GNSS Inter-Frequency Bias Solving Method and System Based on Graph Neural Network

Abstract

The present invention discloses a method and system for solving GNSS inter-frequency bias based on graph neural network. The method includes: obtaining GNSS observation data, and calculating the inter-frequency difference value of phase-smoothed pseudorange and pierce point information; inputting the inter-frequency difference value of phase-smoothed pseudorange and pierce point information into a trained graph neural network model to output the GNSS inter-frequency bias. The training of the graph neural network model includes: obtaining GNSS observation data and precise inter-frequency bias products of global stations, and calculating the inter-frequency difference value of phase-smoothed pseudorange and pierce point information for each station; using the inter-frequency difference value of phase-smoothed pseudorange and pierce point information as input and the precise inter-frequency bias product as output to construct a data set; based on the graph neural network architecture, using satellites and receivers as nodes and pierce point information and the inter-frequency difference value of phase-smoothed pseudorange as edge attributes to perform embedding transformation to form a graph neural network model; using the constructed data set to train the graph neural network model and output the trained model.

Description

Technical Field

[0001] The present invention belongs to the field of solving GNSS inter-frequency bias parameters, and particularly relates to a method and system for solving GNSS inter-frequency bias based on a graph neural network. Background Art

[0002] In the positioning of the Global Navigation Satellite System (GNSS), the differential code bias (hereinafter referred to as "DCB") is an error caused by the difference in the propagation speed of signals with different frequencies in the ionosphere. Accurately solving the DCB is crucial for improving the positioning accuracy of GNSS.

[0003] Traditional DCB solving methods usually rely on GNSS observation data worldwide, establish a smooth correction model for ionospheric delay based on the phase-smoothed pseudo-range inter-frequency difference value and spherical harmonic functions, and separate and solve the DCB while calculating the model coefficients. This method assumes that the ionospheric correction value is smooth in space, but the actual ionospheric correction value often has significant spatio-temporal non-stationarity. Due to this assumption deviation, the systematic error of the ionospheric correction model will be absorbed by the DCB during the solving process, thus affecting the solving accuracy of the DCB, especially when the ionospheric conditions are complex or dynamically changing, the error is more obvious.

[0004] The term explanations involved in the specification of the present invention are as follows:

[0005] GNSS: Global Navigation Satellite System, the global navigation satellite system;

[0006] DCB: Differential Code Bias, differential code bias, that is, inter-frequency bias. Summary of the Invention

[0007] To solve the problem that the existing method depends on the smooth assumption and affects the DCB solving accuracy, the present invention provides a method and system for solving GNSS inter-frequency bias based on a graph neural network, which directly uses the spatio-temporal dependence relationship between satellites and receivers for modeling, and combines the graph neural network for node information transmission to achieve accurate prediction of the DCB.

[0008] According to one aspect of the specification of the present invention, a method for solving GNSS inter-frequency bias based on a graph neural network is provided, including:

[0009] Obtain GNSS observation data, and calculate the phase-smoothed pseudo-range inter-frequency difference value and pierce point information according to the GNSS observation data;

[0010] Input the calculated phase-smoothed pseudorange inter-frequency difference value and the pierce point information into the trained graph neural network model to output the GNSS inter-frequency bias; wherein, the training of the graph neural network model includes:

[0011] Obtain the GNSS observation data and precise inter-frequency bias products of global stations, and calculate the phase-smoothed pseudorange inter-frequency difference value and the pierce point information for each station based on the GNSS observation data;

[0012] Construct a data set with the calculated phase-smoothed pseudorange inter-frequency difference value and the pierce point information as the input and the precise inter-frequency bias product as the output;

[0013] Based on the graph neural network architecture, use satellites and receivers as nodes and the pierce point information and the phase-smoothed pseudorange inter-frequency difference value as edge attributes to perform embedding transformation to form a graph neural network model;

[0014] Use the constructed data set to train the graph neural network model and output the trained model.

[0015] As a further technical solution, the calculation of the pierce point information includes:

[0016] Calculate the longitude, latitude, satellite elevation angle and azimuth angle of the pierce point.

[0017] As a further technical solution, constructing the data set further includes:

[0018] Use the phase-smoothed pseudorange inter-frequency difference value, the pierce point information and the precise inter-frequency bias product of one day to form a data sample, and form a data set according to the data samples of several days.

[0019] As a further technical solution, using the constructed data set to train the graph neural network model includes:

[0020] Input the data set into the constructed graph neural network model, use the mean square error MSE as the loss function, and use the Adam optimization algorithm for backpropagation to perform model training.

[0021] According to one aspect of the specification of the present invention, there is provided a GNSS inter-frequency bias solving system based on a graph neural network, including:

[0022] A first calculation module for obtaining GNSS observation data and calculating the phase-smoothed pseudorange inter-frequency difference value and the pierce point information according to the GNSS observation data;

[0023] A second calculation module for inputting the calculated phase-smoothed pseudorange inter-frequency difference value and the pierce point information into the trained graph neural network model to output the GNSS inter-frequency bias; wherein, the training of the graph neural network model includes:

[0024] Obtain GNSS observation data and precise inter-frequency bias products for global stations, and calculate the inter-frequency difference value of phase-smoothed pseudorange and pierce point information for each station based on the GNSS observation data;

[0025] Use the calculated inter-frequency difference value of phase-smoothed pseudorange and pierce point information as input, and the precise inter-frequency bias product as output to construct a data set;

[0026] Based on the graph neural network architecture, with satellites and receivers as nodes, and pierce point information and the inter-frequency difference value of phase-smoothed pseudorange as edge attributes, perform embedding transformation to form a graph neural network model;

[0027] Use the constructed data set to train the graph neural network model and output the trained model.

[0028] According to one aspect of the specification of the present invention, there is provided a GNSS inter-frequency bias solving device based on a graph neural network, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the GNSS inter-frequency bias solving method based on the graph neural network.

[0029] According to one aspect of the specification of the present invention, there is provided a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the GNSS inter-frequency bias solving method based on the graph neural network.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] The present invention designs a graph neural network applicable to GNSS DCB and implements a high-precision solving method for DCB. This method no longer depends on the smooth assumption of the ionospheric correction value, but bypasses the estimation of the ionospheric correction and directly models the spatio-temporal dependence relationship between satellites and receivers. By utilizing the characteristic that the graph neural network is naturally suitable for processing such strongly spatio-temporally correlated data, through the information transfer between nodes, the deep connection between historical precise DCB data and observation values is adaptively learned. Through this data-driven method, the model can more accurately predict DCB without relying on traditional smooth assumptions and show higher robustness and adaptability in a dynamic ionospheric environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0033] Figure 1 It is a schematic flowchart of the GNSS inter-frequency bias solving method based on graph neural network provided by the embodiment of the present invention.

[0034] Figure 2 It is a schematic flowchart of the training process of the graph neural network model provided by the embodiment of the present invention.

[0035] Figure 3 It is a schematic structural diagram of the GNSS inter-frequency bias solving system based on graph neural network provided by the embodiment of the present invention. Detailed implementation manners

[0036] It should be noted that:

[0037] When solving the DCB in the prior art, it is assumed that the ionospheric correction value is smooth in space. Due to this assumption deviation, the systematic error of the ionospheric correction model is absorbed by the DCB during the solving process, thus affecting the solving accuracy of the DCB, especially when the ionospheric conditions are complex or dynamically changing, the error is more obvious. To solve this problem, considering that the neural network, as a powerful data fitting and feature extraction tool, has made remarkable progress in many fields, the present invention proposes a GNSS inter-frequency bias solving method based on graph neural network, which solves the problem that the solving accuracy of the DCB in the prior art is affected by the smooth assumption of the ionospheric correction value.

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention. In addition, the technical features in each embodiment or a single embodiment provided by the present invention can be combined with each other arbitrarily to form a new technical solution. This combination is not restricted by the order of steps and / or the structural composition mode, but must be based on the fact that those of ordinary skill in the art can implement it. When the combination of technical solutions results in contradictions or cannot be implemented, it should be considered that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0039] An embodiment of the present invention provides a method for solving the GNSS inter-frequency bias based on a graph neural network. As Figure 1 shown, first, GNSS observation data is obtained, and the inter-frequency difference value of the phase-smoothed pseudorange and the pierce point information are calculated according to the GNSS observation data; then, the calculated inter-frequency difference value of the phase-smoothed pseudorange and the pierce point information are input into the trained graph neural network model, and the GNSS inter-frequency bias is output.

[0040] The method provided by the embodiment of the present invention bypasses the estimation of ionospheric correction, directly utilizes the spatio-temporal dependence relationship between the satellite and the receiver, and combines the graph neural network model to solve the GNSS inter-frequency bias, that is, the precise DCB product. It not only avoids the dependence on the smoothness assumption but also shows higher robustness and adaptability in a dynamic ionospheric environment.

[0041] The embodiment of the present invention is described based on the IGS global observation stations and the precise DCB product for model training. As Figure 2 shown, the training of the graph neural network model includes:

[0042] S1, Obtain the GNSS observation data and the precise inter-frequency bias product of the global stations, and calculate the inter-frequency difference value of the phase-smoothed pseudorange and the pierce point information for each station based on the GNSS observation data.

[0043] Specifically, it further includes:

[0044] S11, Download the GNSS observation data and the precise DCB product of the global stations. For example, download the GNSS observation data of the IGS global stations from 2020 to 2023 and the daily precise DCB product of the CODE analysis center.

[0045] S12, Calculate the inter-frequency difference observation value of the phase-smoothed pseudorange of the GNSS station for each station, as well as the pierce point position, satellite elevation angle, and azimuth angle at the intersection of the observation ray and the ionosphere.

[0046] S2, Using the calculated inter-frequency difference value of the phase-smoothed pseudorange and the pierce point information as the input, and the precise inter-frequency bias product as the output, construct a data set.

[0047] Specifically, construct data samples with single-day data. Use the pierce point information (including pierce point longitude and latitude, satellite elevation angle, azimuth angle) and the calculated inter-frequency difference value of the phase-smoothed pseudorange as the input data, and the precise DCB product as the output data. The input data and output data of one day form a data sample, and the data samples of multiple days form the entire data set.

[0048] S3, Based on the graph neural network architecture, using satellites and receivers as nodes respectively, and using the pierce point information and the inter-frequency difference value of the phase-smoothed pseudorange as the observation edge attributes, perform embedding transformation respectively and then construct the graph neural network model.

[0049] Specifically, when constructing the graph neural network model, satellites and receivers are used as two types of nodes respectively, and the longitude and latitude of the piercing point, the satellite elevation angle, azimuth angle, and inter-frequency difference value are used as edge attributes. Embedding transformation is performed to map these discrete data into a low-dimensional continuous space, so that similar data is close in the embedding space, facilitating the neural network to capture relationships. Through information transmission and aggregation operations between nodes, spatio-temporal data modeling is carried out.

[0050] The graph convolution is expressed as:

[0051]

[0052] Among them, is the feature vector of node i at layer t, is the set of neighbor nodes of node i, is the weight matrix, is the bias term, is the activation function (such as ReLU).

[0053] S4. Use the constructed dataset to train the graph neural network model and output the trained model.

[0054] Divide the dataset into a training set, a validation set, and a test set, and apply the graph neural network model. Train the graph neural network model according to the input of training set data, the iteration of graph convolution operation features, the minimization of the loss function, and the optimization of parameters by the backpropagation algorithm until the training converges.

[0055] Furthermore, input the dataset into the constructed model, and at the same time use the mean square error MSE as the loss function, and adopt the Adam optimization algorithm for backpropagation to perform model training.

[0056] In the actual application of the embodiments of the present invention, the trained graph neural network model and the latest site observation data are used to calculate spatio-temporal data, and inferences are made based on the calculated spatio-temporal data to obtain the corresponding DCB. Compared with the existing methods, the DCB obtained by the embodiments of the present invention has higher accuracy and shows higher robustness and adaptability in a dynamic ionospheric environment.

[0057] The implementation basis of each embodiment of the present invention is achieved through programmed processing by a device with processor functions. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this actual situation, on the basis of the above embodiments, an embodiment of the present invention provides a GNSS inter-frequency bias solving system based on a graph neural network, and this system is used to execute the GNSS inter-frequency bias solving method based on a graph neural network in the above method embodiments.

[0058] SeeFigure 3 The system includes: a first calculation module, configured to obtain GNSS observation data and calculate the differential value of the phase-smoothed pseudorange frequency difference and the puncture point information according to the GNSS observation data; a second calculation module, configured to input the calculated differential value of the phase-smoothed pseudorange frequency difference and the puncture point information into a trained graph neural network model to output the GNSS inter-frequency bias; wherein the training of the graph neural network model includes: obtaining the GNSS observation data and precise inter-frequency bias products of global stations, and calculating the differential value of the phase-smoothed pseudorange frequency difference and the puncture point information for each station based on the GNSS observation data; using the calculated differential value of the phase-smoothed pseudorange frequency difference and the puncture point information as inputs and the precise inter-frequency bias products as outputs to construct a data set; based on the graph neural network architecture, using satellites and receivers as nodes and the puncture point information and the differential value of the phase-smoothed pseudorange frequency difference as edge attributes to perform embedding transformation to form a graph neural network model; and training the graph neural network model using the constructed data set to output a trained model.

[0059] The GNSS inter-frequency bias solving system based on a graph neural network provided by an embodiment of the present invention addresses the problem that the existing method relies on a smooth assumption, resulting in the accuracy of DCB solving being affected. It adopts Figure 3 several modules, directly models the spatio-temporal dependence relationship between satellites and receivers, combines the graph neural network for node information transmission, and realizes accurate prediction of DCB.

[0060] It should be noted that the system embodiment provided by the present invention is used not only to implement the method in the above method embodiment, but also to implement the methods in other method embodiments provided by the present invention. The difference is only in setting corresponding functional modules, and its principle is basically the same as that of the above system embodiment provided by the present invention. As long as those skilled in the art, based on the above system embodiment, refer to the specific technical solutions in other method embodiments, obtain corresponding technical means by combining technical features, and the technical solutions composed of these technical means, and on the premise of ensuring the practicality of the technical solutions, improve the modules in the above system embodiment to obtain corresponding system-like embodiments for implementing the methods in other method-like embodiments. For example:

[0061] Based on the content of the above system embodiment, as a preferred embodiment, in the GNSS inter-frequency bias solving system based on a graph neural network provided by an embodiment of the present invention, the first calculation module is further configured to execute the following instructions:

[0062] Calculate the longitude, latitude, satellite elevation angle, and azimuth angle of the puncture point.

[0063] Based on the content of the above system embodiments, as a preferred embodiment, in the GNSS inter-frequency bias solving system based on graph neural network provided in the embodiments of the present invention, the second calculation module is further configured to execute the following instructions:

[0064] Construct a data sample with the one-day phase-smoothed pseudorange inter-frequency difference value, puncture point information, and precise inter-frequency bias product, and form a data set according to the data samples of several days.

[0065] Based on the content of the above system embodiments, as a preferred embodiment, in the GNSS inter-frequency bias solving system based on graph neural network provided in the embodiments of the present invention, the second calculation module is further configured to execute the following instructions:

[0066] Input the data set into the constructed graph neural network model, use the mean squared error (MSE) as the loss function, and perform backpropagation using the Adam optimization algorithm for model training.

[0067] Based on the same inventive concept as the above embodiments, the embodiments of the present invention further provide a GNSS inter-frequency bias solving device based on graph neural network, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the GNSS inter-frequency bias solving method based on graph neural network.

[0068] In the embodiments of the present invention, the memory can be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or a volatile memory, such as a random-access memory (RAM). The memory is any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in the embodiments of the present invention can also be a circuit or any other device capable of implementing a storage function, for storing program instructions and / or data.

[0069] In the embodiments of the present invention, the processor can be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor can be a microprocessor or any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0070] Based on the same inventive concept as the above embodiments, an embodiment of the present invention further provides a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the GNSS inter-frequency bias solving method based on a graph neural network described above.

[0071] In summary of the above embodiments, the present invention designs a graph neural network applicable to GNSS DCB and realizes a high-precision solving method for DCB, which has the following advantages:

[0072] The present invention no longer relies on the smooth assumption of the ionospheric correction value, but skips the estimation of the ionospheric correction and directly models the spatio-temporal dependence relationship between satellites and receivers.

[0073] The present invention utilizes the characteristics of the graph neural network suitable for processing such strongly spatio-temporally correlated data, and adaptively learns the deep connection between the historical precise DCB data and the observed values through information transmission between nodes.

[0074] Without relying on the traditional smooth assumption, the present invention can more accurately predict DCB and show higher robustness and adaptability in a dynamic ionospheric environment.

[0075] The terms "including" and "having" in the specification, claims and drawings of the present invention, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A GNSS inter-frequency deviation solution method based on graph neural network, characterized in that: Independent of relying on smooth assumptions about ionospheric corrections, the method comprises: Acquire GNSS observation data, and calculate phase-smoothed pseudorange inter-frequency difference value and puncture point information according to the GNSS observation data; The calculated phase-smoothed pseudorange frequency difference value and puncture point information are input into the trained graph neural network model to output the GNSS frequency deviation; wherein the training of the graph neural network model includes: Obtain GNSS observation data and precise inter-frequency deviation products of global sites, and calculate the phase-smoothed pseudo-range inter-frequency difference value and puncture point information for each site based on the GNSS observation data; The calculated phase-smoothed pseudorange frequency difference value and puncture point information are used as input, and the precise frequency deviation product is used as output to construct a data set; Based on the graph neural network architecture, satellites and receivers are used as nodes, puncture point information and phase-smoothed pseudorange frequency difference values ​​are used as edge attributes, and embedded transformation is performed to form a graph neural network model. Use the constructed dataset to train the graph neural network model and output the trained model.

2. According to the method for solving GNSS inter-frequency deviation based on graph neural network in claim 1, it is characterized in that: The calculation of the puncture point information includes: Calculate the latitude and longitude, satellite altitude angle and azimuth of the puncture point.

3. According to the GNSS inter-frequency deviation solution method based on graph neural network according to claim 1, it is characterized in that: Building a dataset also includes: A data sample is composed of one day's phase-smoothed pseudorange frequency difference value, puncture point information, and precise frequency deviation product, and a data set is composed of several days' data samples.

4. The GNSS inter-frequency deviation solution method based on graph neural network according to claim 1 is characterized in that: Use the constructed dataset to train the graph neural network model, including: The data set is input into the constructed graph neural network model, the mean square error (MSE) is used as the loss function, and the Adam optimization algorithm is used for back propagation to perform model training.

5. The GNSS inter-frequency deviation solution system based on graph neural network is characterized by: Independent of relying on smooth assumptions about ionospheric corrections, the system comprises: A first calculation module is used to obtain GNSS observation data and calculate the phase smoothed pseudorange frequency difference value and puncture point information according to the GNSS observation data; The second calculation module is used to input the calculated phase-smoothed pseudorange frequency difference value and puncture point information into the trained graph neural network model, and output the GNSS frequency deviation; wherein the training of the graph neural network model includes: Obtain GNSS observation data and precise inter-frequency deviation products of global sites, and calculate the phase-smoothed pseudo-range inter-frequency difference value and puncture point information for each site based on the GNSS observation data; The calculated phase-smoothed pseudorange frequency difference value and puncture point information are used as input, and the precise frequency deviation product is used as output to construct a data set; Based on the graph neural network architecture, satellites and receivers are used as nodes, puncture point information and phase-smoothed pseudorange frequency difference values ​​are used as edge attributes, and embedded transformation is performed to form a graph neural network model. Use the constructed dataset to train the graph neural network model and output the trained model.

6. A GNSS inter-frequency deviation solving device based on graph neural network, characterized in that: It includes a memory and a processor, the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the GNSS inter-frequency deviation solution method based on graph neural network as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which enable the computer to execute the steps of the GNSS inter-frequency deviation solution method based on graph neural network as described in any one of claims 1 to 4.

Images