Processing cloud platform for dynamically adjusting and optimizing Beidou measurement data

Abstract

The invention discloses a processing cloud platform for dynamically adjusting and optimizing Beidou measurement data. The processing cloud platform comprises an acquisition module which is responsible for acquiring Beidou measurement data from different area measurement points and data of an ionosphere real-time monitoring network in real time; the buffer is responsible for temporarily storing the Beidou measurement data sent by the analysis module and the processing module; the processing module is responsible for processing Beidou measurement data and ionosphere real-time monitoring network data in real time, performing error correction by adopting a dynamic compensation model and utilizing ionosphere data, and adopting an M estimation robust adjustment algorithm in a baseline resolving process; the analysis module is used for performing single-point analysis and precision analysis on the Beidou measurement data, performing reinforcement learning by using a PPO algorithm, constructing a prior knowledge base and accelerating the training process of reinforcement learning; the client module is used for receiving the charts and the reports sent by the analysis module, outputting the charts and the reports and issuing monitoring information; and the database is used for storing Beidou measurement data, ionosphere data and other related information.

Description

Technical Field

[0001] The present invention relates to the technical field of Beidou data processing, and particularly to a processing cloud platform for dynamically adjusting and optimizing Beidou measurement data. Background Art

[0002] With the development of global information technology, satellite navigation systems have become one of the indispensable basic facilities for military and civilian use. It can quickly and conveniently provide high-precision positioning information. GNSS (Global Navigation Satellite System) mainly includes the four major navigation systems: GPS of the United States, GLONASS of Russia, Beidou system of China, and GALILEO of the European Union. Among these systems, the Beidou satellite navigation system is a global satellite navigation system independently developed by China, which has great strategic significance for China's national defense security and people's livelihood. The measurement data of the Beidou system has a wide range of applications in many fields such as aviation, surveying and mapping, transportation, exploration, time service, etc. By dynamically adjusting and optimizing the Beidou measurement data through a cloud platform, the errors in the data can be eliminated or reduced, and the accuracy and reliability of the data can be improved.

[0003] In the prior art, error correction models are usually used to correct the errors in the observed data to improve the positioning accuracy. However, in terms of atmospheric correction, if only the Klobuchar model is used without considering real-time meteorological data or regional ionospheric grid information, it may lead to excessive model simplification. During the peak period of solar activity, the ionospheric delay error will increase significantly, which will have a serious impact on the positioning accuracy. Measured data shows that the correction rate of the Klobuchar model can only reach 60%-70%. Therefore, a processing cloud platform for dynamically adjusting and optimizing Beidou measurement data is proposed. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art, and a processing cloud platform for dynamically adjusting and optimizing Beidou measurement data is proposed.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data, comprising:

[0007] A collection module: responsible for collecting in real time the Beidou measurement data from measurement points in different regions and the data of the ionospheric real-time monitoring network (CMONOC). The Beidou measurement data includes Beidou observation data streams and broadcast ephemeris data streams, and is transmitted to the processing module through a 4G / 5G communication network;

[0008] A buffer: responsible for temporarily storing the Beidou measurement data sent by the analysis module and the processing module, and finally transmitting it to the database. At the same time, it is also responsible for obtaining the required data from the database;

[0009] Processing module: Responsible for real-time processing of Beidou measurement data and data from the China Monitoring Network for Ionosphere (CMONOC), including data quality verification, baseline solution, adjustment of the real-time monitoring network, and evaluation of solution quality. It uses a dynamic compensation model to correct errors using ionospheric data. During the baseline solution process, the M-estimation robust adjustment algorithm is used to reduce the impact of outliers on the solution results;

[0010] Analysis module: Conducts single-point analysis and accuracy analysis on Beidou measurement data, generates charts and reports, uses the Proximal Policy Optimization (PPO) algorithm for reinforcement learning, constructs a prior knowledge base, and accelerates the training process of reinforcement learning;

[0011] Client module: Receives the charts and reports sent by the analysis module, and performs output and monitoring information publishing. It introduces a kinematic extrapolation model and an LSTM neural network to predict position deviations, synthesizes the prediction results of kinematic extrapolation and the LSTM neural network, compensates and corrects the received position data, improves positioning accuracy and real-time performance, supports the deployment of a lightweight UKF on the drone side for high-frequency IMU data fusion, and runs a robust adaptive EKF in the cloud for precise processing;

[0012] Database: Stores Beidou measurement data, ionospheric data, and other relevant information, provides data query and retrieval functions, and stores the received data according to time periods (such as single-hour arcs, multi-hour arcs, and daily arcs).

[0013] The above technical solution further includes:

[0014] Furthermore, the acquisition module captures satellite signals through high-precision satellite receiving equipment, decodes the received satellite signals, extracts observation data, including key information such as pseudorange and carrier phase, and generates an observation data stream in a predetermined format from the decoded observation data;

[0015] The acquisition module receives navigation information broadcast by Beidou satellites, including satellite orbit parameters, clock parameters, etc., analyzes the received navigation information, extracts information useful for data processing, and generates a navigation message data stream in a predetermined format from the analyzed navigation information;

[0016] The acquisition module obtains ionospheric parameters, such as Total Electron Content (TEC), etc., in real time through ionospheric monitoring equipment, preprocesses the obtained ionospheric data, including data verification, format conversion, etc., and generates a data stream in a predetermined format from the preprocessed ionospheric data.

[0017] Further, the buffer uses memory as a temporary storage medium. The buffer organizes and manages Beidou measurement data using a data queue and needs to have the ability to handle concurrency. This can be achieved by using technologies such as multithreading, multiprocessing, or asynchronous I / O to ensure that data can be received and processed simultaneously.

[0018] Further, the specific data processed by the processing module for Beidou measurement data and data of the China Ionospheric Monitoring Network (CMONOC) are as follows:

[0019] Dynamic compensation model: An ionospheric delay model is established using historical ionospheric data or real-time ionospheric monitoring data. The ionospheric delay model describes the influence of the ionosphere on the satellite signal propagation path and is expressed as Δt ion = a·f -2 + b·TEC, where Δt ion is the predicted ionospheric delay value, f is the signal frequency, TEC (Total Electron Content) is the total electron content, a and b are model parameters. The ionospheric model and the dynamic compensation model are combined to correct the ionospheric delay of the original measurement data. The combined model is expressed as Δt corr = Δt meas - k·(Δt ion - Δt meas ), where Δt corr is the corrected ionospheric delay, Δt meas represents the measured ionospheric delay value, Δt ion is the predicted ionospheric delay value calculated according to the ionospheric model, and k is the dynamic compensation coefficient used to adjust the correction amplitude. This coefficient is adjusted in real time according to the change of the ionospheric state;

[0020] Data quality check: Perform quality check on the corrected data, including checks on aspects such as data integrity, continuity, and accuracy, to ensure the reliability of subsequent processing;

[0021] Time synchronization and alignment: Deploy an IEEE 1588v2 hardware timestamp module. Through the timestamp information provided by the IEEE 1588v2 hardware timestamp module, align the data of different stations in time, and further optimize by introducing a sliding window time alignment algorithm during the time alignment process;

[0022] Baseline solution: Use double-difference observations (inter-station difference and inter-satellite difference) of the received satellite observation data for baseline solution. During the baseline solution process, adopt the M-estimation robust adjustment algorithm to reduce the influence of outliers on the solution result;

[0023] Real-time monitoring network adjustment: Based on baseline solution, the entire monitoring network is adjusted to obtain position information;

[0024] Solution quality assessment: The quality of the solution results is evaluated, including accuracy assessment, reliability analysis, etc., to ensure the accuracy and reliability of the solution results;

[0025] Output of processing results: The processed data and solution results are formatted and output to the user or stored in the database. The output content includes parameters such as baseline length, coordinate difference, speed, acceleration, and their accuracy information.

[0026] Furthermore, the time synchronization and alignment include the following steps;

[0027] Hardware deployment: Deploy a hardware timestamp module that supports the IEEE 1588v2 protocol at each measurement station. The timestamp module is connected to the network switch to receive and send time synchronization information;

[0028] Configuration and calibration: Configure the hardware timestamp module to ensure that it can correctly receive and process IEEE 1588v2 time synchronization information. At the same time, regularly calibrate the timestamp module to maintain its time accuracy;

[0029] Timestamp generation: When the measurement station receives Beidou satellite measurement data, the hardware timestamp module adds a time tag to the measurement data;

[0030] In Beidou measurement data processing, time alignment is performed through a sliding window time alignment algorithm to eliminate time differences;

[0031] Define the sliding window: Set a sliding window that contains a certain number of consecutive observation data points. The size of the sliding window is adjusted according to the actual situation to balance alignment accuracy and computational complexity;

[0032] Calculate the time difference: For the data of each measurement station, calculate the time difference between its time tag and the reference time (such as the time tag of the main measurement station);

[0033] Alignment processing:

[0034] Initial alignment: Select the first data point within the sliding window as the alignment starting point and adjust its time tag to the reference time;

[0035] Sliding alignment: For the subsequent data points within the sliding window, perform corresponding time shift operations according to their time differences to achieve time alignment;

[0036] Window update: When the data points within the sliding window are processed, move the window to the next set of data points and repeat the above alignment process.

[0037] Further, the specific steps of the baseline solution are as follows:

[0038] Calculate the double-difference observation value: Assume that the observation values received by station A and station B are P A and P B respectively, the satellite clock biases are C1 and C2 (since the double-difference observation value is used, the influence of the satellite clock bias has actually been eliminated here), and the receiver clock biases are R A and R B (also eliminated), then the double-difference observation value is expressed as

[0039] Apply the Huber loss function: Assume that the error of the double-difference observation value is where truevalue is the true value of the baseline vector (in practical applications, it is estimated by solution), and calculate the weight of the double-difference observation value according to the Huber loss function;

[0040] Construct and solve the weighted normal equation: According to the Huber loss function and the weight of the double-difference observation value, construct the weighted normal equation, and then use the iterative method to solve the algorithm equation to obtain the estimated value of the baseline vector. The Huber loss function is expressed as where a is the error of the double-difference observation value and δ is the threshold parameter.

[0041] Further, the specific steps of the analysis module for single-point analysis and accuracy analysis of Beidou measurement data are as follows:

[0042] Data preprocessing: Preprocess the received data, including data cleaning (removing noise, outliers, etc.), format conversion, etc., to ensure the accuracy and efficiency of subsequent analysis;

[0043] Single-point analysis and accuracy analysis: Conduct single-point positioning analysis on the preprocessed data. Use the Kalman filtering algorithm, combine the measurement data of multiple satellites, and iteratively calculate the accurate position of the receiver. First, predict the state at the current moment according to the state estimate value at the previous moment. The state prediction formula is expressed as where is the state estimate value at the current moment, A is the state transition matrix, is the state estimate value at the previous moment, B is the control matrix (if there is control input), u k-1 is the control input at the previous moment. Next, update the prediction error covariance matrix to reflect the uncertainty of the prediction. The prediction error covariance matrix update formula is expressed as P k =AP k-1 A T +Q, where P k is the prediction error covariance matrix at the current moment, P k-1is the predicted error covariance matrix at the previous moment, A T is the transpose of the state transition matrix, Q is the process noise covariance matrix, calculate the Kalman gain, which is used to weigh the relative confidence between the measured value and the predicted value. The Kalman gain formula is expressed as K k = P k H T (HP k H T + R) -1 , where, K k is the Kalman gain at the current moment, H is the observation matrix (which maps the state to the observation space), R is the measurement noise covariance matrix. Use the measured value and the Kalman gain to update the state estimate. The state update formula is expressed as where, is the updated state estimate, z k is the measured value at the current moment. Finally, update the predicted error covariance matrix to reflect the updated uncertainty. The predicted error covariance matrix update (post) formula is expressed as P k ' = (1 - K kH )P k , where, P k ' is the updated predicted error covariance matrix. Evaluate the accuracy of the single-point positioning result. Use the root mean square error (RMSE) to evaluate the accuracy of the positioning result. Analyze the error sources. The root mean square error (RMSE) is expressed as where, is the positioning result, x i is the true position, such as satellite orbit error, receiver clock error, atmospheric delay, etc., and consider correcting them in subsequent processing. Assume that in a certain single-point positioning, the obtained RMSE is 5 meters. Through error analysis, it is found that the atmospheric delay is the main error source. For the atmospheric delay, a more accurate atmospheric model can be used for correction, or external data sources (such as meteorological data) can be introduced for auxiliary correction;

[0044] Reinforcement Learning and Filter Parameter Adjustment: Extract the filter parameters and the corresponding positioning accuracy from historical data as training samples. Use the Proximal Policy Optimization (PPO) algorithm to build a reinforcement learning model. Set the objective function as the improvement of positioning accuracy. Through iterative training, continuously adjust the filter parameters to maximize the objective function. During the training process, utilize the historical data and analysis results in the prior knowledge base to accelerate convergence. According to the training results, obtain the optimal combination of filter parameters and apply these parameters to real-time data processing to improve the positioning accuracy. Assume that during the reinforcement learning training process, it is found that the process noise covariance matrix Q of the Kalman filter has a significant impact on the positioning accuracy. Continuously adjust the value of Q through the PPO algorithm, and finally obtain a set of optimal Q value combinations. Apply this set of optimal Q values to real-time data processing, and it is found that the positioning accuracy has been significantly improved;

[0045] Generate Charts and Reports: Generate charts such as position trajectory charts, accuracy distribution charts, etc. based on the results of single-point analysis and accuracy analysis to visually display the analysis results. Compile a detailed report including analysis results, error analysis, recommended measures, etc. for users' reference and decision-making;

[0046] Data Output and Communication: Output the generated charts and reports to the specified storage location or display interface. The processing module and the client module conduct one-way communication to ensure the timely output and sharing of analysis results. The processing module is used for further data processing or fusion, while the client module is used to display the results to users or conduct subsequent analysis and decision-making.

[0047] Furthermore, the analysis module adopts a three-level degradation control logic, and the specific steps are as follows:

[0048] Fault Detection: Monitor the operating status of the analysis module in real time and detect potential faults or anomalies;

[0049] Fault Recovery: When a fault is detected, recover according to the preset fault recovery mechanism;

[0050] Level 1 Degradation Strategy: When the CMONOC data delay > 10s, enable the regional TEC prediction model as an alternative solution for ionospheric compensation to reduce the impact of the ionosphere on positioning accuracy;

[0051] Level 2 Degradation Strategy: When the regional TEC prediction model is unavailable, switch to the Klobuchar global model for ionospheric compensation to ensure the continuity and stability of positioning;

[0052] Level 3 Degradation Strategy: When both of the above models are unavailable, turn off the ionospheric compensation function and adopt a conservative strategy to ensure the stability of data processing, although this may reduce the positioning accuracy.

[0053] Furthermore, the specific steps of the client module:

[0054] Initialization and Connection Establishment: The user starts the client application and performs initialization operations, including loading necessary configuration parameters, establishing a network connection, etc. The client establishes a stable communication connection with the analysis module through a preset network protocol and address to ensure real-time data transmission and reception;

[0055] Data Reception and Processing: The client receives processed chart and report data from the analysis module, and preprocesses the received data, such as data format conversion, decoding, etc., for subsequent display and publication;

[0056] Hybrid Prediction Compensation: The client module receives a set of Beidou measurement data and compensates for its position deviation. First, a kinematic extrapolation model is used to predict the position at a future time based on the motion state of the object. The basic formula of the kinematic extrapolation model is where, x t+1 represents the position x at a future time t represents the position at the current time, v t represents the velocity at the current time, a t represents the acceleration at the current time, and Δt represents the time interval. Then, an LSTM neural network is used to correct the predicted position to reduce the position deviation. Finally, the corrected position is output as the final positioning result;

[0057] Monitoring Information Publication and Output: Based on the processed data, the client generates a monitoring report, including position information, trajectory map, accuracy evaluation, etc., and publishes the monitoring report to a specified platform or user terminal for users to view and analyze;

[0058] Disconnection and Resource Cleaning: When the client completes all data processing and publication tasks, it disconnects from the analysis module, releases network resources, and cleans up the cached data and log files in the client to ensure the clean and efficient operation of the system.

[0059] Furthermore, the client module adopts edge-cloud collaborative computing, and the specific steps are as follows:

[0060] Data Processing on the Drone Side: For mobile devices such as drones, the client supports deploying a lightweight UKF algorithm on the device side to perform preliminary fusion processing on high-frequency IMU data;

[0061] Precise Processing in the Cloud: The preliminarily processed data is uploaded to the cloud, and the cloud runs a robust adaptive EKF algorithm for further precise processing to obtain higher-precision position information;

[0062] Data Synchronization and Update: The data processed in the cloud is synchronized back to the client to ensure that the data displayed on the client is the latest and most accurate.

[0063] The present invention has the following beneficial effects:

[0064] 1. In the present invention, by using the real-time monitored ionospheric data, a dynamic compensation model is constructed to dynamically compensate the Beidou measurement data, effectively reducing the influence of ionospheric delay on the measurement accuracy. Combining with the optimized baseline solution and adjustment algorithms, in the process of baseline solution, the M-estimation robust adjustment algorithm is adopted to reduce the influence of outliers on the solution result and improve the data processing accuracy.

[0065] 2. In the present invention, a kinematic extrapolation model and an LSTM neural network are introduced to predict and compensate for the position deviation, reducing the position deviation and improving the positioning accuracy and real-time performance. In addition, the client module also supports edge-cloud collaborative computing. The lightweight UKF (Unscented Kalman Filter) is deployed on the drone side for high-frequency IMU data fusion, and the robust adaptive EKF is run on the cloud side for precise processing to solve the real-time bottleneck problem of adaptive filtering.

[0066] 3. In the present invention, the PPO algorithm is used for reinforcement learning to construct a prior knowledge base and accelerate the training process of reinforcement learning. During the reinforcement learning training process, the value of the process noise covariance matrix Q of the Kalman filter is continuously adjusted through the PPO algorithm to obtain the optimal Q value combination and improve the positioning accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 FIG. is a system block diagram of a processing cloud platform for dynamically adjusting and optimizing Beidou measurement data proposed by the present invention;

[0068] Figure 2 FIG. is a flowchart of the processing module for processing Beidou measurement data and ionospheric real-time monitoring network data. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0069] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0070] Please refer to Figure 1 as shown, the present invention is a processing cloud platform for dynamically adjusting and optimizing Beidou measurement data, including:

[0071] Acquisition module: responsible for real-time acquisition of Beidou measurement data from measurement points in different regions and data of the ionospheric real-time monitoring network (CMONOC). The Beidou measurement data includes Beidou observation data streams and broadcast ephemeris data streams, and is transmitted to the processing module through a 4G / 5G communication network;

[0072] Buffer: Responsible for temporarily storing the Beidou measurement data sent by the analysis module and the processing module, and finally sending it to the database. At the same time, it is also responsible for obtaining the required data from the database;

[0073] Processing module: Responsible for real-time processing of Beidou measurement data and data of the ionospheric real-time monitoring network (CMONOC), including data quality verification, baseline solution, real-time monitoring network adjustment, and solution quality evaluation. It uses a dynamic compensation model to correct errors using ionospheric data. During the baseline solution process, the M-estimation robust adjustment algorithm is used to reduce the influence of outliers on the solution result;

[0074] Analysis module: Conducts single-point analysis and accuracy analysis on Beidou measurement data, generates charts and reports, uses the PPO (Proximal Policy Optimization) algorithm for reinforcement learning, constructs a prior knowledge base, and accelerates the training process of reinforcement learning;

[0075] Client module: Receives the charts and reports sent by the analysis module, and performs output and monitoring information publishing. It introduces a kinematic extrapolation model and an LSTM neural network to predict position deviation, synthesizes the prediction results of kinematic extrapolation and the LSTM neural network, compensates and corrects the received position data, improves positioning accuracy and real-time performance, supports the deployment of a lightweight UKF on the drone side for high-frequency IMU data fusion, and runs a robust adaptive EKF in the cloud for precise processing;

[0076] Database: Stores Beidou measurement data, ionospheric data, and other relevant information, provides data query and retrieval functions, and stores the received data according to time periods (such as single-hour arcs, multi-hour arcs, and daily arcs).

[0077] In one embodiment, the acquisition module captures satellite signals through high-precision satellite receiving equipment, decodes the received satellite signals, extracts observation data, including key information such as pseudorange and carrier phase, and generates an observation data stream in a predetermined format from the decoded observation data;

[0078] The acquisition module receives the navigation information broadcast by Beidou satellites, including satellite orbit parameters, clock parameters, etc., parses the received navigation information, extracts the information useful for data processing, and generates a navigation message data stream in a predetermined format from the parsed navigation information;

[0079] The acquisition module obtains ionospheric parameters, such as total electron content (TEC), etc., in real time through ionospheric monitoring equipment, preprocesses the obtained ionospheric data, including data verification, format conversion, etc., and generates a data stream in a predetermined format from the preprocessed ionospheric data.

[0080] In one embodiment, the buffer uses memory as a temporary storage medium. The buffer uses a data queue to organize and manage Beidou measurement data, and the buffer needs to have the ability to handle concurrency. This can be achieved by using technologies such as multi-threading, multi-processing, or asynchronous I / O to ensure that data can be received and processed simultaneously.

[0081] In one embodiment, the processing module processes the specific data of Beidou measurement data and data of the Ionospheric Real-time Monitoring Network (CMONOC):

[0082] Dynamic compensation model: An ionospheric delay model is established using historical ionospheric data or real-time ionospheric monitoring data. The ionospheric delay model describes the influence of the ionosphere on the satellite signal propagation path, and the ionospheric delay model is expressed as Δt ion = a·f -2 + b·TEC, where Δt ion is the predicted value of ionospheric delay, f is the signal frequency, TEC (Total Electron Content) is the total electron content, a and b are model parameters. The ionospheric model and the dynamic compensation model are combined to correct the ionospheric delay of the original measurement data. The combined model is expressed as Δt corr = Δt meas - k·(Δt ion - Δt meas ), where Δt corr is the corrected ionospheric delay, Δt meas represents the measured value of ionospheric delay, Δt ion is the predicted value of ionospheric delay calculated according to the ionospheric model, and k is the dynamic compensation coefficient used to adjust the correction amplitude. This coefficient is adjusted in real time according to the change of the ionospheric state;

[0083] Data quality verification: Perform quality verification on the corrected data, including checks on aspects such as data integrity, continuity, and accuracy to ensure the reliability of subsequent processing;

[0084] Time synchronization and alignment: Deploy an IEEE 1588v2 hardware timestamp module. Through the timestamp information provided by the IEEE 1588v2 hardware timestamp module, align the data of different stations in time, and further optimize by introducing a sliding window time alignment algorithm during the time alignment process;

[0085] Baseline solution: Use double-difference observations (inter-station difference and inter-satellite difference) of the received satellite observation data for baseline solution. During the baseline solution process, adopt the M-estimation robust adjustment algorithm to reduce the influence of outliers on the solution result;

[0086] Real-time monitoring network adjustment: Based on baseline solution, the entire monitoring network is adjusted to obtain position information;

[0087] Solution quality assessment: The quality of the solution results is assessed, including accuracy evaluation, reliability analysis, etc., to ensure the accuracy and reliability of the solution results;

[0088] Output processing results: The processed data and solution results are formatted and output to the user or stored in the database. The output content includes parameters such as baseline length, coordinate difference, speed, acceleration, etc. and their accuracy information.

[0089] In one embodiment, the time synchronization and alignment includes the following steps;

[0090] Hardware deployment: Deploy a hardware timestamp module that supports the IEEE 1588v2 protocol at each measurement station. The timestamp module is connected to the network switch to receive and send time synchronization information;

[0091] Configuration and calibration: Configure the hardware timestamp module to ensure that it can correctly receive and process IEEE 1588v2 time synchronization information. At the same time, regularly calibrate the timestamp module to maintain its time accuracy;

[0092] Timestamp generation: When the measurement station receives Beidou satellite measurement data, the hardware timestamp module adds a time tag to the measurement data;

[0093] In Beidou measurement data processing, time alignment is performed through a sliding window time alignment algorithm to eliminate time differences;

[0094] Define the sliding window: Set a sliding window that contains a certain number of consecutive observation data points. The size of the sliding window is adjusted according to the actual situation to balance alignment accuracy and computational complexity;

[0095] Calculate the time difference: For the data of each measurement station, calculate the time difference between its time tag and the reference time (such as the time tag of the main measurement station);

[0096] Alignment processing:

[0097] Initial alignment: Select the first data point within the sliding window as the alignment starting point and adjust its time tag to the reference time;

[0098] Sliding alignment: For the subsequent data points within the sliding window, perform corresponding time shift operations according to their time differences to achieve time alignment;

[0099] Window update: When the data points within the sliding window are processed, move the window to the next data point set and repeat the above alignment process.

[0100] In one embodiment, the specific steps of the baseline solution are as follows:

[0101] Calculate the double-difference observation value: Assume that the observation values received by station A and station B are P A and P B , the satellite clock biases are C1 and C2 respectively (since the double-difference observation value is adopted, the influence of the satellite clock bias has actually been eliminated here), and the receiver clock biases are R A and R B (similarly eliminated), then the double-difference observation value is expressed as

[0102] Apply the Huber loss function: Assume that the error of the double-difference observation value is where truevalue is the true value of the baseline vector (in practical applications, it is estimated by solution), and calculate the weight of the double-difference observation value according to the Huber loss function;

[0103] Construct and solve the weighted normal equation: According to the Huber loss function and the weight of the double-difference observation value, construct the weighted normal equation, and then use the iterative method to solve the algorithm equation to obtain the estimated value of the baseline vector. The Huber loss function is expressed as where a is the error of the double-difference observation value and δ is the threshold parameter.

[0104] In one embodiment, the specific steps of the analysis module for performing single-point analysis and accuracy analysis on Beidou measurement data are as follows:

[0105] Data preprocessing: Preprocess the received data, including data cleaning (removing noise, outliers, etc.), format conversion, etc., to ensure the accuracy and efficiency of subsequent analysis;

[0106] Single-point analysis and accuracy analysis: Perform single-point positioning analysis on the preprocessed data, use the Kalman filtering algorithm, combine the measurement data of multiple satellites, and iteratively calculate the accurate position of the receiver. First, predict the state at the current moment according to the state estimate value at the previous moment. The state prediction formula is expressed as where is the state estimate value at the current moment, A is the state transition matrix, is the state estimate value at the previous moment, B is the control matrix (if there is a control input), u k-1 is the control input at the previous moment. Next, update the prediction error covariance matrix to reflect the uncertainty of the prediction. The prediction error covariance matrix update formula is expressed as P k =AP k-1 A T +Q, where P k is the prediction error covariance matrix at the current moment, Pk-1 is the prediction error covariance matrix at the previous moment, A T is the transpose of the state transition matrix, Q is the process noise covariance matrix, calculate the Kalman gain, which is used to weigh the relative trust between the measurement value and the predicted value. The Kalman gain formula is expressed as K k = P k H T (HP k H T + R) -1 , where K k is the Kalman gain at the current moment, H is the observation matrix (which maps the state to the observation space), R is the measurement noise covariance matrix. Use the measurement value and the Kalman gain to update the state estimate value. The state update formula is expressed as where is the updated state estimate value, z k is the measurement value at the current moment. Finally, update the prediction error covariance matrix to reflect the updated uncertainty. The prediction error covariance matrix update (post) formula is expressed as P' k =(1 - K kH )P k , where P' k is the updated prediction error covariance matrix. Evaluate the accuracy of the single-point positioning result. Use the root mean square error (RMSE) to evaluate the accuracy of the positioning result. Analyze the error sources. The root mean square error (RMSE) is expressed as where is the positioning result, x i is the true position, such as satellite orbit error, receiver clock error, atmospheric delay, etc., and consider correcting them in subsequent processing. Assume that in a certain single-point positioning, the obtained RMSE is 5 meters. Through error analysis, it is found that the atmospheric delay is the main error source. For the atmospheric delay, a more accurate atmospheric model can be used for correction, or external data sources (such as meteorological data) can be introduced for auxiliary correction;

[0107] Reinforcement Learning and Filter Parameter Adjustment: Extract filter parameters and corresponding positioning accuracies from historical data as training samples. Use the Proximal Policy Optimization (PPO) algorithm to build a reinforcement learning model. Set the objective function as the improvement of positioning accuracy. Through iterative training, continuously adjust the filter parameters to maximize the objective function. During the training process, utilize the historical data and analysis results in the prior knowledge base to accelerate convergence. According to the training results, obtain the optimal combination of filter parameters and apply these parameters to real-time data processing to improve positioning accuracy. Assume that during the reinforcement learning training process, it is found that the process noise covariance matrix Q of the Kalman filter has a significant impact on positioning accuracy. Continuously adjust the value of Q through the PPO algorithm, and finally obtain a set of optimal Q value combinations. Apply this set of optimal Q values to real-time data processing and find that the positioning accuracy has been significantly improved;

[0108] Generate Charts and Reports: Generate charts such as position trajectory charts, accuracy distribution charts, etc. based on the results of single-point analysis and accuracy analysis to visually display the analysis results. Compile a detailed report including analysis results, error analysis, recommended measures, etc. for users' reference and decision-making;

[0109] Data Output and Communication: Output the generated charts and reports to the specified storage location or display interface. The processing module and the client module conduct one-way communication to ensure the timely output and sharing of analysis results. The processing module is used for further data processing or fusion, while the client module is used to display the results to users or perform subsequent analysis and decision-making.

[0110] In one embodiment, the analysis module adopts a three-level degradation control logic. The specific steps are as follows:

[0111] Fault Detection: Real-time monitor the operating status of the analysis module and detect potential faults or anomalies;

[0112] Fault Recovery: When a fault is detected, perform recovery according to the preset fault recovery mechanism;

[0113] Level 1 Degradation Strategy: When the CMONOC data delay > 10s, enable the regional TEC prediction model as an alternative solution for ionospheric compensation to reduce the impact of the ionosphere on positioning accuracy;

[0114] Level 2 Degradation Strategy: When the regional TEC prediction model is unavailable, switch to the Klobuchar global model for ionospheric compensation to ensure the continuity and stability of positioning;

[0115] Level 3 Degradation Strategy: When both of the above models are unavailable, turn off the ionospheric compensation function and adopt a conservative strategy to ensure the stability of data processing, although this may reduce the positioning accuracy.

[0116] In one embodiment, the specific steps of the client module are as follows:

[0117] Initialization and connection establishment: The user starts the client application and performs initialization operations, including loading necessary configuration parameters, establishing a network connection, etc. The client establishes a stable communication connection with the analysis module through a preset network protocol and address to ensure real-time data transmission and reception.

[0118] Data reception and processing: The client receives the processed chart and report data from the analysis module and preprocesses the received data, such as data format conversion, decoding, etc., for subsequent display and publication.

[0119] Hybrid prediction compensation: The client module receives a set of Beidou measurement data and compensates for its position deviation. First, a kinematic extrapolation model is used to predict the position at a future time based on the motion state of the object. The basic formula of the kinematic extrapolation model is where x t+1 represents the position at the future time x t represents the position at the current time, v t represents the velocity at the current time, a t represents the acceleration at the current time, and Δt represents the time interval. Then, an LSTM neural network is used to correct the predicted position to reduce the position deviation. Finally, the corrected position is output as the final positioning result.

[0120] Monitoring information publication and output: Based on the processed data, the client generates a monitoring report, including position information, trajectory map, accuracy evaluation, etc., and publishes the monitoring report to a specified platform or user terminal for users to view and analyze.

[0121] Disconnection and resource cleaning: When the client completes all data processing and publication tasks, it disconnects from the analysis module, releases network resources, and cleans the cached data and log files in the client to ensure the clean and efficient operation of the system.

[0122] In one embodiment, the client module adopts edge-cloud collaborative computing, and the specific steps are as follows:

[0123] Data processing on the drone side: For mobile devices such as drones, the client supports deploying a lightweight UKF algorithm on the device side to perform preliminary fusion processing on high-frequency IMU data.

[0124] Precise processing in the cloud: The preliminarily processed data is uploaded to the cloud, and the cloud runs a robust adaptive EKF algorithm for further precise processing to obtain higher-precision position information.

[0125] Data synchronization and update: Synchronize the data processed in the cloud back to the client to ensure that the data displayed on the client is the latest and most accurate.

[0126] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data, characterized in that, Including: Collection module: Responsible for collecting in real time Beidou measurement data from measurement points in different regions and data from the ionospheric real-time monitoring network. The Beidou measurement data includes Beidou observation data streams and broadcast ephemeris data streams, and is transmitted to the processing module through a 4G / 5G communication network. Buffer: Responsible for temporarily storing the Beidou measurement data sent by the analysis module and the processing module, and finally sending it to the database. At the same time, it is also responsible for obtaining the required data from the database. Processing module: Responsible for processing in real time the Beidou measurement data and the data of the ionospheric real-time monitoring network, including data quality verification, baseline solution, real-time monitoring network adjustment, and solution quality assessment. A dynamic compensation model is used to correct errors using ionospheric data. During the baseline solution process, the M-estimation robust adjustment algorithm is used to reduce the influence of outliers on the solution result. Analysis module: Conducts single-point analysis and accuracy analysis on the Beidou measurement data, generates charts and reports, uses the PPO algorithm for reinforcement learning, constructs a prior knowledge base, and accelerates the training process of reinforcement learning. Client module: Receives the charts and reports sent by the analysis module, and performs output and monitoring information release. Introduces a kinematic extrapolation model and an LSTM neural network to predict position deviations, synthesizes the prediction results of kinematic extrapolation and the LSTM neural network, and compensates and corrects the received position data. Database: Stores Beidou measurement data, ionospheric data, and other relevant information, provides data query and retrieval functions, and stores the received data according to time periods.

2. The processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 1, wherein The collection module captures satellite signals through a satellite receiving device, decodes the received satellite signals, and generates an observation data stream from the decoded observation data in a predetermined format. The collection module receives the navigation information broadcast by Beidou satellites, analyzes the received navigation information, extracts the information useful for data processing, and generates a broadcast ephemeris data stream from the analyzed navigation information in a predetermined format. The collection module obtains ionospheric parameters in real time through ionospheric monitoring devices, preprocesses the obtained ionospheric data, and generates a data stream from the preprocessed ionospheric data in a predetermined format.

3. A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 1, characterized in that, The buffer uses memory as a temporary storage medium, and the buffer uses a data queue to organize and manage Beidou measurement data.

4. A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 1, characterized in that, Specific data for which the processing module processes Beidou measurement data and data of the ionospheric real-time monitoring network: Dynamic compensation model: An ionospheric delay model is established using historical ionospheric data or real-time ionospheric monitoring data. The ionospheric delay model describes the influence of the ionosphere on the satellite signal propagation path, and the ionospheric delay model is expressed as Δt ion = a·f -2 + b·TEC, where Δt ion is the predicted value of ionospheric delay, f is the signal frequency, TEC is the total electron content, a and b are model parameters. The ionospheric model is combined with the dynamic compensation model to correct the ionospheric delay of the original measurement data. The combined model is expressed as Δt corr = Δt meas - k·(Δt ion - Δt meas ), where Δt corr is the corrected ionospheric delay, Δt meas represents the measured value of ionospheric delay, Δt ion is the predicted value of ionospheric delay calculated according to the ionospheric model, and k is the dynamic compensation coefficient used to adjust the amplitude of the correction; Data quality verification: Verifies the quality of the corrected data. Time synchronization and alignment: Deploys an IEEE 1588v2 hardware timestamp module. Through the timestamp information provided by the IEEE 1588v2 hardware timestamp module, the data of different measurement stations is time-aligned. During the time alignment process, a sliding window time alignment algorithm is introduced for further optimization. Baseline solution: The received satellite observation data is used for baseline solution using double-difference observations. During the baseline solution process, the M-estimation robust adjustment algorithm is used to reduce the influence of outliers on the solution result. Real-time monitoring network adjustment: On the basis of the baseline solution, the entire monitoring network is adjusted to obtain position information. Solution quality assessment: Conduct quality assessment on the solution results; Output processing results: Format and output the processed data and solution results to the user or store them in the database.

5. The processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 3, characterized in that, The time synchronization and alignment, include the following steps; Hardware deployment: Deploy a hardware timestamp module that supports the IEEE 1588v2 protocol at each measurement station. The timestamp module is connected to the network switch to receive and send time synchronization information; Configuration and calibration: Configure the hardware timestamp module and, at the same time, regularly calibrate the timestamp module; Timestamp generation: When the measurement station receives the measurement data of Beidou satellites, the hardware timestamp module adds a time tag to the measurement data; In Beidou measurement data processing, time alignment is performed through a sliding window time alignment algorithm to eliminate time differences; Define the sliding window: Set a sliding window that contains a certain number of consecutive observation data points; Calculate the time difference: For the data of each measurement station, calculate the time difference between its time tag and the reference time; Alignment processing: Initial alignment: Select the first data point within the sliding window as the alignment starting point and adjust its time tag to the reference time; Sliding alignment: For the subsequent data points within the sliding window, perform corresponding time shift operations according to their time differences to achieve time alignment; Window update: When the data points within the sliding window are processed, move the window to the next set of data points and repeat the above alignment process.

6. The processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 3, wherein, The specific steps of the baseline solution: Calculate the double-difference observation value: Let the observation values received by station A and station B be P A and P B , the satellite clock errors be C1 and C2 respectively, and the receiver clock errors be R A and R B , then the double-difference observation value is expressed as Apply the Huber loss function: Let the error of the double-difference observation be where truevalue is the true value of the baseline vector, and the weight of the double-difference observation is calculated according to the Huber loss function; Construct a weighted normal equation and solve it: According to the Huber loss function and the weights of double-difference observations, construct a weighted normal equation, and then use an iterative method to solve the normal equation to obtain an estimated value of the baseline vector. The Huber loss function is expressed as where a is the error of the double-difference observation and δ is the threshold parameter.

7. A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 1, characterized in that, The specific steps of the analysis module for single-point analysis and accuracy analysis of Beidou measurement data: Data preprocessing: Preprocess the received data; Single-point analysis and accuracy analysis: Perform single-point positioning analysis on the preprocessed data. Use the Kalman filtering algorithm, combine the measurement data of multiple satellites, iteratively calculate the accurate position of the receiver, evaluate the accuracy of the single-point positioning result, use the root mean square error to evaluate the accuracy of the positioning result, analyze the error sources, and the root mean square error is expressed as where is the positioning result, and x i is the true position; Reinforcement learning and filtering parameter adjustment: Extract the filtering parameters and the corresponding positioning accuracy from historical data as training samples. Use the proximal policy optimization algorithm to build a reinforcement learning model. Set the objective function as the improvement of positioning accuracy. Through iterative training, continuously adjust the filtering parameters to maximize the objective function. During the training process, use the historical data and analysis results in the prior knowledge base to accelerate convergence. According to the training results, obtain the optimal combination of filtering parameters and apply these parameters to real-time data processing; Generate charts and reports: Generate charts and write reports based on the results of single-point analysis and accuracy analysis; Data output and communication: Output the generated charts and reports to the specified storage location or display interface, and the processing module and the client module conduct one-way communication.

8. A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 1, characterized in that, The analysis module adopts a three-level degradation control logic, specific steps: Fault detection: Real-time monitor the operating status of the analysis module and detect potential faults or anomalies; Fault recovery: When a fault is detected, perform recovery according to the preset fault recovery mechanism; Level 1 degradation strategy: When the CMONOC data delay > 10s, enable the regional TEC prediction model as an alternative for ionospheric compensation; Level 2 degradation strategy: When the regional TEC prediction model is unavailable, switch to the Klobuchar global model for ionospheric compensation; Level 3 degradation strategy: When both of the above models are unavailable, turn off the ionospheric compensation function and adopt a conservative strategy.

9. A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 1, characterized in that, The specific steps of the client module: Initialization and connection establishment: The user starts the client application for initialization. The client establishes a communication connection with the analysis module through a preset network protocol and address. Data reception and processing: The client receives processed chart and report data from the analysis module and preprocesses the received data. Hybrid Prediction Compensation: The client module receives a set of Beidou measurement data and performs position deviation compensation on it. First, the kinematic extrapolation model is used to predict the position at a future time based on the motion state of the object. The basic formula of the kinematic extrapolation model is where x t+1 represents the position x at a future time t represents the position at the current time, v t represents the velocity at the current time, a t represents the acceleration at the current time, and Δt represents the time interval. Then, the LSTM neural network is used to correct the predicted position. Finally, the corrected position is output as the final positioning result; Monitoring information publication and output: Based on the processed data, the client generates a monitoring report and publishes the monitoring report to a specified platform or user terminal. Disconnection and resource cleaning: When the client completes all data processing and publication tasks, it disconnects from the analysis module, releases network resources, and cleans up the cached data and log files in the client.

10. A processing cloud platform for dynamically adjusting and optimizing Beidou measurement data according to claim 1, characterized in that, The client module adopts edge-cloud collaborative computing, and the specific steps are as follows: Data processing on the drone side: For mobile devices, the client supports deploying a lightweight UKF algorithm on the device side to perform preliminary fusion processing on high-frequency IMU data. Fine processing in the cloud: The preliminarily processed data is uploaded to the cloud, and the cloud runs a robust adaptive EKF algorithm for further fine processing to obtain position information. Data synchronization and update: The data processed in the cloud is synchronized back to the client.

Images