be trained based on the sample set obtained from the above embodiments, specifically including the following steps: The sample set is used as input, and the sample feature vectors are processed by the nonlinear mapping function of the error prediction model in the horizontal, vertical and vertical directions respectively, and the error value of the corresponding direction axis is output. Based on the supervised learning labels, an error label is determined for each direction axis. An optimization objective function is constructed based on the difference between the error label and the predicted error value for the corresponding direction axis. The model parameters of the error prediction model are adjusted based on the optimization objective function until the model converges, thus obtaining the trained error prediction model.
[0058] Specifically, the enhanced XGBoost regression algorithm can be used to train three directional error models separately, as shown in the following formula: ; in, Represents the residual in the horizontal direction. Represents the residual along the vertical axis. Represents the residual in the vertical direction. This represents a nonlinear mapping function along the horizontal axis. This represents a nonlinear mapping function along the vertical axis. This represents a nonlinear mapping function along the vertical axis. This represents the eigenvector of epoch t.
[0059] The objective function is to minimize the mean squared error (MSE), which is expressed by the following formula: ; ; ; Specifically, during training, KFold cross-validation and early stopping strategies can be used to avoid overfitting. Parameters such as tree depth, learning rate, and regularization coefficient can be set to enhance the model's generalization ability, thereby obtaining a well-trained error prediction model.
[0060] In some embodiments, in S103, the localization error of the corresponding epoch can be predicted by the trained error prediction model based on the epoch feature vector constructed in S102, and is expressed as: ; in, The epochal eigenvectors constructed for S102, The positioning error at epoch t predicted by the model. These represent the components of the positioning error in the corresponding direction for epoch t predicted by the model.
[0061] In some embodiments, in S104, the initial positioning coordinates obtained based on the GNSS observation data can be corrected based on the positioning error to obtain the corrected positioning coordinates, expressed by the formula: ; in, This represents the initial positioning coordinates of the user terminal obtained using the single-point positioning method based on GNSS observation data from S101. This indicates the corrected positioning coordinates.
[0062] Specifically, the corrected positioning coordinates can be output in standard POS or CSV format for navigation, surveying, or other high-precision positioning applications.
[0063] The following are apparatus embodiments of this application, which can be used to execute the method embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the method embodiments of this application.
[0064] Please see below. Figure 2 The diagram below illustrates a machine learning-based positioning coordinate error correction device, which is an exemplary embodiment of this application. The device includes: The data acquisition unit is used to acquire GNSS observation data through the user terminal, preprocess the GNSS observation data, and obtain preprocessed GNSS observation data. The feature extraction unit is used to extract features based on the preprocessed GNSS observation data to obtain the epoch feature vector for each epoch. An error calculation unit is used to input the epoch feature vector into a trained error prediction model, and to predict the positioning error of the corresponding epoch through the error prediction model. An error correction unit is used to correct the initial positioning coordinates obtained from the GNSS observation data based on the positioning error, so as to obtain the corrected positioning coordinates.
[0065] It should be noted that the apparatus provided in the above embodiments, when executing a machine learning-based positioning coordinate error correction method, is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and their implementation process is detailed in the method embodiments, which will not be repeated here.
[0066] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods described above.
[0067] Please see Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application.
[0068] like Figure 3 As shown, the electronic device includes a processor and a memory.
[0069] In this embodiment, the processor is the control center of the computer system, and can be a processor of a physical machine or a processor of a virtual machine. The processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor can be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array).
[0070] A processor can also include a main processor and a coprocessor. The main processor is used to process data in the wake-up state and is also called the CPU (Central Processing Unit). The coprocessor is a low-power processor used to process data in the standby state.
[0071] The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments of this application, the non-transitory computer-readable storage media in the memory are used to store at least one instruction, which is executed by a processor to implement the methods in the embodiments of this application.
[0072] In some embodiments, the electronic device further includes a peripheral device interface and at least one peripheral device. The processor, memory, and peripheral device interface are connected via a bus or signal line. Each peripheral device is connected to the peripheral device interface via a bus, signal line, or circuit board. Specifically, the peripheral device includes: a display screen, a camera, and audio circuitry. The peripheral device interface can be used to connect at least one I / O (Input / Output) related peripheral device to the processor and memory.
[0073] In some embodiments of this application, the processor, memory, and peripheral device interfaces are integrated on the same chip or circuit board; in other embodiments of this application, any one or two of the processor, memory, and peripheral device interfaces can be implemented on separate chips or circuit boards. This application does not specifically limit the implementation in this regard.
[0074] The electronic device structural block diagrams shown in the embodiments of this application do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or use different component arrangements.
[0075] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the methods in any of the foregoing embodiments. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.
[0076] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A machine learning-based method for correcting positioning coordinate errors, characterized in that, include: GNSS observation data is collected by the user terminal, and the GNSS observation data is preprocessed to obtain preprocessed GNSS observation data. Feature extraction is performed based on the preprocessed GNSS observation data to obtain the epoch feature vector for each epoch. The epoch feature vector is input into the trained error prediction model, and the positioning error of the corresponding epoch is predicted by the error prediction model. The initial positioning coordinates obtained from the GNSS observation data are corrected based on the positioning error to obtain the corrected positioning coordinates.
2. The machine learning-based positioning coordinate error correction method according to claim 1, characterized in that, The preprocessing of the GNSS observation data to obtain preprocessed GNSS observation data includes: Convert the observation time of the GNSS observation data into epoch time to unify the time format; The observation data from different GNSS systems were processed with unified column names, and missing values were filled for each type of observation. Preprocessed GNSS observation data are obtained.
3. A machine learning-based positioning coordinate error correction method according to claim 1 or 2, characterized in that, The feature extraction based on the preprocessed GNSS observation data to obtain the epoch feature vector for the corresponding epoch includes: For each epoch, the set of observed satellites for that epoch is extracted based on the preprocessed GNSS observation data; Satellite signal strength characteristics are obtained based on the statistical results of the signal strength of each satellite in the observed satellite set; The satellite quantity characteristics are obtained based on the number of satellites in each type of GNSS system within the observed satellite set; The satellite position parameter characteristics are obtained based on the statistical results of the elevation angle and azimuth angle of each satellite in the observed satellite set; The satellite observation characteristics are obtained based on the statistical results of the corrected pseudorange and Doppler shift of each satellite in the observation satellite set; Based on the satellite signal strength characteristics, the satellite quantity characteristics, the satellite position parameter characteristics, and the satellite observation value characteristics, an epoch feature vector for the corresponding epoch is constructed.
4. The machine learning-based positioning coordinate error correction method according to claim 3, characterized in that, The satellite signal strength characteristics include the average, standard deviation, maximum, and minimum signal strength of all satellites within the observed satellite set; The satellite position parameter features include the average and standard deviation of the elevation angles of all satellites in the observation satellite set, as well as the signal strength-weighted average elevation angle, and also the average and standard deviation of the azimuth angles of all satellites in the observation satellite set. The satellite observation characteristics include the average and standard deviation of the corrected pseudorange of all satellites in the observation satellite set, as well as the average and standard deviation of the Doppler frequency shift of all satellites in the observation satellite set.
5. The machine learning-based positioning coordinate error correction method according to claim 3, characterized in that, The method further includes: Sample GNSS observation data for historical periods were collected using GNSS receivers of the same type deployed in the same area as the user terminals. Based on the sample GNSS observation data, sample feature vectors for multiple epochs are constructed, and the single-point positioning coordinates of the GNSS receiver in each epoch are determined based on the sample GNSS observation data. Obtain the true coordinates of the GNSS receiver at each epoch, and obtain the error label based on the difference between the true coordinates and the single-point positioning coordinates of the corresponding epoch; Based on the sample feature vector of the GNSS receiver in each epoch and the single-point positioning coordinates in the corresponding epoch as sample inputs, and the error label as the supervised learning label, a sample set for training the error prediction model is constructed.
6. The machine learning-based positioning coordinate error correction method according to claim 5, characterized in that, The training process of the error prediction model includes: The sample set is used as input, and the sample feature vectors are processed by the nonlinear mapping function of the error prediction model in the horizontal, vertical and vertical directions respectively, and the error value of the corresponding direction axis is output. Based on the supervised learning labels, an error label is determined for each direction axis. An optimization objective function is constructed based on the difference between the error label and the predicted error value for the corresponding direction axis. The model parameters of the error prediction model are adjusted based on the optimization objective function until the model converges, thus obtaining the trained error prediction model.
7. The machine learning-based positioning coordinate error correction method according to claim 6, characterized in that, The optimization objective function is expressed as: ； ； ； Where N represents the total number of epochs, and t represents the epoch number. Error labels indicating the horizontal axis direction. Error labels indicating the vertical axis direction. Error labels indicating the vertical axis direction. This represents a nonlinear mapping function along the horizontal axis. This represents a nonlinear mapping function along the vertical axis. This represents a nonlinear mapping function along the vertical axis. This represents the eigenvector of epoch t.
8. A positioning coordinate error correction device based on machine learning, characterized in that, include: The data acquisition unit is used to acquire GNSS observation data through the user terminal, preprocess the GNSS observation data, and obtain preprocessed GNSS observation data. The feature extraction unit is used to extract features based on the preprocessed GNSS observation data to obtain the epoch feature vector for each epoch. An error calculation unit is used to input the epoch feature vector into a trained error prediction model, and to predict the positioning error of the corresponding epoch through the error prediction model. An error correction unit is used to correct the initial positioning coordinates obtained from the GNSS observation data based on the positioning error, so as to obtain the corrected positioning coordinates.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1 to 6.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.