Low-bandwidth link-oriented NRTK differential data optimization method and system

Abstract

The invention discloses an NRTK differential data optimization method and system for a low-bandwidth link, and the method comprises the following steps: obtaining a visible satellite signal, and packaging the visible satellite signal into an RTCM differential message protocol frame; performing satellite-by-satellite information content scoring on the RTCM differential message protocol frames, retaining observation data of Top-k optimal satellites, and re-framing to form simplified frames; performing single-point positioning calculation on the simplified frame to obtain a coarse positioning result of the edge node; triggering to generate a virtual base station request; a virtual base station is generated in the range, close to the short base line, of the coarse positioning point, and the virtual base station generates differential correction data; and the differential correction data and the simplified frame are fused, extended Kalman filtering is executed, a least square ambiguity decorrelation adjustment algorithm is carried out, and a final positioning result is output. The problem that traditional NRTK positioning depends on a high-bandwidth link technology in non-edge solution is effectively solved, and the accuracy and stability of a low-cost positioning technology are improved.

Description

Technical Field

[0001] This invention relates to the field of differential positioning data processing technology, specifically to an NRTK differential data optimization method and system for low-bandwidth links. Background Technology

[0002] The BeiDou Navigation Satellite System already possesses real-time centimeter-level dynamic positioning capabilities. By continuously broadcasting differential correction information to field users through a ground-based augmentation network, millimeter-level displacement monitoring can be achieved. Existing ground-based differential networks can also provide network differential data download functionality, enabling network real-time dynamic differential positioning (NRTK positioning). However, the raw observation data from satellites typically grows linearly with the number of visible satellites, resulting in a relatively large data volume. In long-term deformation monitoring in remote locations such as power transmission towers, slopes, dilapidated buildings, and bridges, due to cost and power consumption limitations, on-site data transmission often relies on narrowband wireless links such as LoRa, NB-IoT, and BLE Long-Range. These links have low air speeds and small single-packet payloads. Directly transmitting traditional observation data packets can easily lead to transmission queuing, packet retransmission, or even data loss, causing data from the observation station to fail to reach the terminal in a timely manner, making it difficult to fix integer ambiguity, and resulting in drifting positioning results.

[0003] To alleviate bandwidth pressure, simplification methods such as thinning visible satellites, reducing precision bits, or enabling general lossless compression are commonly used. However, while reducing bandwidth, this often sacrifices satellite geometry or increases processor computational burden, ultimately resulting in longer initial fixed-time and reduced reacquisition capability. Other solutions replace the original observations with state domain parameters, which can significantly reduce byte length but requires the terminal to support new service protocols, increasing hardware upgrade costs. Therefore, how to enable differential data to successfully traverse narrowband links while maintaining centimeter-level positioning accuracy and rapid reacquisition performance without changing existing low-cost single-frequency or multi-frequency RTK module hardware or increasing power consumption has become a major obstacle restricting the large-scale application of BeiDou high-precision IoT monitoring. Summary of the Invention

[0004] To overcome the defects and shortcomings of existing technologies, this invention provides an NRTK differential data optimization method and system for low-bandwidth links. This method effectively solves the dependence of traditional NRTK positioning on high-bandwidth link technology in non-edge calculations. Through the designed data filtering and platform-side calculation optimization methods, NRTK technology can be widely applied to non-edge calculation scenarios with low cost and low power consumption requirements, thereby improving the accuracy and stability of low-cost positioning technology.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] This invention provides an NRTK differential data optimization method for low-bandwidth links, comprising the following steps:

[0007] Acquire visible satellite signals and encapsulate them into RTCM differential message protocol frames;

[0008] The information content of each satellite is scored for the RTCM differential message protocol frame, the observation data of the top-k best satellites are retained and re-framed to form a simplified frame;

[0009] Perform single-point localization calculation on the simplified frame to obtain coarse localization results for edge nodes;

[0010] Trigger a request to generate a virtual reference station;

[0011] A virtual reference station is generated within the range of the short baseline near the coarse positioning point, and differential correction data is generated from the virtual reference station.

[0012] The differential correction data and simplified frames are fused, extended Kalman filtering is performed, and a least-squares ambiguity decorrelation adjustment algorithm is applied to output the final localization result.

[0013] As a preferred technical solution, visible satellite signals are acquired. Specifically, visible satellite signals are captured through an RTK module and a satellite communication antenna. The RTK module completes satellite signal reception, demodulation, and spread spectrum recovery, and outputs raw observation data in RTCM3.3 format.

[0014] As a preferred technical solution, the information content of RTCM differential message protocol frames is scored on a satellite-by-satellite basis, specifically including:

[0015] At the single-star level, it is divided into signal-to-noise ratio score and elevation angle score;

[0016] The signal-to-noise ratio score is expressed as:

[0017] ;

[0018] in, This represents the signal-to-noise ratio of a single satellite. This represents the minimum signal-to-noise ratio of all observable satellites. This represents the maximum signal-to-noise ratio of all observable satellites;

[0019] The elevation angle rating is expressed as follows:

[0020] ;

[0021] in, This indicates the elevation angle relationship between the satellite and the initial positioning.

[0022] The final basic rating at the single-star level is expressed as follows:

[0023] ;

[0024] in, This represents the basic rating at the single-star level. and It is the weighting of signal-to-noise ratio and elevation angle. and These are the signal-to-noise ratio and elevation angle score of a single star, respectively.

[0025] At the multi-star level, the geometric contribution is obtained from the contribution of a single star to the satellite geometry, and the geometric contribution is expressed as:

[0026] ;

[0027] ;

[0028] ;

[0029] in, Indicates the geometric contribution of a single star. This represents a satellite direction vector used in the process of calculating the average direction vector. Represents the set of all observed satellites. This represents the direction vector of a single star. Indicates the altitude angle of a single star. Indicates the azimuth of a single star. This represents the average direction vector of all observable satellites;

[0030] The final information content scoring formula is:

[0031] ;

[0032] in, Indicates the weight.

[0033] As a preferred technical solution, the wireless link adopts the LoRa-mesh communication network.

[0034] As a preferred technical solution, differential correction data and simplified frames are fused, extended Kalman filtering is performed, and a least-squares ambiguity decorrelation adjustment algorithm is applied to output the final localization result, specifically including:

[0035] Before performing extended Kalman filtering, time window constraint matching is performed to construct a time window-based storage space to store the base station data;

[0036] When the raw observation data arrives, the system searches for the same epoch within the time window of the stored data. If the same epoch is found, the differential correction data issued by the corresponding virtual base station is used. If not found, the differential data of the base station with the smallest epoch difference is used.

[0037] The propagation state covariance during short-term link interruption is specifically represented as follows:

[0038] ;

[0039] ;

[0040] in, It is the posterior state estimate at time k. It is the posterior state estimate at time k-1. It is the posterior state estimation covariance matrix at time k. It is the posterior state estimation covariance matrix at time k-1. It describes model incompleteness and unmodeled disturbances, and is used to characterize the accumulation of uncertainty during interruptions.

[0041] The present invention also provides an NRTK differential data optimization system for low-bandwidth links, comprising: edge nodes, platform, and network CORS center;

[0042] Edge nodes acquire visible satellite signals and encapsulate them into RTCM differential message protocol frames;

[0043] Edge nodes score the information content of each satellite in the RTCM differential message protocol frame, retain the observation data of the top-k best satellites and reassemble the frames to form a simplified frame;

[0044] The simplified frame is transmitted through the narrowband antenna of the edge node, and after fading through the wireless link, it reaches the platform in a single-hop or multi-hop manner to achieve low-bandwidth backhaul.

[0045] The platform performs single-point localization calculations on the simplified frame to obtain coarse localization results for the edge nodes;

[0046] The platform uploads the coarse positioning results to the network CORS center, triggering a request to generate a virtual base station;

[0047] The network CORS center generates a virtual reference station within the short baseline range near the coarse positioning point. The virtual reference station generates differential correction data and transmits the corresponding differential correction data back to the platform.

[0048] The platform integrates differential correction data and simplified frames, performs extended Kalman filtering, and applies a least-squares ambiguity decorrelation adjustment algorithm to output the final positioning result.

[0049] As a preferred technical solution, the edge node acquires visible satellite signals. Specifically, it captures visible satellite signals through an RTK module and a satellite communication antenna. The RTK module completes satellite signal reception, demodulation, and spread spectrum recovery, and outputs raw observation data in RTCM3.3 format.

[0050] As a preferred technical solution, the information content of RTCM differential message protocol frames is scored on a satellite-by-satellite basis, specifically including:

[0051] At the single-star level, it is divided into signal-to-noise ratio score and elevation angle score;

[0052] The signal-to-noise ratio score is expressed as:

[0053] ;

[0054] in, This represents the signal-to-noise ratio of a single satellite. This represents the minimum signal-to-noise ratio of all observable satellites. This represents the maximum signal-to-noise ratio of all observable satellites;

[0055] The elevation angle rating is expressed as follows:

[0056] ;

[0057] in, This indicates the elevation angle relationship between the satellite and the initial positioning.

[0058] The final basic rating at the single-star level is expressed as follows:

[0059] ;

[0060] in, This represents the basic rating at the single-star level. and It is the weighting of signal-to-noise ratio and elevation angle. and These are the signal-to-noise ratio and elevation angle score of a single star, respectively.

[0061] At the multi-star level, the geometric contribution is obtained from the contribution of a single star to the satellite geometry, and the geometric contribution is expressed as:

[0062] ;

[0063] ;

[0064] ;

[0065] in, Indicates the geometric contribution of a single star. This represents a satellite direction vector used in the process of calculating the average direction vector. Represents the set of all observed satellites. This represents the direction vector of a single star. Indicates the altitude angle of a single star. Indicates the azimuth of a single star. This represents the average direction vector of all observable satellites;

[0066] The final information content scoring formula is:

[0067] ;

[0068] in, Indicates the weight.

[0069] As a preferred technical solution, the wireless link adopts the LoRa-mesh communication network.

[0070] As a preferred technical solution, the platform integrates differential correction data and simplified frames, performs extended Kalman filtering, and applies a least-squares ambiguity decorrelation adjustment algorithm to output the final localization result, specifically including:

[0071] Before performing extended Kalman filtering, time window constraint matching is performed to construct a time window-based storage space to store the base station data;

[0072] When the raw observation data arrives, the system searches for the same epoch within the time window of the stored data. If the same epoch is found, the differential correction data issued by the corresponding virtual base station is used. If not found, the differential data of the base station with the smallest epoch difference is used.

[0073] The propagation state covariance during short-term link interruption is specifically represented as follows:

[0074] ;

[0075] ;

[0076] in, It is the posterior state estimate at time k. It is the posterior state estimate at time k-1. It is the posterior state estimation covariance matrix at time k. It is the posterior state estimation covariance matrix at time k-1. It describes model incompleteness and unmodeled disturbances, and is used to characterize the accumulation of uncertainty during interruptions.

[0077] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0078] This invention breaks down the high-bandwidth link and edge computing terminal into a two-stage architecture: the terminal selects only satellites, and the platform performs the computation. The edge node uses a low-cost RTK module to collect and filter satellites in real time, generating byte-level simplified frames and transmitting them back via a narrowband link. On the platform side, it uses a first-in-first-out time window to match virtual reference station data, and then runs extended Kalman filtering and least-squares ambiguity decorrelation adjustment algorithms. At the same time, it uses the model noise Q to naturally expand the covariance during short-term interruptions, and can continuously output centimeter-level results without additional hardware. Overall, it eliminates the need for a terminal computing chip and has the three major characteristics of low power consumption, low cost, and high availability, enabling NRTK to be truly deployed in battery-powered narrowband IoT monitoring scenarios such as poles, slopes, and dilapidated buildings. Attached Figure Description

[0079] Figure 1 This is a flowchart illustrating the NRTK differential data optimization method for low-bandwidth links according to the present invention.

[0080] Figure 2 This is a comparison chart of the mean standard deviation of localization corresponding to different k values ​​under the topk filtering method of this invention;

[0081] Figure 3 This is a schematic diagram of the overall architecture of the NRTK differential data optimization system for low-bandwidth links according to the present invention. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0083] Example 1

[0084] like Figure 1 As shown, this embodiment provides an NRTK differential data optimization method for low-bandwidth links. This method effectively solves the dependence of traditional NRTK positioning on high-bandwidth link technology in non-edge calculations. Through designed data filtering and platform-side calculation optimization methods, it enables NRTK technology to be widely applied to non-edge calculation scenarios with low-cost and low-power requirements, improving the accuracy and stability of low-cost positioning technology. The specific steps include the following:

[0085] Step 1: Edge nodes capture visible satellite signals through their onboard RTK modules and satellite communication antennas, and encapsulate the raw observation information into differential GPS data format standard (RTCM differential message protocol frame) to complete local data acquisition;

[0086] In this embodiment, the RTK module only performs satellite signal reception, demodulation and spread spectrum recovery, and outputs raw observation data in RTCM3.3 format. The terminal does not perform any positioning calculations, thus saving expensive calculation chips and additional power consumption.

[0087] In this embodiment, the RTK module adopts a single-frequency satellite signal receiving mode. Although multi-frequency signals can bring extreme improvement in positioning accuracy, they will also increase the burden on the communication link. However, single-frequency network RTK technology can already achieve centimeter-level positioning accuracy. Therefore, using a single-frequency signal module with better communication cost-effectiveness can better maintain the stability of the solution link.

[0088] Step 2: The microprocessor of the edge node scores the information content of each satellite in the RTCM differential message protocol frame, and executes a fast selection strategy based on the score to retain the observation data of the top-k best satellites and reassemble the frames to form a simplified frame.

[0089] In this embodiment, satellites are ranked by calculating the information content score of each satellite, and a Top-k selection method (k can be dynamically adjusted) is used to further select a suitable satellite constellation. The quality of the raw satellite observation data is related to the signal quality of individual satellites and the spatial geometric distribution of satellites. Therefore, the designed information content scoring method is studied from both single-satellite and multi-satellite levels. Furthermore, considering the scenario of edge nodes with limited computing resources and the need for rapid satellite selection, the information content score of a satellite consists of the following components:

[0090] At the single-satellite level, it is divided into signal-to-noise ratio score and elevation angle score;

[0091] The signal-to-noise ratio (SNR) scoring formula is:

[0092] ;

[0093] in, The signal-to-noise ratio of a single satellite can be obtained from the raw RTCM3 data; This represents the minimum signal-to-noise ratio of all satellites that can be observed by this edge node; This represents the maximum signal-to-noise ratio of all satellites that the edge node can observe.

[0094] The formula for scoring the elevation angle is:

[0095] ;

[0096] in, This represents the elevation angle relationship between the satellite and the initial positioning. The higher the elevation angle of the satellite, the fewer the line-of-sight obstacles between it and the node, and the higher the signal quality. The initial positioning can be obtained from historical positioning results, directly from NMEA data, or by selecting all satellites and obtaining the initial positioning after completing one positioning.

[0097] The final base rating for a single star is:

[0098] ;

[0099] in, Represents the basic rating at the single-star level; and The weighting of signal-to-noise ratio and elevation angle should satisfy the standard that the sum is 1. and These are the signal-to-noise ratio and elevation angle score of a single satellite, respectively.

[0100] At the multi-star level, the geometric contribution of a single star to the satellite geometry is obtained, and the formula for geometric contribution is:

[0101] ;

[0102] ;

[0103] ;

[0104] in, Represents the direction vector of a single star. Represents the altitude angle of a single star. The azimuth of a single star. This represents the average direction vector of all satellites that can be observed by the edge node. This means the collection of all observation satellites. This represents a satellite's direction vector used in the process of calculating the average direction vector. Represents the geometric contribution of a single star.

[0105] The final information content scoring formula is:

[0106] ;

[0107] like Figure 2 As shown, the information content score of each observable satellite can be obtained and ranked according to the information content scoring formula. Satellites are selected from high to low according to the Top-k strategy, where k can be dynamically adjusted. Figure 3 The relationship between k and satellite positioning accuracy measurement shows that as k increases, the mean standard deviation of positioning accuracy in the three directions of east, north, and sky continuously decreases. When k reaches 14 to 17, the standard deviation drops to its lowest point. However, as the number of satellites k increases, the standard deviation will rebound. This is because satellites with poor signal quality may be selected, thus interfering with the solution quality.

[0108] Taking k=15 as an example, when the base address is selected by fast filtering, for the MSM4 type message of RTCM, the amount of raw observation data can be compressed to 160 bytes, which is in line with the single packet communication capacity of narrowband communication technology.

[0109] Step 3: The compressed simplified frame is transmitted through the node's narrowband antenna. After fading through the wireless link, it arrives at the platform in a single-hop or multi-hop manner to achieve low-bandwidth backhaul.

[0110] In this embodiment, the communication link used is a LoRa-mesh communication network. LoRa falls exactly at the optimal cut-off point of RTK in the bandwidth-power consumption-coverage triangle. However, the power consumption problem of NB-IoT technology does not match the low-power scenario discussed. The transmission distance and transmission performance of BLELong-Range greatly affect the scalability of the network. Therefore, LoRa communication, which has better performance in low power consumption and long distance communication, is chosen for narrowband communication. In addition, with the mesh self-organizing network, the coverage of the network can be extended through multi-hop and other mechanisms, thus expanding the application scenarios of the system.

[0111] Step 4: The platform uses the received simplified frame to perform single-point localization calculation to obtain coarse localization results of edge nodes, providing initial coordinate values ​​for subsequent differential calculation;

[0112] Step 5: The platform uploads the coarse positioning results to the continuously operating reference station network center (network CORS center) via Ethernet, triggering a request to generate a virtual reference station;

[0113] Step 6: The network CORS center generates a virtual reference station within the short baseline range near the coarse positioning point. The virtual reference station generates differential correction data and transmits the corresponding differential correction data back to the platform via Ethernet to complete the preparation of the differential data source.

[0114] Step 7: The platform integrates the differential correction data sent by CORS with the compressed simplified frame received in Step 3, runs extended Kalman filtering and performs least squares ambiguity decorrelation adjustment algorithm (LAMBDA integer ambiguity is fixed), and outputs the final centimeter-level positioning result for monitoring operations.

[0115] In this embodiment, the differential correction data sent by the base station in step 7 and the original observation data transmitted back from the edge nodes may be misaligned at epochs. In this scenario, the base station data is transmitted via Ethernet faster than the original observation data via a narrowband wireless link. Therefore, when the original observation data arrives at the platform's processing center, the epoch it represents will be behind the current epoch of the base station differential data. Therefore, a time-constrained matching-based RTK differential timeliness optimization method is adopted. This method performs time window constraint matching before the extended Kalman filter is applied at the platform. The platform designs a time window-based storage space, continuously updating and sliding the storage window to store base station data. When the original observation data arrives, it searches for the same epoch within the stored data's time window. If the same epoch is found, the differential correction data sent by the corresponding virtual base station is used directly; otherwise, the base station differential data with the smallest epoch difference is used. Furthermore, the rover's data may contain old data packets due to wireless link issues; these should be discarded to ensure the convergence speed of the positioning.

[0116] In this embodiment, the platform's solution end additionally employs a short-term link interruption compensation algorithm based on state preservation. During RTK positioning, a short-term communication link interruption will cause observations to become unavailable, thus preventing the Kalman filter from performing measurement updates. If the filter state and its covariance are simply frozen during this period, the uncertainty of the true state will be underestimated, and statistical mismatch will occur after the link is restored, thereby affecting the stability of observation acceptance and ambiguity fixation.

[0117] A state-space model-based filtering consistency-preserving method is introduced during link interruptions: under observation-free conditions, only time updates are performed, and the natural growth of uncertainty is characterized by propagating the state covariance without introducing any virtual observations or predictive measurements. When communication is restored, the filter can accept new observations with reasonable statistics, thereby achieving a smooth recovery of the solution state and ambiguity estimates.

[0118] The update formula for Kalman filtering is:

[0119] ;

[0120] ;

[0121] ;

[0122] in, For the prior state estimation at step k, This is the posterior state estimate after observation correction. It is the actual observed vector. For the observation model function, It is its predicted output of the prior state; It is Kalman gain. and Let these represent the prior and posterior estimated covariance matrices, respectively. It is the observation model in Jacobian matrix at the location, It is the identity matrix. To observe the noise covariance matrix;

[0123] When a short-term link is interrupted, the traditional solution method freezes the parameters of the Kalman filter until the latest solvable data arrives. However, by this time, a period of time has passed, and the uncertainty of the filter's state prediction should theoretically increase. But if the parameters of the Kalman filter are frozen, it means that the uncertainty of the filter's state prediction remains unchanged. Therefore, it is necessary to propagate the state covariance when the link is interrupted to characterize the natural growth of uncertainty without introducing any virtual observations or predictive measurements.

[0124] The formula for when the link is interrupted is:

[0125] ;

[0126] ;

[0127] in, It is the posterior state estimate at time k. It is the posterior state estimate at time k-1. It is the posterior state estimation covariance matrix at time k. It is the posterior state estimation covariance matrix at time k-1. It describes model incompleteness and unmodeled disturbances, and is used to characterize the accumulation of uncertainty during interruptions.

[0128] like Figure 3 As shown, this embodiment provides an NRTK differential data optimization system for low-bandwidth links, which uses a narrowband communication network of LoRa-mesh networking as the wireless link medium, including: edge nodes, platform end, and network CORS center;

[0129] Edge nodes acquire visible satellite signals and encapsulate them into RTCM differential message protocol frames. Specifically, edge nodes acquire visible satellite signals through their onboard RTK module and satellite communication antenna, and encapsulate the raw observation information into differential GPS data format standard (RTCM differential message protocol frames) to complete local data acquisition.

[0130] In this embodiment, the RTK module collects BeiDou single-frequency observation data, only performs satellite signal reception, demodulation and spread spectrum recovery, and outputs raw observation data in RTCM3.3 format. The terminal does not perform any positioning calculations, thus saving expensive calculation chips and additional power consumption. The MCU is used for the control logic execution of the edge nodes, the solar cells are used to maintain the long survivability of the device, and the LoRa antenna encodes, modulates and transmits the processed satellite data through the onboard LoRa module.

[0131] Edge nodes score the information content of each satellite in the RTCM differential message protocol frame, retain the observation data of the top-k best satellites and reassemble the frames to form a simplified frame;

[0132] The simplified frame is transmitted through the narrowband antenna of the edge node, and after fading through the wireless link, it reaches the platform in a single-hop or multi-hop manner to achieve low-bandwidth backhaul.

[0133] The platform performs single-point localization calculations on the simplified frame to obtain coarse localization results for the edge nodes;

[0134] The platform uploads the coarse positioning results to the network CORS center, triggering a request to generate a virtual base station;

[0135] The network CORS center generates a virtual reference station within the short baseline range near the coarse positioning point. The virtual reference station generates differential correction data and transmits the corresponding differential correction data back to the platform.

[0136] The platform integrates differential correction data and simplified frames, performs extended Kalman filtering, and applies a least-squares ambiguity decorrelation adjustment algorithm to output the final positioning result.

[0137] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for optimizing NRTK differential data for low-bandwidth links, characterized in that, Includes the following steps: Acquire visible satellite signals and encapsulate them into RTCM differential message protocol frames; The information content of each satellite is scored for the RTCM differential message protocol frame, the observation data of the top-k best satellites are retained and re-framed to form a simplified frame; Perform single-point localization calculation on the simplified frame to obtain coarse localization results for edge nodes; Trigger a request to generate a virtual reference station; A virtual reference station is generated within the range of the short baseline near the coarse positioning point, and differential correction data is generated from the virtual reference station. The differential correction data and simplified frames are fused, extended Kalman filtering is performed, and a least-squares ambiguity decorrelation adjustment algorithm is applied to output the final localization result.

2. The NRTK differential data optimization method for low-bandwidth links according to claim 1, characterized in that, The system acquires visible satellite signals, specifically by using an RTK module and a satellite communication antenna to capture these signals. The RTK module performs satellite signal reception, demodulation, and spread spectrum recovery, and outputs raw observation data in RTCM3.3 format.

3. The NRTK differential data optimization method for low-bandwidth links according to claim 1, characterized in that, The information content of RTCM differential message protocol frames is scored on a satellite-by-satellite basis, specifically including: At the single-star level, it is divided into signal-to-noise ratio score and elevation angle score; The signal-to-noise ratio score is expressed as: ； in, This represents the signal-to-noise ratio of a single satellite. This represents the minimum signal-to-noise ratio of all observable satellites. This represents the maximum signal-to-noise ratio of all observable satellites; The elevation angle rating is expressed as follows: ； in, This indicates the elevation angle relationship between the satellite and the initial positioning. The final basic rating at the single-star level is expressed as follows: ； in, This represents the basic rating at the single-star level. and It is the weighting of signal-to-noise ratio and elevation angle. and These are the signal-to-noise ratio and elevation angle score of a single star, respectively. At the multi-star level, the geometric contribution is obtained from the contribution of a single star to the satellite geometry, and the geometric contribution is expressed as: ； ； ； in, Indicates the geometric contribution of a single star. This represents a satellite direction vector used in the process of calculating the average direction vector. Represents the set of all observed satellites. This represents the direction vector of a single star. Indicates the altitude angle of a single star. Indicates the azimuth of a single star. This represents the average direction vector of all observable satellites; The final information content scoring formula is: ； in, Indicates the weight.

4. The NRTK differential data optimization method for low-bandwidth links according to claim 1, characterized in that, The wireless link uses a LoRa-mesh communication network.

5. The NRTK differential data optimization method for low-bandwidth links according to claim 1, characterized in that, By fusing differentially corrected data and simplified frames, extended Kalman filtering is performed, followed by a least-squares ambiguity decorrelation adjustment algorithm, resulting in the final localization result, which includes: Before performing extended Kalman filtering, time window constraint matching is performed to construct a time window-based storage space to store the base station data; When the raw observation data arrives, the system searches for the same epoch within the time window of the stored data. If the same epoch is found, the differential correction data issued by the corresponding virtual base station is used. If not found, the differential data of the base station with the smallest epoch difference is used. The propagation state covariance during short-term link interruption is specifically represented as follows: ； ； in, It is the posterior state estimate at time k. It is the posterior state estimate at time k-1. It is the posterior state estimation covariance matrix at time k. It is the posterior state estimation covariance matrix at time k-1. It describes model incompleteness and unmodeled disturbances, and is used to characterize the accumulation of uncertainty during interruptions.

6. An NRTK differential data optimization system for low-bandwidth links, characterized in that, include: Edge nodes, platform, and network CORS center; Edge nodes acquire visible satellite signals and encapsulate them into RTCM differential message protocol frames; Edge nodes score the information content of each satellite in the RTCM differential message protocol frame, retain the observation data of the top-k best satellites and reassemble the frames to form a simplified frame; The simplified frame is transmitted through the narrowband antenna of the edge node, and after fading through the wireless link, it reaches the platform in a single-hop or multi-hop manner to achieve low-bandwidth backhaul. The platform performs single-point localization calculations on the simplified frame to obtain coarse localization results for the edge nodes; The platform uploads the coarse positioning results to the network CORS center, triggering a request to generate a virtual base station; The network CORS center generates a virtual reference station within the short baseline range near the coarse positioning point. The virtual reference station generates differential correction data and transmits the corresponding differential correction data back to the platform. The platform integrates differential correction data and simplified frames, performs extended Kalman filtering, and applies a least-squares ambiguity decorrelation adjustment algorithm to output the final positioning result.

7. The NRTK differential data optimization system for low-bandwidth links according to claim 6, characterized in that, Edge nodes acquire visible satellite signals, specifically through RTK modules and satellite communication antennas. The RTK module completes satellite signal reception, demodulation, and spread spectrum recovery, and outputs raw observation data in RTCM3.3 format.

8. The NRTK differential data optimization system for low-bandwidth links according to claim 6, characterized in that, The information content of RTCM differential message protocol frames is scored on a satellite-by-satellite basis, specifically including: At the single-star level, it is divided into signal-to-noise ratio score and elevation angle score; The signal-to-noise ratio score is expressed as: ； in, This represents the signal-to-noise ratio of a single satellite. This represents the minimum signal-to-noise ratio of all observable satellites. This represents the maximum signal-to-noise ratio of all observable satellites; The elevation angle rating is expressed as follows: ； in, This indicates the elevation angle relationship between the satellite and the initial positioning. The final basic rating at the single-star level is expressed as follows: ； in, This represents the basic rating at the single-star level. and It is the weighting of signal-to-noise ratio and elevation angle. and These are the signal-to-noise ratio and elevation angle score of a single star, respectively. At the multi-star level, the geometric contribution is obtained from the contribution of a single star to the satellite geometry, and the geometric contribution is expressed as: ； ； ； in, Indicates the geometric contribution of a single star. This represents a satellite direction vector used in the process of calculating the average direction vector. Represents the set of all observed satellites. This represents the direction vector of a single star. Indicates the altitude angle of a single star. Indicates the azimuth of a single star. This represents the average direction vector of all observable satellites; The final information content scoring formula is: ； in, Indicates the weight.

9. The NRTK differential data optimization system for low-bandwidth links according to claim 6, characterized in that, The wireless link uses a LoRa-mesh communication network.

10. The NRTK differential data optimization system for low-bandwidth links according to claim 6, characterized in that, The platform integrates differential correction data and simplified frames, performs extended Kalman filtering, and applies a least-squares ambiguity decorrelation adjustment algorithm to output the final localization result, which includes: Before performing extended Kalman filtering, time window constraint matching is performed to construct a time window-based storage space to store the base station data; When the raw observation data arrives, the system searches for the same epoch within the time window of the stored data. If the same epoch is found, the differential correction data issued by the corresponding virtual base station is used. If not found, the differential data of the base station with the smallest epoch difference is used. The propagation state covariance during short-term link interruption is specifically represented as follows: ； ； in, It is the posterior state estimate at time k. It is the posterior state estimate at time k-1. It is the posterior state estimation covariance matrix at time k. It is the posterior state estimation covariance matrix at time k-1. It describes model incompleteness and unmodeled disturbances, and is used to characterize the accumulation of uncertainty during interruptions.

Images