A static coordinate solution and verification system based on GNSS measurement control network

Abstract

The present application relates to the technical field of positioning, especially relates to a static coordinate calculation and verification system based on GNSS surveying control network. The static coordinate calculation and verification system based on GNSS surveying control network comprises a static observation data acquisition module, a static coordinate calculation module and a static coordinate calculation verification module. The present application analyzes the error in the static coordinate calculation process through the static coordinate calculation strategy adjustment model, and then adjusts the static coordinate calculation process, so that the error is verified in the corresponding static coordinate calculation process of each iteration, and then adjusted, without relying on the experience of engineers, and the work of static coordinate calculation and verification can be accurately completed.

Description

Technical Field

[0001] This invention relates to the field of positioning technology, and in particular to a static coordinate calculation and verification system based on a GNSS measurement and control network. Background Technology

[0002] GNSS measurement and control networks, as a crucial infrastructure for modern high-precision spatial positioning, are widely used in basic surveying, engineering surveying, and crustal deformation monitoring. Traditional GNSS static coordinate calculation methods typically include baseline vector calculation, data quality verification, ambiguity fixing, and network adjustment. The results rely heavily on manual intervention, fixed parameter configuration, and post-implementation static evaluation. Existing systems generally suffer from the following shortcomings: First, they lack effective means to identify observation anomalies, ambiguity drift, or network error propagation. Anomalies or baselines often rely on manual experience for judgment, which is inefficient and susceptible to subjective influences. Second, they lack adaptive mechanisms in error processing, failing to dynamically adjust weights, eliminate anomalies, or isolate errors based on observation data quality. This results in insufficient stability of the calculation results, particularly prone to error accumulation and coordinate drift in large-scale, multi-device, and complex environments. Third, the calculation and verification processes are disconnected, lacking a closed-loop feedback mechanism, making intelligent diagnosis and proactive correction difficult. Summary of the Invention

[0003] (a) Technical problems to be solved

[0004] This invention aims to provide a static coordinate calculation and verification system based on GNSS measurement and control network. It realizes the analysis of errors in the static coordinate calculation process by adjusting the model through static coordinate calculation strategy, and then adjusts the static coordinate calculation process. This allows the error to be verified and adjusted in the static coordinate calculation process of each iteration, without relying on the experience of engineers, and can accurately complete the static coordinate calculation and verification work.

[0005] (II) Technical Solution

[0006] To achieve the above objectives, the present invention provides the following technical solution: a static coordinate calculation and verification system based on a GNSS measurement and control network, comprising:

[0007] The static observation data acquisition module is used to acquire static observation data of all deployed control points;

[0008] The static coordinate calculation module is used to perform static coordinate calculation on the static observation data of all deployed control points, obtain the standard coordinates corresponding to all deployed control points, and obtain the feature vectors of deployed control points and baseline edge feature vectors from the static coordinate calculation process.

[0009] The static coordinate calculation and verification module is used to assemble a control point feature map from all the feature vectors of the deployed control points and a baseline edge feature map from all the feature vectors of the baseline edges. The control point feature map and the baseline edge feature map are then fed into the static coordinate calculation strategy adjustment model for processing, outputting the static coordinate calculation adjustment strategy. Before each use, the static coordinate calculation strategy adjustment model performs reinforcement learning based on the coordinate change evaluation value. During the application of the static coordinate calculation adjustment strategy to the static coordinate calculation process, static coordinate calculation continues on the static observation data of all deployed control points, with each static coordinate calculation considered as an iteration. The static coordinate calculation and verification are completed when the coordinate change evaluation value is less than the evaluation value threshold. The coordinate change evaluation value is the standard deviation between the coordinate change rates corresponding to all deployed control points in the current iteration.

[0010] Preferably, the system further includes a method applied to the system, comprising:

[0011] Obtain static observation data for all deployed control points;

[0012] Static coordinates are calculated from the static observation data of all control points to obtain the standard coordinates of all control points. The feature vectors of the control points and the feature vectors of the baseline edges are obtained from the static coordinate calculation process.

[0013] A control point feature map is constructed from the feature vectors of all deployed control points, and a baseline edge feature map is constructed from the feature vectors of all baseline edges. These two feature maps are then fed into a static coordinate calculation strategy adjustment model for processing, outputting the static coordinate calculation adjustment strategy. Before each use, the static coordinate calculation strategy adjustment model undergoes reinforcement learning based on the coordinate change evaluation value. During the static coordinate calculation process, the static coordinate calculation continues on the static observation data of all deployed control points, with each static coordinate calculation considered an iteration. The static coordinate calculation and verification are completed when the coordinate change evaluation value is less than the evaluation threshold. This represents the standard deviation of the rate of change of coordinates of all control points in the current iteration. The rate of change of coordinates of control points is the ratio between the change value of coordinates of control points in the current iteration and the change value of coordinates of control points in the previous iteration. Specifically, it is the difference between the change value of coordinates of control points in the current iteration and the change value of coordinates of control points in the previous iteration divided by the change value of coordinates of control points in the current iteration. The change value of coordinates of control points in the current iteration is the distance between the standard coordinates of control points in the current iteration and the standard coordinates of control points in the previous iteration, which can be calculated using the Euclidean distance algorithm.

[0014] Preferably, the static coordinate solution strategy adjustment model includes a node relationship enhancement layer, an error propagation layer, and a static coordinate solution adjustment strategy output layer. The node relationship enhancement layer enhances the feature map of the control points using the baseline edge feature map to obtain an enhanced node feature map. The error propagation layer constructs a node adjacency matrix based on the baseline edge feature map and performs graph convolution on the enhanced node feature map according to the adjacency matrix to obtain a corrected node feature map. The static coordinate solution adjustment strategy output layer performs a fully connected operation on the static coordinate solution analysis vector to output the static coordinate solution adjustment strategy. The static coordinate solution analysis vector is derived from the corrected node feature map. Figure 1 Output after dimensional expansion and baseline edge features Figure 1 The output after dimensional expansion is obtained by splicing the first and last parts together.

[0015] Preferably, the node relationship enhancement layer performs feature enhancement operations on the control point feature map using the baseline edge feature map to obtain the enhanced node feature map, specifically including the following steps:

[0016] The control point feature map is multiplied by the value weight matrix and the key weight matrix to obtain the corresponding control point feature value matrix V and control point feature key matrix K. The baseline edge feature map is multiplied by the query weight matrix to obtain the baseline edge feature query matrix Q. Feature enhancement is then performed using the following formula: H = softmax(QK) T / D 0.5 ), where H is the enhanced node feature map, T is the matrix transpose operation, and D is the dimension of the control point feature key matrix K.

[0017] Preferably, the error propagation layer is used to construct a node adjacency matrix based on the baseline edge feature map, and to perform graph convolution operation on the enhanced node feature map according to the node adjacency matrix to obtain the corrected node feature map, specifically including the following steps:

[0018] A node adjacency matrix is ​​constructed based on the baseline edge feature map. The node adjacency matrix includes the connectivity between the i-th deployed control point and the i-th deployed control point, where i, j = 1, 2, 3, ..., M is the total number of control points.

[0019] The enhanced node feature map is subjected to graph convolution operation using the following formula: F(t+1) = AW(t+1)F(t), where F(t+1) is the local feature map output in the t-th iteration, A is the node adjacency matrix, and W(t+1) is the graph convolution weight matrix corresponding to the (t+1)-th iteration. When the number of iterations reaches T, the final output local feature map is used as the corrected node feature map.

[0020] Preferably, the static coordinate solution strategy adjustment model performs reinforcement learning based on the coordinate change evaluation value before each use, specifically including the following steps:

[0021] The feature maps of the control points and baseline edges are fed into the static coordinate solution strategy adjustment model for processing, and the static coordinate solution adjustment strategy is output. Then, the feature maps of the control points are... Figure 1 Output after dimensional expansion, baseline edge features Figure 1 The output after dimensional expansion and the static coordinate solution adjustment strategy are fed into the static coordinate solution adjustment strategy evaluation network for processing. The static coordinate solution adjustment strategy evaluation network is built based on the BP neural network and outputs the strategy evaluation value, which is based on the sum of the reciprocal of the strategy evaluation value and the coordinate change evaluation value.

[0022] Preferably, the model is trained to adjust the static coordinate calculation strategy, specifically including the following steps:

[0023] Several static coordinate solution strategy adjustment training samples are obtained. The static coordinate solution strategy adjustment training samples include control point feature maps and baseline edge feature maps. The static coordinate solution strategy adjustment training samples are labeled using the static coordinate solution adjustment strategy. All labeled static coordinate solution strategy adjustment training samples are combined into a static coordinate solution strategy adjustment training set. The static coordinate solution strategy adjustment model is trained using the static coordinate solution strategy adjustment training set. The training target is the labeled static coordinate solution strategy.

[0024] Preferably, training the evaluation network for the static coordinate solution adjustment strategy includes the following steps:

[0025] Obtain several training samples for evaluating static coordinate solution adjustment strategies. These training samples include features of control point deployment. Figure 1 Output after dimensional expansion, baseline edge features Figure 1 The output after dimensional expansion and the static coordinate solution adjustment strategy are used to label the static coordinate solution adjustment strategy evaluation training samples through the strategy evaluation value. All labeled static coordinate solution adjustment strategy evaluation training samples are combined into a static coordinate solution adjustment strategy evaluation training set. The static coordinate solution adjustment strategy evaluation model is trained through the static coordinate solution adjustment strategy evaluation training set, and the training target is the labeled strategy evaluation value.

[0026] Compared with the prior art, the present invention provides a static coordinate calculation and verification system based on GNSS measurement and control network, which has the following advantages:

[0027] This application analyzes the errors in the static coordinate calculation process by adjusting the model through a static coordinate solution strategy, and then adjusts the static coordinate calculation process so that the error is verified and adjusted in each iteration of the static coordinate calculation process. This eliminates the need to rely on engineer experience and can accurately complete the static coordinate solution and verification work. Attached Figure Description

[0028] Figure 1 This is a system structure diagram of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Example 1: A static coordinate calculation and verification system based on a GNSS measurement and control network, see [link to example]. Figure 1 As shown, it includes:

[0031] The static observation data acquisition module is used to acquire static observation data of all deployed control points. The static observation data includes the observation value sequence and ephemeris data of each satellite by the GNSS receiver. The observation value sequence includes pseudorange, carrier phase and signal-to-noise ratio, etc., and the ephemeris data includes the orbital parameters and satellite clock correction parameters of each satellite. The static observation data is stored in RINEX data format to provide a data foundation for subsequent static coordinate calculation.

[0032] The static coordinate calculation module is used to perform static coordinate calculations on the static observation data of all control points, obtaining the standard coordinates corresponding to all control points. The standard coordinates are the accurate coordinates of each control point after calculation. The module also obtains the feature vectors of the control points and the baseline edge feature vectors during the static coordinate calculation process. The feature vectors include the calculated coordinates of the control point set, the standard error, and the residual vector. The standard error is the square root of the diagonal element of the covariance matrix corresponding to the control point during the static coordinate calculation process, and the residual vector is obtained from the static observation data. The error between the observed coordinates and the calculated coordinates of the control point set is controlled. The baseline edge feature vector includes the components of the calculated baseline vector in different directions (X-axis, Y-axis, and Z-axis), the length of the baseline edge, the azimuth of the baseline edge, the baseline residual, and the baseline covariance attribute. The baseline edge refers to the connecting edge between two control points, the baseline vector is the vector corresponding to the baseline edge, the baseline residual is the residual between the observed value of the baseline vector before the calculation and the calculated value after the calculation, and the baseline covariance attribute is the covariance between the observed value of the baseline vector before the calculation and the calculated value after the calculation.

[0033] The static coordinate calculation and verification module is used to assemble a control point feature map from the feature vectors of all deployed control points and a baseline edge feature map from the feature vectors of all baseline edges. These feature maps are then fed into the static coordinate calculation strategy adjustment model for processing, outputting a static coordinate calculation adjustment strategy. This strategy consists of a set of weight values ​​for the observation matrix. Before each use, the static coordinate calculation strategy adjustment model performs reinforcement learning based on the coordinate change evaluation value. During the static coordinate calculation process, the static coordinate calculation continues on the static observation data of all deployed control points, with each static coordinate calculation considered an iteration. The process continues until the coordinate change evaluation value is less than the evaluation threshold, at which point the process is complete. The static coordinates are calculated and verified. The coordinate change evaluation value is the standard deviation between the coordinate change rates of all control points in the current iteration. The coordinate change rate of the control point is the ratio between the coordinate change value of the control point in the current iteration and the coordinate change value of the control point in the previous iteration. Specifically, it is the difference between the coordinate change value of the control point in the current iteration and the coordinate change value of the control point in the previous iteration divided by the coordinate change value of the control point in the current iteration. The coordinate change value of the control point in the current iteration is the distance between the standard coordinates of the control point in the current iteration and the standard coordinates of the control point in the previous iteration, which can be calculated using the Euclidean distance algorithm.

[0034] Example 2: A method for static coordinate calculation and verification system based on GNSS measurement and control network, comprising:

[0035] Acquire static observation data for all deployed control points. The static observation data includes the observation sequence and ephemeris data of each satellite from the GNSS receiver. The observation sequence includes pseudorange, carrier phase and signal-to-noise ratio, etc., and the ephemeris data includes the orbital parameters and satellite clock correction parameters of each satellite. The static observation data is stored in RINEX data format to provide a data foundation for subsequent static coordinate calculation.

[0036] It should be noted that during the static coordinate calculation and verification process, several control points will be set up in the corresponding area, and a GNSS receiver will be installed at each control point. Then, the observation mode of the GNSS receiver will be adjusted to static, the observation interval will be set to 5 seconds, the sampling duration will be set to one hour, and static observation data will be continuously acquired.

[0037] Static coordinate calculations are performed on the static observation data of all control points to obtain the standard coordinates corresponding to each control point. The standard coordinates are the accurate coordinates of each control point after calculation. Feature vectors for the control points and baseline edges are obtained from the static coordinate calculation process. The feature vectors include the calculated coordinates of the control point set, the standard error, and the residual vector. The standard error is the square root of the diagonal element of the covariance matrix corresponding to the control point during the static coordinate calculation, and the residual vector is the control observation coordinates obtained from the static observation data. The error between the coordinates of the control point set and the baseline edge feature vector includes the components of the calculated baseline vector in different directions (X-axis, Y-axis and Z-axis), the length of the baseline edge, the azimuth of the baseline edge, the baseline residual and the baseline covariance attribute, etc., where the baseline edge refers to the connecting edge between two control points, the baseline vector is the vector corresponding to the baseline edge, the baseline residual is the residual between the observed value of the baseline vector before the calculation and the calculated value after the calculation, and the baseline covariance attribute is the covariance between the observed value of the baseline vector before the calculation and the calculated value after the calculation.

[0038] The specific steps for static coordinate calculation are as follows: For any two deployment control points A and B, calculate the carrier phase difference for the same satellite at the same time. Then, eliminate the clock difference between the GNSS receiver and the satellite using the double-difference formula to obtain the pure geometric difference. Next, solve the pure geometric difference using the LAMBDA method to obtain the ambiguity N, thus obtaining the accurate distance difference between deployment control points A and B. Finally, calculate the distance difference according to the formula X = (A... T PA) -1 A TP(Y-λN), where X is the baseline vector between control points A and B after settlement, A is the projection of the satellite direction vector, P is the observation matrix, which includes the weight value of each control point, reflecting which observations are more reliable (set by the configuration file), Y is the pure geometric difference, and λ is the wavelength. The control point coordinates are then calculated by vector summation between the baseline vector and the coordinates of known points (e.g., national control points). When a control point is associated with multiple baseline vectors, the calculated control point coordinates are the average of the calculated baseline vectors. Finally, adjustment is performed on all calculated baseline vectors and control point coordinates to complete the static coordinate calculation. Furthermore, the baseline vector can only be constructed when the same satellite is observed at two control points.

[0039] All feature vectors of the deployed control points are combined to form a feature map of the deployed control points, and all feature vectors of the baseline edges are combined to form a feature map of the baseline edges. These feature maps are then fed into a static coordinate calculation strategy adjustment model for processing, outputting a static coordinate calculation adjustment strategy. This strategy consists of a set of weight values ​​for the observation matrix. Before each use, the static coordinate calculation strategy adjustment model undergoes reinforcement learning based on the coordinate change evaluation value. During the static coordinate calculation process, the static coordinate calculation continues on the static observation data of all deployed control points, with each static coordinate calculation considered an iteration. This continues until the coordinate change evaluation value is less than an evaluation value threshold, which is set by the operator. The static coordinates are calculated and verified. The coordinate change evaluation value is the standard deviation between the coordinate change rates of all control points in the current iteration. The coordinate change rate of the control point is the ratio between the coordinate change value of the control point in the current iteration and the coordinate change value of the control point in the previous iteration. Specifically, it is the difference between the coordinate change value of the control point in the current iteration and the coordinate change value of the control point in the previous iteration divided by the coordinate change value of the control point in the current iteration. The coordinate change value of the control point in the current iteration is the distance between the standard coordinates of the control point in the current iteration and the standard coordinates of the control point in the previous iteration, which can be calculated using the Euclidean distance algorithm.

[0040] This application analyzes the errors in the static coordinate calculation process by adjusting the model through a static coordinate solution strategy, and then adjusts the static coordinate calculation process so that the error is verified and adjusted in each iteration of the static coordinate calculation process. This eliminates the need to rely on engineer experience and can accurately complete the static coordinate solution and verification work.

[0041] The static coordinate solution strategy adjustment model includes a node relationship enhancement layer, an error propagation layer, and a static coordinate solution adjustment strategy output layer. The node relationship enhancement layer enhances the feature map of the deployed control points using the baseline edge feature map, resulting in an enhanced node feature map. Since the baseline edge feature map reflects the coordinate relationship between any two deployed control points, it can highlight deployed control points with errors. For example, if the relationship between the calculated coordinates of two deployed control points does not match the baseline variables corresponding to the two deployed control points, then these two deployed control points have a higher probability of error. The error propagation layer constructs a node adjacency matrix based on the baseline edge feature map and performs graph convolution on the enhanced node feature map according to the node adjacency matrix to propagate the error, resulting in a corrected node feature map. Since the coordinates of the deployed control points are calculated based on multiple baseline vectors, when errors occur in the baseline variables, the errors will propagate along with the calculation of the baseline vectors. The static coordinate solution adjustment strategy output layer performs a fully connected operation on the static coordinate solution analysis vectors and outputs the static coordinate solution adjustment strategy. The static coordinate solution analysis vectors are derived from the corrected node feature map. Figure 1 Output after dimensional expansion and baseline edge features Figure 1 The output after dimensional expansion is obtained by splicing the first and last parts together.

[0042] In the static coordinate calculation process, the errors mainly come from two categories: GNSS observation errors and model errors. Observation errors include satellite clock errors, receiver clock errors, ionospheric delay, tropospheric delay, multipath effects, carrier phase observation noise, and antenna phase center deviation. Model errors mainly come from unfixed ambiguity, inaccurate baseline covariance estimation, imperfect satellite orbit and clock error models, and unreasonable weight matrix settings during the adjustment process.

[0043] The node relationship enhancement layer performs feature enhancement operations on the control point feature map using the baseline edge feature map to obtain the enhanced node feature map. This process includes the following steps:

[0044] The control point feature map is multiplied by the value weight matrix and the key weight matrix to obtain the corresponding control point feature value matrix V and control point feature key matrix K. The baseline edge feature map is multiplied by the query weight matrix to obtain the baseline edge feature query matrix Q. Feature enhancement is then performed using the following formula: H = softmax(QK) T / D 0.5 ), where H is the enhanced node feature map, T is the matrix transpose operation, and D is the dimension of the control point feature key matrix K; it should be noted that the feature enhancement operation is set with reference to the self-attention mechanism operation in the Transformer model, and the value weight matrix, key weight matrix and query weight matrix are adjusted according to the static coordinate solution strategy to adjust the training of the model.

[0045] The error propagation layer is used to construct a node adjacency matrix based on the baseline edge feature map, and then perform graph convolution operation on the enhanced node feature map according to the node adjacency matrix to obtain the corrected node feature map. Specifically, it includes the following steps:

[0046] A node adjacency matrix is ​​constructed based on the baseline edge feature map. The node adjacency matrix includes the connectivity between the i-th deployed control point and the i-th deployed control point. If there are several baseline vectors between two deployed control points, then the connectivity is that number. If there are no baseline vectors between two deployed control points, then the connectivity is zero. i, j=1, 2, 3, ..., M is the total number of control points.

[0047] The enhanced node feature map is subjected to graph convolution operation using the following formula: F(t+1) = AW(t+1)F(t), where F(t+1) is the local feature map output in the t-th iteration, A is the node adjacency matrix, and W(t+1) is the graph convolution weight matrix corresponding to the (t+1)-th iteration. This process continues until the number of iterations reaches T, where T is typically 3. The final output local feature map is then used as the corrected node feature map.

[0048] The static coordinate solution strategy adjustment model performs reinforcement learning based on the coordinate change evaluation value before each use, specifically including the following steps:

[0049] The feature maps of the control points and baseline edges are fed into the static coordinate solution strategy adjustment model for processing, and the static coordinate solution adjustment strategy is output. Then, the feature maps of the control points are... Figure 1 Output after dimensional expansion, baseline edge features Figure 1 The output after dimensional expansion and the static coordinate solution adjustment strategy are fed into the static coordinate solution adjustment strategy evaluation network for processing. The static coordinate solution adjustment strategy evaluation network is built based on the BP neural network and outputs the strategy evaluation value, which is based on the sum of the reciprocal of the strategy evaluation value and the coordinate change evaluation value.

[0050] The model is trained to adjust the static coordinate calculation strategy, which includes the following steps:

[0051] Several training samples for adjusting the static coordinate calculation strategy are obtained. These training samples include control point feature maps and baseline edge feature maps, which are obtained by operators based on actual static coordinate calculation experiments. The static coordinate calculation strategy is labeled using the optimal adjustment strategy obtained by operators based on engineers' experience. All labeled training samples are combined into a static coordinate calculation strategy adjustment training set. The static coordinate calculation strategy adjustment model is trained using this training set, with the training objective being the labeled static coordinate calculation strategy. The training condition is determined by whether the training conditions are met. Generally, the accuracy of the static coordinate calculation strategy adjustment model meets expectations. If the training conditions are met, the trained static coordinate calculation strategy adjustment model is output; otherwise, the static coordinate calculation strategy adjustment model is trained again using the static coordinate calculation strategy adjustment training set.

[0052] The training of the evaluation network for the static coordinate solution adjustment strategy includes the following steps:

[0053] Obtain several training samples for evaluating static coordinate solution adjustment strategies. These training samples include features of control point deployment. Figure 1 Output after dimensional expansion, baseline edge features Figure 1 The output after dimensional expansion and the static coordinate calculation adjustment strategy are as follows: the control point feature map and baseline edge feature map are obtained by the operator based on actual static coordinate calculation experiments, and the corresponding static coordinate calculation adjustment strategy is the optimal adjustment strategy obtained by the operator based on the engineer's experience. The static coordinate calculation adjustment strategy evaluation training samples are labeled with the strategy evaluation value, which is generally 1. All labeled static coordinate calculation adjustment strategy evaluation training samples are combined into a static coordinate calculation adjustment strategy evaluation training set. The static coordinate calculation adjustment strategy evaluation model is trained using the static coordinate calculation adjustment strategy evaluation training set. The training objective is the labeled strategy evaluation value. It is then determined whether the training conditions are met. The training conditions are generally that the accuracy of the static coordinate calculation adjustment strategy evaluation model meets expectations. If the training conditions are met, the trained static coordinate calculation adjustment strategy evaluation model is output; otherwise, the static coordinate calculation adjustment strategy evaluation model is trained again using the static coordinate calculation adjustment strategy evaluation training set.

[0054] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A static coordinate calculation and verification system based on a GNSS measurement and control network, characterized in that, include: The static observation data acquisition module is used to acquire static observation data of all deployed control points; The static coordinate calculation module is used to perform static coordinate calculation on the static observation data of all deployed control points. The static coordinate calculation adopts double-difference carrier phase static calculation to obtain the standard coordinates corresponding to all deployed control points, and obtains the feature vectors of deployed control points and baseline edge feature vectors from the static coordinate calculation process. The static coordinate calculation and verification module is used to assemble a control point feature map from the feature vectors of all deployed control points and a baseline edge feature map from the feature vectors of all baseline edges. These feature maps are then fed into the static coordinate calculation strategy adjustment model for processing, outputting a static coordinate calculation adjustment strategy. This strategy consists of a set of weight values ​​for the observation matrix in the static coordinate calculation. Before each use, the static coordinate calculation strategy adjustment model performs reinforcement learning based on the coordinate change evaluation value. During the application of the static coordinate calculation adjustment strategy to the static coordinate calculation process, static coordinate calculation continues on the static observation data of all deployed control points, with each static coordinate calculation considered an iteration. The static coordinate calculation and verification are completed when the coordinate change evaluation value is less than the evaluation value threshold. The coordinate change evaluation value is the standard deviation between the coordinate change rates corresponding to all deployed control points in the current iteration. It also includes a method applied to the system, comprising: Obtain static observation data for all deployed control points; Static coordinates are calculated from the static observation data of all control points. The static coordinate calculation adopts double-difference carrier phase static calculation to obtain the standard coordinates corresponding to all control points. The feature vectors of the control points and the baseline edge feature vectors are obtained from the static coordinate calculation process. A control point feature map is constructed from the feature vectors of all deployed control points, and a baseline edge feature map is constructed from the feature vectors of all baseline edges. These two feature maps are then fed into a static coordinate calculation strategy adjustment model for processing, outputting a static coordinate calculation adjustment strategy. This strategy consists of a set of weight values ​​for the observation matrix used in the static coordinate calculation. Before each use, the static coordinate calculation strategy adjustment model undergoes reinforcement learning based on the coordinate change evaluation value. During the application of the static coordinate calculation adjustment strategy to the static coordinate calculation process, static coordinate calculations are continuously performed on the static observation data of all deployed control points, with each static coordinate calculation considered as one iteration. This process continues until the coordinate change evaluation value is less than the evaluation value threshold, at which point the static coordinate calculation is complete. The coordinate calculation and verification are as follows: the coordinate change evaluation value is the standard deviation between the coordinate change rates of all control points in the current iteration; the coordinate change rate of a control point is the ratio between the coordinate change value of the control point in the current iteration and the coordinate change value of the control point in the previous iteration, specifically the difference between the coordinate change value of the control point in the current iteration and the coordinate change value of the control point in the previous iteration divided by the coordinate change value of the control point in the current iteration; the coordinate change value of the control point in the current iteration is the distance between the standard coordinates of the control point in the current iteration and the standard coordinates of the control point in the previous iteration, which can be calculated using the Euclidean distance algorithm.

2. The static coordinate calculation and verification system based on a GNSS measurement and control network according to claim 1, characterized in that, The static coordinate solution strategy adjustment model includes a node relationship enhancement layer, an error propagation layer, and a static coordinate solution adjustment strategy output layer. The node relationship enhancement layer is used to enhance the feature map of the control points by using the baseline edge feature map to obtain the enhanced node feature map. The error propagation layer is used to construct a node adjacency matrix based on the baseline edge feature map and perform graph convolution operation on the enhanced node feature map according to the node adjacency matrix to obtain the corrected node feature map. The static coordinate solution adjustment strategy output layer is used to perform a fully connected operation on the static coordinate solution analysis vector to output the static coordinate solution adjustment strategy. The static coordinate solution analysis vector is obtained by concatenating the output of the one-dimensional expansion of the corrected node feature map and the output of the one-dimensional expansion of the baseline edge feature map.

3. The static coordinate calculation and verification system based on a GNSS measurement and control network according to claim 2, characterized in that, The node relationship enhancement layer performs feature enhancement operations on the control point feature map using the baseline edge feature map to obtain the enhanced node feature map. This process includes the following steps: The control point feature map is multiplied by the value weight matrix and the key weight matrix to obtain the corresponding control point feature value matrix V and control point feature key matrix K. The baseline edge feature map is multiplied by the query weight matrix to obtain the baseline edge feature query matrix Q. Feature enhancement is then performed using the following formula: H = softmax(QK) T / D 0.5 ), where H is the enhanced node feature map, T is the matrix transpose operation, and D is the dimension of the control point feature key matrix K.

4. The static coordinate calculation and verification system based on a GNSS measurement and control network according to claim 3, characterized in that, The error propagation layer is used to construct a node adjacency matrix based on the baseline edge feature map, and then perform graph convolution operation on the enhanced node feature map according to the node adjacency matrix to obtain the corrected node feature map. Specifically, it includes the following steps: A node adjacency matrix is ​​constructed based on the baseline edge feature map. The node adjacency matrix includes the connectivity between the i-th deployed control point and the i-th deployed control point, where i, j = 1, 2, 3, ..., M is the total number of control points. The enhanced node feature map is subjected to graph convolution operation using the following formula: F(t+1) = AW(t+1)F(t), where F(t+1) is the local feature map output in the t-th iteration, A is the node adjacency matrix, and W(t+1) is the graph convolution weight matrix corresponding to the (t+1)-th iteration. When the number of iterations reaches T, the final output local feature map is used as the corrected node feature map.

5. A static coordinate calculation and verification system based on a GNSS measurement and control network according to claim 4, characterized in that, The static coordinate solution strategy adjustment model performs reinforcement learning based on the coordinate change evaluation value before each use, specifically including the following steps: The control point feature map and baseline edge feature map are fed into the static coordinate solution strategy adjustment model for processing, and the static coordinate solution adjustment strategy is output. Then, the output of the one-dimensional expansion of the control point feature map, the output of the one-dimensional expansion of the baseline edge feature map, and the static coordinate solution adjustment strategy are fed into the static coordinate solution adjustment strategy evaluation network for processing. The static coordinate solution adjustment strategy evaluation network is built based on the BP neural network and outputs the strategy evaluation value, which is based on the sum of the reciprocal of the strategy evaluation value and the coordinate change evaluation value.

6. The static coordinate calculation and verification system based on a GNSS measurement and control network according to claim 5, characterized in that, The model is trained to adjust the static coordinate calculation strategy, which includes the following steps: Several static coordinate solution strategy adjustment training samples are obtained. The static coordinate solution strategy adjustment training samples include control point feature maps and baseline edge feature maps. The static coordinate solution strategy adjustment training samples are labeled using the static coordinate solution adjustment strategy. All labeled static coordinate solution strategy adjustment training samples are combined into a static coordinate solution strategy adjustment training set. The static coordinate solution strategy adjustment model is trained using the static coordinate solution strategy adjustment training set. The training target is the labeled static coordinate solution strategy.

7. A static coordinate calculation and verification system based on a GNSS measurement and control network according to claim 6, characterized in that, The training of the evaluation network for the static coordinate solution adjustment strategy includes the following steps: Several training samples for evaluating static coordinate solution adjustment strategies are obtained. These training samples include the output of the one-dimensional unfolded control point feature map, the output of the one-dimensional unfolded baseline edge feature map, and the static coordinate solution adjustment strategy. The training samples are labeled with the strategy evaluation values. All labeled training samples are combined into a static coordinate solution adjustment strategy evaluation training set. The static coordinate solution adjustment strategy evaluation model is trained using this training set, with the labeled strategy evaluation values ​​as the training objective.

Images