GNSS network RTK reference station distributed data processing method and device

Abstract

The invention discloses a GNSS network RTK base station distributed data processing method and device, belongs to the technical field of satellite navigation and positioning, and is used for solving key problems of base station networking, baseline solution, error correction calculation, model establishment and the like in a large GNSS network RTK system. The method is especially suitable for efficient calculation, task allocation and dynamic load balancing processing of a large-scale base station network. The method comprises the following steps: 1) dynamically networking a core server, determining a baseline to be solved, generating a baseline solving task and establishing an error model; (2) a plurality of data processing servers solve all baselines in parallel, and baseline error correction numbers are extracted; (3) each base station calculates relative corrections of other stations relative to the base station according to the related baseline corrections, and establishes an error model, and (4) dynamically allocates a baseline resolving task and an error modeling task. Through multi-thread and distributed parallel computing, the task processing efficiency and the resource utilization rate of a large-scale base station network are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of satellite navigation and positioning, and particularly relates to a method and device for distributed data processing of a large-scale reference station for GNSS network real-time kinematic (RTK) positioning. Background Art

[0002] GNSS network RTK (Real-Time Kinematic) technology is a high-precision positioning method based on carrier phase observation, which is widely used in fields such as surveying and mapping, engineering monitoring, and geological disaster warning. Its core is to calculate the precise positioning information of the user terminal through the cooperation of multiple reference stations. In this process, the data processing of the reference station includes multiple links such as network formation and baseline solution, error correction number extraction and integration, error model establishment and data dissemination. The baseline solution task needs to calculate the baseline vectors between reference stations and extract error correction numbers, and then integrate the correction numbers of multiple baselines into relative correction numbers to establish a global error model and generate differential data. However, with the continuous expansion of the scale of the reference station network, the existing centralized processing architecture has gradually exposed problems such as low efficiency, insufficient scalability, and poor system stability.

[0003] Currently, the data processing of large-scale reference stations in GNSS network RTK mainly adopts a centralized architecture, that is, the core server uniformly processes all reference station data. This method is more applicable to small-scale reference station networks in the early stage, but when faced with hundreds or even thousands of reference stations, its computing power and task processing efficiency are difficult to meet the requirements. The computing bottleneck is one of the main problems of the centralized architecture. The large amount of baseline solution tasks and error model establishment calculations are huge, and the core server is prone to overloading, resulting in reduced real-time performance. In addition, the centralized system lacks the ability of dynamic management of task allocation and resource utilization, with uneven load distribution and difficulty in expansion. When the scale of the reference station network expands or the core server fails, the system may become paralyzed, seriously affecting the reliability of the positioning service. Summary of the Invention

[0004] In view of the above problems, the distributed data processing technology provides an efficient solution for GNSS network RTK. By introducing a distributed computing architecture and allocating the baseline solution and error model establishment tasks to multiple processing servers, parallel processing of tasks can be achieved, effectively improving the computing efficiency. In addition, through dynamic task scheduling and load balancing mechanisms, the distributed architecture makes full use of the computing resources of each server, while reducing the pressure on the core server. Through the message bus to achieve efficient transmission of reference station data, task allocation and result interaction, the distributed system can also flexibly adapt to the needs of the expansion of the reference station network scale. In summary, the application of distributed data processing in GNSS network RTK has significant performance improvement and technical advantages, and has gradually become the mainstream direction of large-scale reference station network data processing.

[0005] The present invention proposes a GNSS network RTK reference station distributed data processing method and system. Through distributed architecture design and task dynamic allocation mechanism, it solves the problems existing in the existing centralized processing and subnet processing modes, such as computing bottlenecks, low task scheduling efficiency, poor scalability, and complex coordination, and realizes the efficient processing of large-scale reference station network data and real-time error correction services.

[0006] The technical solution of the present invention is a GNSS network RTK reference station distributed data processing method, including the following steps: Step 1, the core server receives the real-time observation data of all reference stations through the message bus, and determines the available reference stations; based on the available reference stations, dynamically groups baselines based on Delaunay triangles, optimizes the number of baselines and geometric conditions, and obtains a baseline set; uses the baseline set as the task to be solved and assigns it to the baseline solution server; uses the available reference station set as the error modeling task and assigns it to the error model establishment server; Step 2, the baseline solution server first calculates the real-time resource occupancy assessment of the baseline solution server, and then performs baseline solution to calculate the double-difference ionospheric delay and double-difference tropospheric delay; Step 3, the error model establishment server calculates the real-time resource occupancy assessment of the error model establishment server, then parallelly solves all baselines, extracts the baseline error correction numbers, calculates the relative correction numbers of other stations relative to the local station, and thus establishes the master station error model; Step 4, the core server calculates the corresponding remaining computing capabilities according to the real-time resource occupancy assessment of the baseline solution server and the real-time resource occupancy assessment of the error model establishment server respectively, and dynamically allocates baseline solution tasks and error modeling tasks according to the corresponding remaining computing capabilities.

[0007] Further, in step 1, the specific implementation method of obtaining the baseline set is as follows: Step 1.2.1, input the coordinates of all reference stations to form a point set , wherein refers to the abscissa and ordinate of the i-th point; Step 1.2.2, initialize the triangle, and select a super triangle containing all points from the point set to cover the entire point set; Step 1.2.3, sequentially insert the point into the current triangulation, that is, the result of the initialized triangle, and check whether the point is inside the circumcircle of a certain triangle. If satisfied, reconstruct the relevant triangles; Step 1.2.4, if a certain side does not meet the Delaunay condition, that is, other points are included in the circumcircle, exchange the sides and update the triangle structure; Step 1.2.5, remove the sides of the super triangle and output the remaining triangle set and baseline set.

[0008] Furthermore, in Step 2, the real-time resource occupancy evaluation calculation method of the baseline processing server is as follows: Set the number of baselines processed by the baseline processing server to , and the CPU and memory resources occupied by each baseline solution are equal. The CPU occupancy rate of the baseline solution process in the current server is , and the memory occupancy rate is , then the CPU occupancy rate of a single baseline is and the memory occupancy rate is respectively:

[0009] . Furthermore, in Step 2, the calculation processes of the double-difference ionospheric delay and double-difference tropospheric delay are as follows: Step 2.2.1, the baseline solution server obtains the coordinate and observation value information of the baseline to be solved from the message bus; Step 2.2.2, based on the HMW combination method, determine the Beidou ultra-wide lane ambiguity, then based on the TCAR method of the ionosphere-free combination, determine the wide lane ambiguity, and finally determine the original ambiguity of the satellite based on the wide lane and dual-frequency ionosphere-free combined ambiguity; Step 2.2.3, according to the already fixed ambiguity and reference station coordinate information, extract the error information related to and independent of the baseline geometry. According to the L1 and L2 carrier double-difference observation equations:

[0010]

[0011] where , , , are the frequencies of carriers L1 and L2 respectively, , are the wavelengths of carriers L1 and L2 respectively, is the double-difference phase observation value on carriers L1 and L2, is the double-difference geometric distance, are the L1 and L2 double-difference ambiguities, is the double-difference tropospheric delay, is the double-difference ionospheric delay; finally, the double-difference ionospheric delay and double-difference tropospheric delay with centimeter-level accuracy at the reference station are obtained, and their estimation formulas are as follows:

[0012] 。

[0013] Further, in step 3, the calculation method for real-time resource occupancy evaluation of the error model establishment server is as follows: Set the current number of sites processed by the error model establishment server to , the CPU and memory resources occupied by the relative error calculation of each site are equal, and the CPU occupancy rate of the main station error model establishment process in the current error model establishment server is , and the memory occupancy rate is , then the CPU occupancy rate for establishing the error model of 1 main station is and the memory occupancy rate is respectively:

[0014] 。

[0015] Further, in step 3, the specific implementation method for establishing the main station error model is as follows: Step 3.2.1, the error model establishment server obtains the relative error information of the baselines related to the site to be solved and several surrounding sites from the message bus; Step 3.2.2, use the overall adjustment algorithm of baseline error correction numbers to extract baseline error correction numbers, which are divided into three steps: 1) Use the triangle network closure difference test to judge whether there are gross errors; 2) Use the error correction numbers of each baseline to perform the overall adjustment calculation of relative error correction numbers; 3) Based on the adjustment results, locate and repair the gross errors and perform adjustment again; Step 3.2.3, integrate the baseline error correction numbers of all main stations to construct the main station error model, which is expressed by the interpolation method as:

[0016] where is the total number of secondary stations, is the secondary station serial number, is the correction number of the secondary station , is the interpolation coefficient of the secondary station , is the secondary station 's coordinate value relative to the position to be interpolated, is the correction number of the position to be interpolated. Further, in 1) of step 3.2.2, the closure difference of the baseline error correction numbers of each side in the same triangle is within the minimum threshold range. Therefore, if the closure difference of the triangle network exceeds the minimum threshold range, there must be at least one baseline error correction number in the triangle network that is incorrect.

[0017] Further, in item 2) of step 3.2.2, according to the basic principle of indirect adjustment, the baseline vector adjustment method is adopted to solve the corrections for a specified master station. According to the relationship of double differences and baseline transmission, the following error equations describing the relationships of each baseline vector are listed: (2) Where are the error corrections of each baseline in the reference station triangle network, is the adjusted estimate of the double-difference relative error correction from the master station to other reference stations, is the coefficient matrix describing the linear relationship between the error corrections of each baseline in the triangular network and the relative error correction of the master station, is the residual of the error correction.

[0018] Further, set the memory and CPU occupancy rates of all the baseline processing servers to and , and the maximum memory and CPU occupancy are both P%. Then the remaining computing power of the baseline processing server is:

[0019] Where, and are the CPU occupancy rate and memory occupancy rate of a single baseline respectively; Set the memory and CPU occupancy rates of all the error model establishment servers to and , and the maximum memory and CPU occupancy are both P%. Then the remaining computing power of the error model establishment server is:

[0020] Where, and are the CPU occupancy rate and memory occupancy rate for establishing the error model of 1 master station respectively; And and meet the following conditions: ,

[0021] The core server dynamically allocates baseline solution tasks and error modeling tasks according to and , that is, it gives priority to the tasks with a large number of baselines to be processed.

[0022] The present invention also provides GNSS network RTK reference station distributed data processing, including: A processor and a memory, where the memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the GNSS network RTK reference station distributed data processing method as described in the above technical solution.

[0023] The present invention provides a GNSS network RTK reference station distributed data processing method and system. Through a distributed architecture and a task dynamic scheduling mechanism, the deficiencies of the existing centralized processing and subnet processing modes are overcome, and the system performance and reliability are significantly improved. The specific beneficial effects are as follows: 1. The calculation efficiency is significantly improved Adopting a distributed computing architecture, the baseline solution and error model establishment tasks are assigned to multiple data processing servers, making full use of the computing resources of multiple servers, realizing parallel processing of tasks, and significantly reducing the calculation time. The dynamic task scheduling mechanism allocates tasks based on the real-time resource status of the servers, avoiding waste of computing resources and improving the overall computing efficiency of the system.

[0024] 2. The system scalability is enhanced The distributed architecture supports flexible expansion and can quickly access new data processing servers according to the increase in the scale of the reference station network. Whether the number of reference stations increases or the complexity of the calculation tasks increases, the system can maintain efficient operation by adding computing nodes. It supports dynamic networking and task allocation strategies, so that the addition or deletion of reference stations or changes in the quality of observation data will not affect the normal operation of the system.

[0025] 3. The reliability and robustness are improved Through the distributed allocation of tasks and the cooperation of multiple servers, the system avoids the risk of single-point failure. Even if some servers fail, other servers can still continue to complete tasks, ensuring the stability of the system. The core server is mainly responsible for task allocation and result integration, and does not directly participate in the calculation, greatly reducing the load pressure and failure risk of the core server.

[0026] 4. The real-time performance of data processing is enhanced Using a message bus to achieve efficient data interaction, supporting multi-threaded asynchronous communication, ensuring the rapid upload of reference station data, the real-time allocation of tasks, and the rapid feedback of calculation results.

[0027] Through distributed parallel processing, the system shortens the processing time of baseline solution and error model establishment, and can meet the requirements of the GNSS network RTK system for real-time high-precision positioning services.

[0028] 5. The data processing accuracy is improved The dynamic networking algorithm optimizes the baseline selection according to the quality of the observation data and geometric conditions of the reference stations, reduces redundant baselines, and improves the accuracy of baseline solution.

[0029] The error correction integration algorithm is based on multi-baseline data. Through weighted average or least squares optimization, it generates a more accurate error model, further improving the positioning accuracy at the user end.

[0030] 6. Optimization of Data Interaction and Resource Utilization Efficiency The message bus mechanism simplifies the data transmission process between the core server and the data processing server, improving the data interaction efficiency.

[0031] Task allocation is dynamically adjusted based on the resource utilization rate and load conditions of the processing server, ensuring uniform distribution of tasks and improving the resource utilization rate of the server.

[0032] 7. Enhancement of Adaptability and Flexibility The system can adjust the networking scheme and task allocation strategy according to the dynamic changes in the distribution of reference stations and the quality of observation data, adapting to the requirements of different scenarios.

[0033] Supports multiple task priority settings, and can give priority to high-priority tasks to meet the real-time requirements of specific scenarios (such as disaster warning).

[0034] 8. Practical Application Value The present invention is widely applicable to fields such as high-precision surveying and mapping, intelligent driving, disaster warning, precision agriculture, etc., and is of great significance for improving the service ability of the GNSS network RTK system.

[0035] The distributed processing mode reduces the costs of hardware upgrade and system maintenance, providing an economical and efficient solution for the popularization and application of large-scale GNSS networks.

[0036] Through the innovative distributed computing architecture and task scheduling mechanism, the present invention solves the technical bottlenecks in the aspects of computing efficiency, real-time performance, scalability, and reliability in the existing centralized processing and sub-network processing modes of GNSS network RTK systems, and has significant technical advantages and broad application prospects. Description of the Drawings

[0037] Figure 1 It is the flowchart of the embodiment of the present invention.

[0038] Figure 2 It is the structural diagram of the reference station network distributed data processing system in the embodiment of the present invention.

[0039] Figure 3 It is the data processing flowchart of the baseline solution server in the embodiment of the present invention.

[0040] Figure 4 It is the data processing flowchart of the error model establishment server in the embodiment of the present invention.

[0041] Figure 5This is the data processing flow chart of the core server in the embodiment of the present invention. Detailed implementation manners

[0042] The implementation process of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0043] In the GNSS network RTK system, the data processing modes are mainly divided into two types: centralized processing and processing in multiple subnets. These two methods have played a certain role in different-scale reference station networks, but when dealing with the requirements of modern large-scale GNSS networks, both have technical problems. The following analyzes from the core characteristics and limitations of the two methods.

[0044] In the GNSS network RTK system, the centralized processing mode relies on a single core server to uniformly process the data of the entire reference station network. All reference station observation data is uploaded to the core server for networking, baseline solution, and error correction number calculation. However, with the expansion of the scale of the reference station network, this mode exposes obvious calculation bottlenecks and poor scalability problems. The computing power of the core server is difficult to meet the real-time requirements of large-scale reference station networks. Especially when facing a large number of baseline solution and error model construction tasks, it is easy to cause task backlog and processing delay. In addition, the centralized transmission of all data to the core server will cause network bandwidth pressure, further exacerbating the system efficiency problem. At the same time, the design relying on a single server increases the risk of single-point failure.

[0045] The mode of processing in multiple subnets divides the entire reference station network into several subnets, and each subnet is responsible for processing by an independent server. This mode alleviates the computing pressure of a single server, but brings new problems. On the one hand, the data coordination between subnets is relatively complex, which is easy to lead to inconsistencies in error correction numbers and model calculations; on the other hand, the calculation results of each subnet need to be summarized to a global server, which still forms a new calculation bottleneck when the number of subnets is large. In addition, the subnet division is usually static and cannot be adjusted according to the dynamic distribution of reference stations and task requirements, resulting in insufficient flexibility and adaptability of the system.

[0046] Neither of the two modes can effectively meet the high computing requirements and real-time requirements of large-scale GNSS network RTK. In practical applications, centralized processing is difficult to expand, and processing in subnets has coordination and consistency problems. Therefore, a more efficient distributed processing architecture is needed to achieve dynamic allocation of computing tasks, load balancing, and network expansion capabilities to meet the development needs of modern GNSS network RTK systems.

[0047] An embodiment of the present invention proposes a large-scale reference station distributed data processing method for GNSS network RTK. Through distributed architecture design and task dynamic allocation mechanism, it solves the problems of computing bottleneck, low task scheduling efficiency, poor scalability, and complex coordination existing in the existing centralized processing and sub-network processing modes, and realizes the efficient processing of large-scale reference station network data and real-time error correction services. Specifically, it includes: core server reference station networking and task list generation; baseline solution server baseline solution and real-time resource occupancy assessment; error model establishment service for modeling the relative error of the master station and real-time resource occupancy assessment; dynamic allocation of baseline solution and error modeling. The connection methods between each step are as Figure 1 shown, and the structure of the reference station network distributed data processing system is as Figure 2 shown, the data processing flow of the baseline solution server is as Figure 3 shown, the data processing flow of the error model establishment server is as Figure 4 shown, and the reference station networking and task allocation flow of the core server is as Figure 5 shown.

[0048] As Figure 2 shown, the GNSS network RTK large-scale reference station distributed data processing system provided by the embodiment of the present invention includes: 1) The system consists of a core server, a baseline solution server, an error model establishment server, and a message bus.

[0049] 2) Core server: Responsible for reference station networking, task generation and allocation, integration of network-wide error correction numbers, task scheduling, and load balancing.

[0050] 3) Data processing server: Responsible for receiving tasks assigned by the core server, performing tasks such as baseline solution, etc., and feeding back the calculation results to the core server through the message bus.

[0051] 4) Error model establishment server: Responsible for receiving tasks assigned by the core server, performing tasks such as error correction number calculation and error model establishment, and feeding back the calculation results to the core server through the message bus.

[0052] 5) Message bus: Used for data transmission and task interaction between the core server and the data processing server to ensure the real-time and reliability of task allocation, data upload, and result feedback.

[0053] The data processing flow of the GNSS network RTK large-scale reference station distributed data processing system is as follows: 1) Reference station networking and task generation. The core server receives the observation data of reference stations through the message bus, and dynamically generates a baseline networking scheme for the entire network according to the geographical distribution, observation data quality, and geometric structure of the reference stations. The core server generates a task list for baseline solution and error model establishment.

[0054] 2) Dynamic task allocation. The core server obtains the real-time resource usage conditions (such as CPU occupancy rate, memory usage rate) of all data processing servers through the message bus. According to the task volume and the resource status of each server, the core server adopts a load balancing strategy to dynamically allocate baseline solution tasks and error model establishment tasks to different data processing servers.

[0055] 3) Baseline solution and calculation of error correction numbers. The baseline solution server receives the observation data of the baseline to be solved, independently completes the baseline solution task, extracts the baseline error correction numbers, and uploads the calculation results to the core server through the message bus.

[0056] 4) Error model establishment. The processing server obtains the relevant baseline correction numbers from the message bus according to the task, calculates the relative correction numbers of other reference stations relative to the local station, and constructs a regional error model, which is sent to the message bus for use by the data broadcast module.

[0057] 5) Distributed task scheduling and data interaction mechanism 51) Dynamic task scheduling: The core server dynamically allocates baseline solution and error model establishment tasks according to the priority of the tasks, the calculation complexity, and the real-time status of the processing servers (such as CPU usage rate, task load situation), ensuring an even distribution of tasks and avoiding overloading of computing resources.

[0058] 52) Efficient data interaction: Use the message bus to achieve data sharing and result interaction between the core server and the data processing servers. The message bus supports multi-threaded asynchronous communication, improves data transmission efficiency, and ensures the real-time nature of observation data upload, task allocation, and result feedback.

[0059] Based on the above GNSS network RTK large-scale reference station distributed data processing system, the specific implementation of the embodiments of the present invention includes the following steps: Step 1: Generation of reference station networking and task list by the core server All reference stations of the GNSS network RTK system form a triangular network, determine the baselines to be solved by the system, and generate a task list for baseline solution for task allocation to the solution server.

[0060] Step 1.1, Data reception of reference stations by the core server The core server receives the real-time observation data of all reference stations through the message bus, including coordinate information, signal strength, observation data quality indicators, etc., and determines the available reference stations.

[0061] Step 1.2, Denaunay triangle reference station networking Based on the available reference stations obtained in Step 1.1, optimize the number of baselines and geometric conditions based on Denaunay triangle dynamic group baselines to ensure the solution quality and efficiency.

[0062] Step 1.2.1, input point set. Input the coordinates of all reference stations to form a point set , refers to the horizontal and vertical coordinates of the i-th point.

[0063] Step 1.2.2, initialize the triangle. Select a super triangle from the point set that contains all points to cover the entire point set.

[0064] Step 1.2.3, insert points one by one. Insert the points into the current triangulation in turn (the result of initializing the triangle in 1.2.2 is the triangulation), and check whether the point is inside the circumcircle of a certain triangle. If it is satisfied, reconstruct the relevant triangle.

[0065] Step 1.2.4, edge exchange optimization. If an edge does not meet the Delaunay condition (that is, there are other points in the circumcircle), update the triangle structure after exchanging the edge.

[0066] In Delaunay triangulation, edge exchange optimization is a step used to ensure that the triangulation meets the Delaunay condition (i.e., the empty circle property). Specifically, the purpose of edge exchange is to adjust the edges of the triangle so that there are no other points inside the circumcircle of any triangle. The following is the specific process of edge exchange: Steps of edge exchange optimization: Judge whether it meets the Delaunay condition: For two adjacent triangles △ ABC and △ ABD , they share an edge AB .

[0067] Check whether the circumcircles of these two triangles contain other points. If there are other points inside the circumcircle, it does not meet the Delaunay condition.

[0068] (2) Exchange the edge: If it does not meet the Delaunay condition, then exchange the shared edge AB for the diagonalCD , that is, replace side AB with side CD .

[0069] After the exchange, the original two triangles △ ABC and △ ABD are replaced by the new two triangles △ ACD and △ BCD .

[0070] (3) Update the triangle structure: After exchanging the sides, update the triangle structure in the triangulation to ensure that the new triangles still satisfy the Delaunay condition.

[0071] If the new triangles still do not satisfy the Delaunay condition, further edge exchanges may be required. (4) Repeat the check: Check again whether the exchanged triangles satisfy the Delaunay condition until all triangles satisfy it.

[0072] Step 1.2.5, output the triangulation result. Remove the edges of the super triangle and output the remaining set of triangles and the set of baselines , where all the edges forming the triangles are the baselines.

[0073] Step 1.3, task list generation It is mainly divided into two types of tasks: 1) baseline solution 2) master station error model establishment. Take the set of baselines as the tasks to be solved and allocate them to the baseline solution server; take the set of available reference stations as the error modeling tasks and allocate them to the error model establishment server, including specific information such as whether to allocate the server and resource occupancy.

[0074] Step 2: Real-time resource occupancy assessment and baseline solution of the baseline processing server Step 2.1, real-time resource occupancy assessment of the baseline processing server Let the number of baselines processed by the server be , and assume that the CPU and memory resources occupied by each baseline solution are equal. Let the CPU occupancy rate of the baseline solution process in the current server be , and the memory occupancy rate be , then the CPU occupancy rate of a single baseline and the memory occupancy rate are respectively:

[0075]

[0076] Step 2.2: Baseline solution and relative error extraction of a single baseline Step 2.2.1: Coordinate and Observation Value Acquisition. The baseline solution server acquires the coordinate and observation value information of the baseline to be solved from the message bus.

[0077] Step 2.2.2: Forming Double Differences and Fixing Double Differences Ambiguity between Reference Stations. The baseline solution method first determines the BeiDou ultra-wide lane ambiguity based on the HMW combination method, then determines the wide lane ambiguity based on the TCAR method of the ionosphere-free combination, and finally determines the original ambiguities of B1I and B3I of all satellites based on the wide lane and dual-frequency ionosphere-free combined ambiguity.

[0078] Step 2.2.3: Baseline Error Extraction According to the fixed ambiguity and reference station coordinate information, etc., extract the error information related to and independent of the baseline geometry. According to the double-difference observation equations of L1 and L2 carriers:

[0079]

[0080] where , , 、 are the frequencies of carriers L1 and L2 respectively, 、 are the wavelengths of carriers L1 and L2 respectively, is the double-difference phase observation value on carriers L1 and L2, is the double-difference geometric distance, are the double-difference ambiguities of L1 and L2, is the double-difference tropospheric delay, is the double-difference ionospheric delay. After fixing the integer ambiguity between reference stations, the order of magnitude of the double-difference carrier phase observation noise is very small, so the observation noise can be ignored, and the double-difference ionospheric delay and double-difference tropospheric delay with centimeter-level accuracy at the reference station are obtained. Their estimation formulas are as follows:

[0081]

[0082] Step 3: Establishing the Relative Error Model of the Master Station and Evaluating Server Resource Occupancy Step 3.1: Evaluating the Real-time Resource Occupancy of the Processing Server Let the number of current sites for the error model establishment server to process relative error calculation be , and assume that the CPU and memory resources occupied by the relative error calculation of each site are equal. Let the CPU occupancy rate occupied by the master station error model establishment process in the current server, and the memory occupancy rate The CPU occupancy rate for establishing 1 master station error model and the memory occupancy rate are respectively as follows:

[0083]

[0084] Step 3.2: Calculation of the relative error of the master station Step 3.2.1: Obtaining the relative error value of the baseline. The solution server obtains the relative error information of the baselines related to the station to be solved and 10 surrounding stations from the message bus.

[0085] Step 3.2.2: Overall adjustment and solution of the relative error. The overall adjustment algorithm for the baseline error correction number is carried out in three steps: 1) Use the triangle network closure error test to judge whether there are gross errors; 2) Use the error correction numbers of each baseline to carry out the overall adjustment solution of the relative error correction number; 3) Based on the adjustment result, locate and repair the gross error and carry out the adjustment again.

[0086] In the first step, the closure error of the error correction numbers of each side baseline in the same triangle must theoretically be 0. Considering the existence of observation errors, its closure error should be within a certain extremely small threshold range. Therefore, if the closure error of the triangle network exceeds the limit, there must be at least one baseline error correction number in the triangle network that is incorrect. In the second step, according to the basic principle of indirect adjustment, the baseline vector adjustment method is used to solve the correction number for a specified master station. According to the relationship of double differences and baseline transmission, the following error equation describing the relationship of each baseline vector can be listed: (2) where are the error correction numbers of each baseline in the reference station triangle network, is the adjusted estimate of the double-difference relative error correction number from the master station to other reference stations, is the coefficient matrix describing the linear relationship between the error correction numbers of each baseline in the triangle network and the relative error correction number of the master station, is the residual of the error correction number.

[0087] Step 3.2.3: Establishment of the regional error model. Integrate the correction numbers of all master stations to construct a regional error model, provide high-precision error correction numbers for the GNSS network RTK rover, improve the positioning accuracy of the rover and shorten the initialization time, which is expressed by the interpolation method as:

[0088] where is the total number of slave stations, is the serial number of the slave station, is the slave station Correction number, is a slave station Interpolation coefficient of, is a slave station Coordinate value relative to the position to be interpolated, is the correction number of the position to be interpolated. Step 4, the core server calculates the corresponding remaining computing power according to the real-time resource occupancy evaluation of the baseline solution server and the server real-time resource occupancy evaluation of the error model establishment server respectively, and dynamically allocates the baseline solution task and the error modeling task according to the corresponding remaining computing power.

[0089] Let the memory and CUP occupancy rates of all baseline processing servers be and , considering the problem of server processing redundancy, the maximum memory and CUP occupancy of the server are both designed to be 80% (which can be set to other values according to actual needs), then the remaining available baseline number for processing (i.e., the remaining computing power of the baseline processing server) is:

[0090] Let the memory and CUP occupancy rates of all error model establishment servers be and , considering the problem of server processing redundancy, the maximum memory and CUP occupancy of the server are both designed to be 80%, then the remaining available baseline number for processing (i.e., the remaining computing power of the error model establishment server) is:

[0091] and Satisfy the following conditions: ,

[0092] The core server dynamically allocates the baseline solution task and the error modeling task according to and , that is, give priority to the task with a large number of processed baselines.

[0093] In specific implementation, the above process can be automatically run by computer software technology, and the system device running the method process of the present invention should also be within the protection scope of the present invention.

[0094] On the other hand, the embodiment of the present invention also provides GNSS network RTK reference station distributed data processing, including: A processor and a memory, the memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the GNSS network RTK reference station distributed data processing method as described in the above technical solution.

[0095] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Those skilled in the art to which the present invention pertains can make various modifications or supplements to the described specific embodiments or use similar means for substitution, but will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. GNSS network RTK reference station distributed data processing method, characterized in that, It includes the following steps: Step 1: The core server receives the real-time observation data of all reference stations through the message bus, and determines the available reference stations; based on the available reference stations, the baseline quantity and geometric conditions are optimized based on the Denaunay triangle dynamic group baselines to obtain a baseline set. Taking the baseline set as the task to be solved, it is assigned to the baseline solution server. Taking the set of available reference stations as the error modeling task, it is assigned to the error model establishment server. Step 2: The baseline solution server first calculates the real-time resource occupancy evaluation of the baseline solution server, and then performs baseline solution to calculate the double-difference ionospheric delay and the double-difference tropospheric delay. Step 3: The error model establishment server calculates the real-time resource occupancy evaluation of the error model establishment server, then solves all baselines in parallel, extracts the baseline error correction numbers, calculates the relative correction numbers of other stations relative to the local station, and thus establishes the master station error model. Step 4: The core server calculates the corresponding remaining computing capabilities respectively according to the real-time resource occupancy evaluation of the baseline solution server and the real-time resource occupancy evaluation of the error model establishment server, and dynamically allocates the baseline solution task and the error modeling task according to the corresponding remaining computing capabilities.

2. The GNSS network RTK reference station distributed data processing method according to claim 1, characterized in that: In Step 1, the specific implementation method for obtaining the baseline set is as follows: Step 1.2.1, input the coordinates of all reference stations to form a point set , refers to the horizontal and vertical coordinates of the i-th point; Step 1.2.2, initialize the triangle, select a super triangle that contains all the points from the point set to cover the entire point set; ​ Step 1.2.3, sequentially insert the points into the result of the current triangulation, i.e., the initialized triangle, and check whether the point is inside the circumcircle of a certain triangle. If so, reconstruct the relevant triangle; Step 1.2.4: If a certain side does not meet the Delaunay condition, that is, other points are included in the circumcircle, exchange the sides and update the triangle structure. Step 1.2.5: Remove the sides of the super triangle and output the remaining triangle set and baseline set.

3. The GNSS network RTK reference station distributed data processing method according to claim 1, characterized in that: In Step 2, the calculation method for the real-time resource occupancy evaluation of the baseline processing server is as follows: Set the number of baselines processed by the baseline processing server to , and the CPU and memory resources occupied by each baseline solution are equal. The CPU occupancy rate of the baseline solution process in the current server is , and the memory occupancy rate is . Then the CPU occupancy rate of a single baseline is and the memory occupancy rate is respectively as follows: 。 4. The GNSS network RTK reference station distributed data processing method according to claim 1, wherein: In Step 2, the calculation process of the double-difference ionospheric delay and the double-difference tropospheric delay is as follows: Step 2.2.1: The baseline solution server obtains the coordinate and observation value information of the baseline to be solved from the message bus. Step 2.2.2: Based on the HMW combination method, determine the Beidou ultra-wide lane ambiguity, then based on the TCAR method of the ionosphere-free combination, determine the wide lane ambiguity, and finally determine the original ambiguity of the satellite based on the wide lane and the double-frequency ionosphere-free combination ambiguity. Step 2.2.3: According to the fixed ambiguity and the reference station coordinate information, extract the error information related to and independent of the baseline geometry, and according to the L1 and L2 carrier double-difference observation equations: Among them , , 、 are the frequencies of carrier waves L1 and L2 respectively, 、 are the wavelengths of carrier waves L1 and L2 respectively, is the double-difference phase observation value on carrier waves L1 and L2, is the double-difference geometric distance, are the double-difference ambiguities of L1 and L2, is the double-difference tropospheric delay, is the double-difference ionospheric delay; finally, the double-difference ionospheric delay and double-difference tropospheric delay with centimeter-level accuracy at the reference station are obtained, and their estimation formulas are as follows: 。 5. The GNSS network RTK reference station distributed data processing method according to claim 1, characterized in that: In Step 3, the calculation method for the real-time resource occupancy evaluation of the error model establishment server is as follows: Set the current number of sites processed by the error model establishment server to , the CPU and memory resources occupied by the relative error calculation of each site are equal, and the CPU occupancy rate of the master station error model establishment process in the current error model establishment server , and the memory occupancy rate . Then the CPU occupancy rate of the master station error model establishment and the memory occupancy rate are respectively: 。 6. The GNSS network RTK reference station distributed data processing method according to claim 1, characterized in that: In Step 3, the specific implementation method for establishing the master station error model is as follows: Step 3.2.1: The error model establishment server obtains the relative error information of the baselines related to the station to be solved and several surrounding stations from the message bus. Step 3.2.2: Use the overall adjustment algorithm of the baseline error correction numbers to extract the baseline error correction numbers, which are carried out in three steps: 1) Use the triangular network closure difference test to judge whether there are gross errors; 2) Use the error correction numbers of each baseline to perform the overall adjustment calculation of the relative error correction numbers; 3) Based on the adjustment result, locate and repair the gross errors and perform adjustment again. Step 3.2.3: Integrate the baseline error correction numbers of all master stations to construct the master station error model, which is expressed by the interpolation method as: wherein is the total number of slave stations, is the serial number of the slave station, is the slave station correction, is the slave station interpolation coefficient, is the slave station coordinate value relative to the position to be interpolated, is the correction of the position to be interpolated.

7. The GNSS network RTK reference station distributed data processing method according to claim 6, characterized in that: In step 3.2.2 (1), the closure error of the baseline error corrections for each side within the same triangle is within the minimum threshold range. Therefore, if the closure error of the triangulation network exceeds the minimum threshold range, there must be at least one incorrect baseline error correction in the triangle network.

8. The GNSS network RTK reference station distributed data processing method according to claim 6, wherein: In step 3.2.2 (2), according to the basic principle of indirect adjustment, the baseline vector adjustment method is used to solve for the corrections with respect to a specified master station. According to the relationship of double differences and baseline transfer, the following error equations describing the relationships of each baseline vector are listed: （2） where is the error correction of each baseline in the reference station triangular network, is the adjusted estimate of the double-difference relative error correction from the master station to other reference stations, is the coefficient matrix describing the linear relationship between the error corrections of each baseline in the triangular network and the relative error correction of the master station, is the residual of the error correction.

9. The GNSS network RTK reference station distributed data processing method according to claim 1, wherein: In step 4, first calculate the remaining computing power of the baseline processing server and the remaining computing power of the error model establishment server; Set the memory and CPU occupancy rates of all baseline processing servers to and , with the maximum memory and CPU occupancy both being P%. Then the remaining computing power of the baseline processing server is: Among them, and are the CPU occupancy rate and memory occupancy rate of a single baseline respectively; Set the memory and CPU occupancy rates of the error model establishment server to be and , respectively. If the maximum memory and CPU occupancy are both P%, then the remaining computing power of the error model establishment server is: Among them, and are the CPU occupancy rate and memory occupancy rate established for one main station error model respectively; and and meet the following conditions: , The core server, according to and dynamically allocates baseline solution tasks and error modeling tasks, that is, it preferentially assigns tasks with large remaining computing capabilities for processing.

10. The distributed data processing device for the GNSS network RTK reference station is characterized in that, Including: A processor and a memory. The memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the GNSS network RTK reference station distributed data processing method according to any one of claims 1-9.

Images