A track state domain parameter continuity guarantee method and system

Abstract

The application discloses a kind of track state domain parameter continuity guarantee method and system, belong to satellite navigation and positioning technical field.The method includes: receiving global reference station data and broadcast ephemeris data, calculating satellite real-time orbit correction value;Real-time orbit correction value is predicted based on reinforcement learning model, and it is compared with third party precise orbit product, and difference data is used to train optimization model;When real-time calculation is interrupted, the predicted value is used to encode broadcast, to guarantee parameter continuity.The application can automatically learn the change mode of orbit correction number from data, adapt to the prediction needs under different satellites and environmental conditions, improve the timeliness and accuracy of prediction.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of satellite navigation and positioning, and particularly relates to an orbit state domain parameter continuity guarantee method and system. BACKGROUND

[0002] PPP-RTK (Precise Point Positioning-Real Time Kinematic) is a high-precision positioning technology based on state domain (SSR) parameters. Through comprehensive processing of global framework network and backbone network reference station data, state correction values including satellite orbits, clock errors, ionospheric layers, etc. are generated and sent to the user end for real-time high-precision position calculation, supporting real-time high-precision positioning needs of industries such as automatic driving, unmanned farms, and ocean ranches. High-continuity and high-reliability satellite orbit products are important conditions for guaranteeing real-time continuity of navigation and positioning. In particular, high-reliability safety in unmanned and automatic driving fields has higher requirements for the continuity and stability of satellite orbit state domain parameters.

[0003] Generally, satellite orbit state parameters are directly sent to the encoding broadcast system for broadcast after being calculated in the cloud. If the reference station data or the transmission network is interrupted, the user cannot use the service during the fault period, resulting in positioning failure and causing related serious accidents. Traditional continuity monitoring of satellite orbit state domain parameters is usually based on monitoring after parameter broadcast. Interruptions can only provide information, and cannot provide a high-precision data backup broadcast for users in use, often causing interruptions in real-time positioning of users and affecting experience. Therefore, orbit state domain parameter continuity guarantee method and broadcast system are needed to guarantee. SUMMARY

[0004] To solve the above technical problems, the application provides an orbit state domain parameter continuity guarantee method and system, which can automatically learn the change pattern of orbit correction numbers from data, adapt to the prediction needs of different satellites and environmental conditions, and effectively cope with rapidly changing orbit environments through real-time model training, improving the timeliness and accuracy of prediction.

[0005] To achieve the above purpose, the technical scheme adopted by the application is as follows:

[0006] An orbit state domain parameter continuity guarantee method, the method comprising:

[0007] Step 1, receiving global reference station data and broadcast ephemeris data, and calculating satellite real-time orbit correction values;

[0008] Step 2, predicting the real-time orbit correction value based on a reinforcement learning model, comparing the predicted orbit correction value with a precise orbit product provided by a third-party service center, and continuously training and optimizing the reinforcement learning model using the obtained difference value data.

[0009] Step 3, when the real-time orbit correction value calculation is interrupted, the predicted real-time orbit correction value is used for encoding and broadcasting to ensure the continuity of the orbit state domain parameters.

[0010] In another aspect, the present application provides an orbit state domain parameter continuity guarantee system, comprising:

[0011] A real-time orbit correction state domain calculation end is configured to receive global reference station data and broadcast ephemeris data, and calculate a satellite real-time orbit correction value.

[0012] A continuity guarantee service end is configured to predict the real-time orbit correction value based on a reinforcement learning model, compare the predicted orbit correction value with a precise orbit product provided by a third-party service center, and continuously train and optimize the reinforcement learning model using the obtained difference value data.

[0013] An encoding and broadcasting end is configured to use the predicted real-time orbit correction value for encoding and broadcasting when the real-time orbit correction value calculation is interrupted, so as to ensure the continuity of the orbit state domain parameters.

[0014] In a third aspect, the present application provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned orbit state domain parameter continuity guarantee method.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium having stored executable instructions, which, when executed by a processor, can enable the processor to implement the aforementioned orbit state domain parameter continuity guarantee method.

[0016] The present application has the following beneficial effects:

[0017] The continuity guarantee service end predicts the orbit correction number through an artificial intelligence reinforcement learning model, and corrects the learning model through comparison with other service agencies to improve the accuracy and availability of the predicted orbit correction number; in comparison with other service agency products, the predicted orbit correction number is provided in the event of data interruption, which on one hand guarantees the continuity of real-time positioning of the user end, and on the other hand avoids the principle that the public products of other service agencies are not commercially available. The continuity guarantee end is added between the calculation and broadcast of the traditional orbit correction product to realize the continuity of the broadcast of the satellite orbit state domain parameters. When the real-time orbit correction state domain calculation end is interrupted due to network and data reasons, the predicted correction number of the continuity guarantee service end is encoded and broadcast in a timely manner.

[0018] In addition, the present application can automatically learn the change mode of the orbit correction number from data, adapt to the prediction requirements under different satellites and environmental conditions, and through real-time updating of the model training, the method can effectively respond to the rapidly changing orbit environment, improve the timeliness and accuracy of the prediction, and has wide application potential. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A flow chart of the orbit state domain parameter continuity guarantee method of the present application;

[0020] Figure 2 A block diagram of the orbit state domain parameter continuity guarantee system of the present application;

[0021] Figure 3 A neural network structure diagram selected by the present application. DETAILED DESCRIPTION

[0022] The present application will be further described below in combination with the drawings and examples.

[0023] As shown in the drawings, Figure 1 The present application provides an orbit state domain parameter continuity guarantee method, which comprises:

[0024] Step S01: receiving global reference station data and broadcast ephemeris data, and calculating satellite real-time orbit correction value; in the real-time orbit correction state domain calculation end, the global reference station real-time observation data and broadcast ephemeris data provided by the service center such as IGS are used to perform satellite real-time precise orbit determination. The center of mass coordinates of the Beidou navigation satellite in the earth-fixed system are obtained through real-time orbit determination, and the coordinates of the satellite in the inertial system are obtained through coordinate conversion; the satellite position in the inertial system after antenna phase center correction is obtained through antenna correction; the satellite position in the corrected earth-fixed system is further calculated; the real-time orbit correction value is obtained by using the satellite position in the broadcast ephemeris and the satellite position in the corrected earth-fixed system. The specific calculation process is as follows:

[0025] The three components of the satellite antenna phase center bias are in the satellite-fixed coordinate system, and the broadcast orbit correction numbers are in the orbital coordinate system. The conversion between the satellite-fixed coordinate system, the earth-centered earth-fixed coordinate system (hereinafter referred to as the earth-fixed system), the earth-centered inertial system (hereinafter referred to as the inertial system) and the orbital coordinate system is needed to correct the antenna phase center to the satellite center of mass position.

[0026] The center of mass coordinates of the Beidou navigation satellite in the earth-fixed coordinate system are calculated by using the real-time observation data and the broadcast ephemeris, and the coordinates of the satellite in the inertial system are obtained by the following formula:

[0027] ,

[0028] Among them, ( , , ) is the inertial coordinate of the satellite corresponding to the reference epoch, such as the Beidou satellite, which represents the inertial coordinate corresponding to the epoch J2000; ( , , ) is the coordinate of the satellite in the earth-fixed coordinate system, represents the matrix of the precession, nutation, earth rotation and polar motion at time t, represents the auxiliary matrix.

[0029] The unit vectors of the coordinate axes of the satellite-fixed coordinate system in the inertial system can be expressed as:

[0030] ,

[0031] Among them, is the inertial vector of the satellite relative to the earth center calculated; is the unit vector from the sun to the satellite, , , respectively represent the unit vectors of the X direction, Y direction and Z direction of the coordinate axes of the satellite-fixed coordinate system in the inertial system. The satellite antenna phase center bias in the inertial system is:

[0032] ,

[0033] Among them, ( , , ) is the antenna phase center correction value in the satellite-fixed system, ( , , ) is the antenna phase center correction value in the inertial system. The position of the satellite in the inertial system after the antenna phase center correction can be expressed as:

[0034] ,

[0035] in,( , , The corrected satellite position in the inertial frame is denoted as: (The corrected satellite position in the Earth-fixed frame can be identified as follows:)

[0036] ,

[0037] The superscript -1 indicates inversion. According to the above formula, the real-time orbital correction value of the Earth-Fixed System satellite can be expressed as:

[0038] ,

[0039] in,( , , ) represents the satellite position correction based on the phase center in the Earth-fixed system. , , ( ) represents the satellite position in the broadcast ephemeris. Using the satellite positions in the Earth-Fixed System (EFS) to participate in the above equation, we obtain the satellite position correction values ​​based on the centroid in the EFS. The calculation method is as follows:

[0040] ,

[0041] in,( , , The values ​​are the satellite orbit corrections in the orbit coordinate system, which are the satellite radial, tangential, and normal directions, respectively. These are the state domain parameters that are ultimately broadcast to the user. The velocity vector of the satellite in the Earth-fixed system; This represents the satellite position vector in the Earth-fixed system. , , These represent the unit vectors in the radial, tangential, and normal directions of the satellite, respectively.

[0042] The calculated real-time orbit corrections are directly encoded in traditional broadcasting methods and then broadcast via the internet or satellite-based communication. However, calculating these corrections requires receiving real-time data from global reference stations and broadcast ephemeris data. Any interruption in data transmission or discontinuous observations can cause interruptions in the calculated real-time orbit corrections, directly resulting in a loss of continuous broadcast information. Therefore, the next step is required.

[0043] Step S02: Predict the real-time track correction value based on the reinforcement learning model, compare the predicted track correction value with the precision track product provided by the third-party service center, and use the obtained difference value data to continuously train and optimize the reinforcement learning model;

[0044] On the continuity assurance service side, the real-time track correction value calculated in step S01 is acquired in real time. Based on reinforcement learning, the track correction value is used to make reinforcement predictions, and compared with precise track products provided by third-party service centers such as IGS and CAS. The difference value is used as a further reward / penalty parameter for the reinforcement learning model. Through a certain number of reinforcements, the model can make more accurate track correction predictions. When the real-time track correction value generated in step S01 is transmitted abnormally, the predicted track correction value generated in step S02 is sent to the encoding and broadcasting end to ensure the continuity and accuracy of the service. Specifically, it can be described as follows:

[0045] In practice, several reinforcement learning models can be selected. This implementation specifically uses a neural network for GNSS satellite orbit prediction, taking a multilayer perceptron (MLP) model as an example. This model is suitable for handling nonlinear problems and can capture complex relationships, enabling accurate prediction of GNSS satellite orbit corrections. This method utilizes a large amount of historical data and improves prediction accuracy by learning complex patterns in the data. The neural network structure is as follows... Figure 3 As shown. The specific training process is as follows:

[0046] Step S021, Learning Dataset Access and Preprocessing: To ensure continuity, the server receives real-time track correction information and constructs a track learning dataset. ,in The feature vectors are selected based on the orbital parameter update time and the satellite PRN number. The three-way correction values ​​for the radial, tangential, and normal directions of the track are calculated in step S01 and provided by other third-party service centers such as IGS and GFZ that are connected in real time.

[0047] For eigenvectors Standardization is performed so that the mean of each feature vector is 0 and the standard deviation is 1. The standardization formula is:

[0048] ,

[0049] in, The characteristic average value, The average variance of the characteristic.

[0050] Step S022: Construct a Multilayer Perceptron (MLP) model: A basic MLP model includes an input layer, several hidden layers, and an output layer. Input Layer: Determined by the dimension of the feature vector. In this application, if the update time is converted into a feature, plus the uniformly sorted and encoded satellite PRN number (e.g., GPS has 32 different PRN numbers), then the size of the GPS satellite system's input layer will be 1 + 32 = 33. Hidden Layers: Contains at least one hidden layer. Each hidden layer uses an activation function f. Output Layer: The size of the output layer depends on the variables... To match the dimensions, it is necessary to predict the corrections in three directions of the orbit. The dimension is set to 3, but the dimension can be increased according to the required rate of change.

[0051] Output of each layer The following formula can be used to calculate:

[0052] ,

[0053] in, and Let f be the weights and biases of the Lth layer, and f be the activation function (such as the ReLU function, f(z)=max(0,z)).

[0054] Output of the output layer The predicted orbital correction is calculated as follows:

[0055] ,

[0056] Step S023: Construct the loss function and optimizer: Use the mean squared error (MSE) as the loss function to evaluate the model's predicted values. Compared with the actual three-party corrections The differences between them.

[0057] ,

[0058] In the formula, n is the predicted value of the evaluation model. The number of parameters, specifically, the parameters mentioned in this aspect are mainly the corrections for the radial, tangential, and normal directions of the track, with n taking the value of 3. This is expressed as the difference between the predicted radial, tangential, and normal orbital corrections and the actual radial, tangential, and normal orbital corrections, expressed using... , , Therefore, the final calculation result of MSE is equivalent to the specific calculation method of the following formula:

[0059] ,

[0060] Use an optimizer such as Adam to minimize the loss function and update the weights W and biases b.

[0061] Step S024, Model Training and Evaluation: Through multiple iterations, the model parameters are continuously updated using optimization algorithms to minimize the loss function on the dataset. The RMS error or other relevant metrics are used to evaluate the model's performance on the test set to ensure the accuracy and reliability of the predictions.

[0062] During normal operation of the real-time orbit correction state domain calculation terminal, real-time orbit correction values ​​can be continuously provided for model training. The model can also be compared with precise orbit products provided by third-party service providers to correct its performance. Typically, navigation satellite orbits in space have relative stability. A model is considered stable when the radial error is better than 10cm or the 3D direction error is better than 15cm. The predicted orbit correction values ​​output by this model can then meet users' real-time, high-precision positioning needs.

[0063] Step S03: When the real-time track correction calculation is interrupted, the predicted real-time track correction value is used for encoding and broadcasting to ensure the continuity of the track state domain parameters. To ensure continuity, the server parses the track correction information provided by the real-time track correction state domain calculation terminal to obtain time-series state information. If the difference between the current and subsequent time states exceeds a certain threshold, such as 60 seconds, the system determines that the real-time track correction state domain calculation terminal has experienced an interruption anomaly and promptly sends the track correction value predicted using the model training to the encoding and broadcasting terminal. If the state is normal, the track correction information provided by the real-time track correction state domain calculation terminal is sent to the encoding and broadcasting terminal.

[0064] The encoding and broadcasting end primarily follows the RTCM-SSR format, encoding orbit corrections into SSR (State Domain) correction messages for broadcast. The encoding format conforms to the RTCM standard, and SSR messages typically include a header and a body. Specific data encoding is shown in Tables 1 and 2 below. Table 1 shows the GNSS satellite orbit SSR message header CGM01, and Table 2 shows the GNSS satellite orbit SSR satellite data body content. The SSR version number informs the user whether the currently received orbit product was actually generated or generated based on a learning model prediction under interruption conditions. For example, the number 1 indicates actually generated data, and the number 2 indicates data predicted by a learning model. Users can decide whether to continue using the data based on the SSR version number.

[0065] Table 1

[0066]

[0067] Table 2

[0068]

[0069] On the other hand, such asFigure 2 As shown, the present invention provides a track state domain parameter continuity assurance system, which includes various parts that can implement the various steps of the aforementioned method. Specifically, it includes a real-time track correction state domain calculation terminal P01, a continuity assurance service terminal P02, and an encoding and broadcasting terminal P03.

[0070] The real-time orbit correction state domain calculation terminal P01 mainly calculates the real-time orbit correction of the satellite by receiving data from global reference stations and broadcast ephemeris data, and obtains the correction values ​​and rates of change in the radial, tangential and normal directions of the orbit.

[0071] The continuity assurance service P02 receives real-time satellite orbit correction data, utilizes the characteristic that satellite orbits are usually smooth and unchanging, and combines an artificial intelligence reinforcement learning model to predict satellite orbit correction parameters. By comparing these parameters with orbit products provided by service centers such as IGS, GFZ, and CAS, the learning model is optimized and the parameter accuracy is improved.

[0072] The encoding and broadcasting terminal P03 is primarily responsible for promptly encoding and broadcasting the predicted correction values ​​from the continuity assurance server when the real-time orbit correction state domain calculation is interrupted due to network or data issues. The real-time satellite orbit correction values ​​are encoded according to the internationally standardized RTCM-SSR data format and broadcast via the internet or satellite-based communication, enabling global users to utilize high-precision real-time positioning.

[0073] Thirdly, the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method for ensuring the continuity of orbital state domain parameters.

[0074] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned method for ensuring the continuity of orbital state domain parameters.

[0075] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for guaranteeing continuity of track state domain parameters, characterized in that, The method comprises: Step 1, receiving global reference station data and broadcast ephemeris data, and calculating satellite real-time orbit correction values; Step 2, predicting the real-time orbit correction values based on a reinforcement learning model, comparing the predicted orbit correction values with precise orbit products provided by a third-party service center, and continuously training and optimizing the reinforcement learning model by using the obtained difference value data; wherein the reinforcement learning model comprises a multilayer perceptron model, the input layer comprises orbit parameter update time and satellite PRN number features; the hidden layer adopts a ReLU activation function; and the output layer outputs orbit correction value prediction results in three directions of radial, tangential and normal directions; Step 3, when the real-time orbit correction value calculation is interrupted, the predicted real-time orbit correction values are used for encoding and broadcasting to ensure the continuity of the orbit state domain parameters, wherein the encoding and broadcasting end encodes the orbit correction values according to the RTCM-SSR standard format, and identifies in the SSR version number that the current orbit product is actually generated or predicted generated, specifically comprising: analyzing the orbit correction information, obtaining time sequence state information, and determining that the real-time orbit correction calculation is abnormal when the difference between the current and previous time states is greater than a threshold value, the orbit correction values predicted by the reinforcement learning model are sent to the encoding and broadcasting end in time, and when the time sequence state is normal, the real-time orbit correction information is sent to the encoding and broadcasting end to encode the orbit correction values according to the RTCM-SSR standard format; when it is detected that the real-time orbit correction value calculation is interrupted for more than 60 seconds, the predicted orbit correction values are automatically switched for broadcasting.

2. The method of claim 1, wherein In step 1, the satellite position in the earth-fixed system is converted into the inertial system coordinates through coordinate conversion; and then converted back to the earth-fixed system coordinates after antenna phase center correction; and the satellite position in the orbit coordinate system in the radial, tangential and normal directions is calculated by using the satellite position in the broadcast ephemeris data and the corrected satellite position in the earth-fixed system.

3. The method of claim 1, wherein In step 2, the reinforcement learning model uses mean square error as a loss function, and the model is trained by an Adam optimizer; and when the 3D direction error between the prediction result and the precise orbit product provided by the third-party service center is less than 15 cm, it is determined that the model reaches a stable state.

4. The method of claim 3, wherein, In step 2, the model reaches a stable state is determined by the following formula: ， wherein, is the radial correction number difference, is the tangential correction number difference, is the radial correction number difference, is the 3D directional error.

5. A track state domain parameter continuity assurance system, characterized by, It comprises: A real-time orbit correction state domain calculation end for receiving global reference station data and broadcast ephemeris data, and calculating satellite real-time orbit correction values; A continuity guarantee service end for predicting the real-time orbit correction values based on a reinforcement learning model, comparing the predicted orbit correction values with precise orbit products provided by a third-party service center, and continuously training and optimizing the reinforcement learning model by using the obtained difference value data; wherein the reinforcement learning model comprises a multilayer perceptron model, the input layer comprises orbit parameter update time and satellite PRN number features; the hidden layer adopts a ReLU activation function; and the output layer outputs orbit correction value prediction results in three directions of radial, tangential and normal directions; The encoding broadcast end is used for encoding broadcast when the real-time orbit correction value calculation is interrupted, and the predicted real-time orbit correction value is used for encoding broadcast, so as to guarantee the continuity of the orbit state domain parameter, wherein the encoding broadcast end encodes the orbit correction value according to the RTCM-SSR standard format, and identifies the current orbit product as actual generation or prediction generation in the SSR version number, and specifically comprises: analyzing the orbit correction information, obtaining the time sequence state information, the difference between the current and the previous time state is greater than the threshold value, the system determines that the real-time orbit correction calculation is interrupted, the predicted orbit correction value is sent to the encoding broadcast end in time, and when the time sequence state is normal, the real-time orbit correction information is sent to the encoding broadcast end to encode the orbit correction value according to the RTCM-SSR standard format; when it is detected that the real-time orbit correction value calculation is interrupted for more than 60 seconds, the predicted orbit correction value is automatically switched to broadcast.

6. An electronic device, comprising: Comprise: One or more processors; Memory for storing one or more programs; When one or more programs are executed by the one or more processors, the one or more processors realize the orbit state domain parameter continuity guarantee method of any one of claims 1-4.

7. A computer readable storage medium characterized in that, Executable instructions are stored thereon, which can make the processor realize the orbit state domain parameter continuity guarantee method of any one of claims 1-4 when executed by the processor.

Images