Networking positioning processing method and device and related equipment

Abstract

The application provides a network positioning processing method and device and related equipment. The method is applied to a first reference station and includes: in the case where network positioning of the first reference station fails, obtaining first observation data streams of M second reference stations; based on the first observation data streams and a position spatial relationship, generating second observation data streams of a virtual reference station, the coordinate position of the virtual reference station being the same as the coordinate position of the first reference station, and the position spatial relationship being the position relationship of the second reference stations and the first reference station in the distance and direction dimensions; inputting the second observation data streams into a trained AI model to perform observation data prediction, obtaining third observation data streams of the virtual reference station, the AI model being used to generate data that is a virtual reference station of input data and meets the observation data statistical characteristics of the first reference station, and the third observation data streams being used to replace the observation data streams of the first reference station to perform network positioning.

Description

Technical Field

[0001] This application relates to the field of satellite navigation technology, and in particular to a network positioning processing method, apparatus and related equipment. Background Technology

[0002] With the rapid development of satellite navigation technology, high-precision positioning services have been widely used. For example, real-time kinematic (RTK) positioning services can rely on a network of base stations and use satellite navigation technology to achieve high-precision positioning.

[0003] Currently, when a base station is interrupted or of poor quality, preventing it from participating in the triangulation network, related technologies typically remove this base station from the network. The remaining base stations are then re-networked. To ensure service availability, base station interruptions can be addressed by encrypting the base stations or implementing master-slave failover, but these methods are costly. Furthermore, while related technologies can maintain service through network element reconstruction, data extrapolation, or reselecting a master station when a base station fails, they cannot solve the interruption problem in scenarios with insufficient base station density and no master-slave failover, thus restricting the continuity and accuracy of high-precision positioning services. Summary of the Invention

[0004] This application provides a network positioning processing method, apparatus, and related equipment, which can solve the technical problem in the related art where the failure of base station network positioning leads to the restriction of the continuity and accuracy of high-precision positioning services.

[0005] In a first aspect, embodiments of this application provide a network positioning processing method applied to a first base station, the method comprising:

[0006] If the first base station fails to network positioning, the first observation data stream of M second base stations is acquired. The distance between the second base station and the first base station is within a preset range. The second base station can participate in network positioning. M is an integer greater than 1.

[0007] Based on the first observation data stream and the spatial relationship, a second observation data stream of a virtual reference station is generated. The coordinate position of the virtual reference station is the same as that of the first reference station. The spatial relationship is the positional relationship between the second reference station and the first reference station in the distance and direction dimensions.

[0008] The second observation data stream is input into the successfully trained AI model to predict the observation data, thereby obtaining the third observation data stream of the virtual base station. The AI ​​model is used to generate data that takes the input data as the physical constraint of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. The third observation data stream is used to replace the observation data stream of the first base station for network positioning.

[0009] Secondly, embodiments of this application provide a network positioning processing device applied to a first base station, the device comprising:

[0010] The acquisition module is used to acquire the first observation data stream of M second reference stations when the first reference station fails to network positioning. The distance between the second reference stations and the first reference station is within a preset range, and the second reference stations can participate in network positioning. M is an integer greater than 1.

[0011] The generation module is used to generate a second observation data stream of a virtual reference station based on the first observation data stream and the spatial relationship between the virtual reference station and the first reference station. The coordinate position of the virtual reference station is the same as the coordinate position of the first reference station. The spatial relationship between the second reference station and the first reference station is the positional relationship between them in the distance and direction dimensions.

[0012] The prediction module is used to input the second observation data stream into the successfully trained AI model to predict the observation data and obtain the third observation data stream of the virtual base station. The AI ​​model is used to generate data that takes the input data as the physical constraints of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. The third observation data stream is used to replace the observation data stream of the first base station for network positioning.

[0013] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, it implements the steps of the network positioning processing method as described in the first aspect.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the network positioning processing method as described in the first aspect.

[0015] Fifthly, embodiments of this application provide a computer program product, including computer instructions that, when executed by a processor, implement the steps of the network positioning processing method as described in the first aspect.

[0016] In this embodiment, when the network positioning of the first reference station fails, a first observation data stream of M second reference stations is acquired. The distance between the second reference stations and the first reference station is within a preset range, and the second reference stations can participate in network positioning. M is an integer greater than 1. Based on the first observation data stream and the spatial relationship, a second observation data stream of a virtual reference station is generated. The coordinate position of the virtual reference station is the same as that of the first reference station. The spatial relationship is the positional relationship between the second reference station and the first reference station in the distance and direction dimensions. The second observation data stream is input into a successfully trained AI model to predict the observation data, resulting in a third observation data stream of the virtual reference station. The AI ​​model is used to generate data that uses the input data as the physical constraint of the virtual reference station and conforms to the statistical characteristics of the observation data of the first reference station. The third observation data stream is used to replace the observation data stream of the first reference station for network positioning. In this way, if the first base station fails to establish a network, a virtual base station can be generated in place at the offline base station location. By combining the generalization ability of the AI ​​model with the physical constraints of the virtual base station, the authenticity of the observation data of the generated virtual base station is guaranteed, as well as its rationality in the space environment. This avoids the situation where the observation data of the failed base station is completely interrupted, which would restrict the continuity and accuracy of high-precision positioning services, thereby improving the continuity and accuracy of high-precision positioning services. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of a scenario where a failed reference station S0 is networked.

[0019] Figure 2 This is a schematic diagram of the process of removing failed reference stations and re-networking in related technologies;

[0020] Figure 3 This is a flowchart of a network positioning processing method provided in an embodiment of this application;

[0021] Figure 4 This is a schematic diagram illustrating the selection of the second base station in an embodiment of this application;

[0022] Figure 5 This is a schematic flowchart of a specific example of a network positioning processing method provided in an embodiment of this application;

[0023] Figure 6 This is a schematic diagram illustrating the application of the network positioning processing method in the embodiments of this application;

[0024] Figure 7 This is a schematic diagram of the structure of a network positioning processing device provided in an embodiment of this application;

[0025] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] First, let's introduce the technical background of the embodiments of this application.

[0028] Network RTK technology: The server integrates observation data from all reference stations in the region and accurately models various spatial correlation errors (mainly ionospheric delay, tropospheric delay and satellite orbit error) in the region through carrier phase differential processing, generating high-precision differential corrections (such as RTCM format).

[0029] Triangular network topology: Base stations are typically deployed in a triangular mesh configuration. This topology ensures the spatial uniformity of the error model and the accuracy of interpolation. User terminals receive these corrections, enabling them to eliminate most errors in their observation data in real time during movement, thus obtaining centimeter-level high-precision positioning results. If data anomalies occur at a base station due to power outages, communication interruptions, obstructions, or other reasons, the base stations in that area will be re-networked to ensure normal service in that region.

[0030] Baseline length constraint: In network RTK processing, the distance between two reference stations is called the baseline. The correlation of spatially related errors (especially ionospheric delay) weakens as the distance between stations increases. To ensure the high accuracy and reliability of the generated regional error model, an upper limit threshold must be set for the baseline length (typically 70km-100km, depending on the region and technology).

[0031] Traditional triangulation models rely on the healthy operation of each reference station. If any reference station (S0) in the network fails due to power outage, communication interruption, physical damage, or equipment failure, the observation data for its location is lost. After detecting the failure of station S0, the system typically attempts "network reconfiguration," which automatically excludes S0 and tries to rebuild the triangulation using the remaining healthy reference stations. There must be a sufficient number of healthy reference stations, and their distribution must be even. In peripheral areas, mountainous regions, or coastal areas, the reference station density is inherently low, and S0 may be a critical node. After its failure, the remaining reference stations cannot form effective triangular coverage (i.e., becoming "orphaned stations" or causing network fragmentation), such as... Figure 1 As shown, the recombination algorithm will fail completely, causing a complete interruption of high-precision services in the region.

[0032] In related technologies, such as Figure 2 As shown, if a base station is interrupted or its quality is too poor to participate in the triangular network, it will be removed from the network, and the remaining base stations will be re-networked. Several scenarios will arise during this re-networking process:

[0033] 1. When the base stations are sufficiently dense, after the network is reassembled, the broadcast differential data may be interrupted for 5-30 seconds while continuing to cover the area.

[0034] 2. After the network is reorganized, the remaining base station baselines will be too long, resulting in some areas not being covered.

[0035] To ensure that services are not interrupted in the region due to base station failures and to improve the overall availability of services, related technologies usually involve increasing the density of base stations or using a dual base station mode for primary / backup switching. However, increasing the density of base stations or using a dual base station mode for primary / backup switching will increase the cost of building and maintaining base stations.

[0036] Based on this, this application provides a new network positioning processing method, the core of which is to use real-time observation data around the failed reference station to generate observation data of the failed reference station, and to fuse it with the statistical features of the observation data of the failed reference station learned by the AI ​​model as physical constraints. In this way, the observation data of the virtual reference station is given unique features of the failed reference station. Thus, a virtual reference station can be generated on the spot at the location of the offline reference station, avoiding the complete interruption of the observation data of the failed reference station.

[0037] See Figure 3 , Figure 3 This is a flowchart of a network positioning processing method provided in an embodiment of this application, such as... Figure 3 As shown, the method includes the following steps:

[0038] Step 301: If the first base station fails to network positioning, acquire the first observation data stream of M second base stations. The distance between the second base stations and the first base station is within a preset range. The second base stations can participate in network positioning. M is an integer greater than 1.

[0039] Step 302: Based on the first observation data stream and the spatial relationship, generate a second observation data stream for a virtual reference station. The coordinate position of the virtual reference station is the same as that of the first reference station. The spatial relationship is the positional relationship between the second reference station and the first reference station in the distance and direction dimensions.

[0040] Step 303: Input the second observation data stream into the successfully trained AI model to predict the observation data and obtain the third observation data stream of the virtual base station. The AI ​​model is used to generate data that takes the input data as the physical constraint of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. The third observation data stream is used to replace the observation data stream of the first base station for network positioning.

[0041] It should be noted that the embodiments of this application involve technologies such as satellite navigation, high-precision positioning, spatial model algorithms, and AI-based data-driven technologies.

[0042] In step 301, when the data interruption or abnormal health status of the first base station S0 is detected, it can be determined that the first base station has failed to form a network and cannot participate in the network.

[0043] In some embodiments, if the first base station detects a network positioning failure, the first base station can broadcast a message indicating that its network positioning has failed. Correspondingly, the second base station in its vicinity can receive the message and send its observation data stream to the first base station, so that the first base station can obtain the observation data stream of the second base station in its vicinity.

[0044] In some embodiments, the first base station can receive observation data streams from the surrounding second base stations in real time. Correspondingly, if the first base station fails to establish a network positioning system, it can acquire observation data streams from the surrounding second base stations.

[0045] The M second reference stations can have at least two azimuths. There can be one or more second reference stations in a single azimuth. A second reference station can be the base station closest to the first reference station among all reference stations in its azimuth. The preset range indicates that the second reference stations are relatively close to the first reference station, and are reference stations surrounding the first reference station.

[0046] In some embodiments, M is 4, and the locations of the M second reference stations satisfy preset conditions; wherein...

[0047] The preset conditions include:

[0048] Centered on the first base station, the M second base stations are located in four different directions from the first base station;

[0049] For any given second reference station, the second reference station is the reference station that is closest to the first reference station among all reference stations in the azimuth of the second reference station.

[0050] In other words, with the first reference station as the center, the first reference station can receive in real time observation data streams from its nearest healthy second reference stations (identified as S1, S2, S3, and S4 in its four cardinal directions), such as... Figure 4 As shown. In this way, the spatial relationship between the first and second reference stations can be fully utilized, and the observation data stream of the virtual reference station can be optimally interpolated based on the observation data stream of the second reference station.

[0051] The first reference station can receive real-time observation data streams from the four nearest healthy second reference stations (S1, S2, S3, S4) in the east, south, west, and north directions. This data stream undergoes preprocessing, such as gross error detection, data cleaning, and integrity verification. Afterward, the preprocessed data stream is formatted and aligned; for example, all observations from the second reference stations are standardized to the same timestamp and sampling rate and mapped to the same satellite list, forming a standardized observation matrix, thus obtaining the first observation data stream.

[0052] The first observation data stream may include the time of data reception, the relative position of the base station and the visible satellite, etc.

[0053] In step 302, the virtual reference station can refer to a reference station in the same location as the first reference station. The spatial relationship can refer to the positional relationship between the second reference station and the first reference station in the distance and direction dimensions. For example, the second reference station S1 is located to the east of the first reference station and is 70 meters away.

[0054] In some embodiments, a second observation data stream of a virtual reference station can be generated based on the first observation data stream and the spatial relationship between the locations, using a physical spatial model. The physical spatial model can refer to simulating the data of unknown points using data from known points and the spatial relationship between the known and unknown points.

[0055] In some embodiments, the physical space model can be a spatial delay model based on Kriging interpolation, which treats the ionospheric delay (Iono) and tropospheric delay (Tropo) of each satellite as regional spatial random fields and optimally interpolates the delay values ​​of unknown points (such as the failed reference station S0) using observation data from known points (such as the second reference station). The physical space model can also be other models, which are not specifically limited here.

[0056] In some embodiments, the physical space model can be a spatial delay model based on Kriging interpolation, and step 302 specifically includes:

[0057] Based on the first observation data stream, the first delay information for the satellite signals received by the M second reference stations is determined;

[0058] Based on the first delay information and the spatial relationship, a variogram model is generated. The variogram model is used to characterize the variation of the delay information of the satellite signal received by the reference station with the distance between reference stations.

[0059] Based on the variogram model and the spatial relationship, the weight information of the M second base stations is determined. The weight of the second base station is related to the target spatial location of the second base station. The target spatial location includes the distance and direction of the second base station relative to the first base station.

[0060] The second delay information of the virtual reference station receiving satellite signals is obtained by weighting the first delay information and the weight information. The second observation data stream includes the second delay information.

[0061] The first delay information can be ionospheric delay (Iono) and / or tropospheric delay (Tropo).

[0062] In some embodiments, the first delay information of the satellite signals received by the M second reference stations can be directly determined based on the first observation data stream.

[0063] In some embodiments, determining the first delay information of the satellite signals received by the M second reference stations based on the first observation data stream includes:

[0064] Based on the first observation data stream, determine the single-difference observation values ​​among the M second reference stations;

[0065] Based on the single-difference observations, the first delay information for the satellite signals received by the M second reference stations is determined.

[0066] The specific process of generating the second observation data stream of the virtual reference station using a physical space model based on spatial delay modeling using Kriging interpolation can be as follows:

[0067] Step 1: Extract the single-difference observations between stations based on the first observation data stream.

[0068] Using a certain healthy second reference station as a reference, the single difference between it and other second reference stations is calculated to eliminate common errors such as satellite clock errors, improve the accuracy of delay information determination, and thus improve the accuracy of the second observation data stream of the virtual reference station generated by interpolation.

[0069] Step 2: Calculate the delay.

[0070] The ionospheric delay can be accurately extracted using the geometry-free combination of dual-frequency observations, L4 = L1 - L2. The dry tropospheric delay can be calculated using the Saastamoinen or Hopfield models, and the wet delay component estimated, thus obtaining the tropospheric delay. The first delay information can include both the ionospheric delay and the tropospheric delay.

[0071] Step 3: Variogram Modeling.

[0072] This is the core of Kriging, which can analyze the spatial correlation of delay within a region based on location and spatial relationships, in order to fit a variogram model (such as an exponential model or a Gaussian model). This variogram model describes the law of delay variation with distance and orientation between stations. For example, the second reference station has a greater impact on the observation data of the first reference station S0 in the eastern orientation. The farther away the second reference station is from the first reference station S0, the smaller the impact of the second reference station on the observation data of the first reference station S0.

[0073] Step 4: Best linear unbiased estimate (BLUE).

[0074] A weight λ_i can be calculated for the delay of each healthy second reference station based on the variogram model. The weight depends on the distance and direction between the second reference station and the failed reference station S0. The estimated delay Î0 at the failed reference station S0 is obtained by weighted averaging of the delay of the second reference station and the weight, such as Î0 = Σ (λ_i * I_i), where i ranges from 1 to 4, representing different second reference stations. Here, I_i represents the delay of the second reference station, and the weight λ_i is obtained by solving the Kriging equations, i.e., the variogram model, ensuring that the variance of the estimation error is minimized. Î0 represents the observation data stream of the virtual reference station optimally interpolated by the physical space model based on Kriging interpolation for spatial delay modeling.

[0075] Accordingly, for each common-view satellite, the ionospheric delay estimate (Iono_est) and tropospheric delay estimate (Tropo_est) at the coordinates of the failed reference station S0 can be output. The second observation data stream can include the ionospheric delay estimate (Iono_est) and the tropospheric delay estimate (Tropo_est).

[0076] In step 303, the statistical characteristics of the observed data may include the root mean square (RMS) of the multipath error, the mean signal-to-noise ratio, and the cycle slip frequency.

[0077] Among them, the AI ​​model can be a deep learning model. The core task of the data-driven (i.e., second observation data stream V_physical driven) AI model is to learn the unique observation characteristics of the failed reference station S0 (such as multipath error, antenna phase center change, local environmental interference, etc.) and generate a third observation data stream that is highly consistent with the physical constraint V_physical, but also has the observation characteristics of the failed reference station S0 itself. This third observation data stream is a virtually generated observation data stream.

[0078] The AI ​​model can be trained using historical observation data from the first base station. In other words, as long as the first base station can still participate in network positioning normally, it can obtain its observation data stream for training the AI ​​model, enabling the AI ​​model to learn the unique observation characteristics of the first base station.

[0079] like Figure 5 As shown, in this embodiment, when the first base station group cannot participate in network positioning, a virtual base station S00 of the first base station is generated by the real-time observation data streams of the surrounding second base stations (second base stations S1, S2, S3, and S4 in the four directions of east, west, south, and north respectively). Specifically, after preprocessing the observation data to obtain the first observation data stream, a second observation data stream of the virtual base station S00 is generated based on the first observation data stream through a preset physical space model. Then, the second observation data stream is passed through a data-driven model, i.e., an AI model. This AI model can add the unique characteristics of the failed base station S0, i.e., observation features, to the second observation data stream of the virtual base station S00 to obtain a third observation data stream that uses the second observation data stream as the physical constraint of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. Then, the third observation data stream replaces the observation data stream of the failed first base station S0 and participates in normal network calculation.

[0080] This embodiment employs a virtual-real fusion approach, using AI to learn the historical behavior of failed reference stations. It incorporates real-time observation data from healthy reference stations surrounding the failed station and fuses this data with a physical spatial model (such as an ionospheric / tropospheric delay model) and a data-driven model (AI) to generate the most probable and physically consistent observation data at the coordinates of the failed base station. This method combines the generalization ability of AI with the physical constraints of the virtual reference station, ensuring both the authenticity of the generated data and its rationality within the spatial environment. It also avoids the elimination of identical data points during subsequent network calculations.

[0081] In this embodiment, a physical constraint and AI fusion mechanism is used. Real-time observation data from multiple surrounding second reference stations are used as physical constraints and fused with the AI ​​model to generate observation data for a virtual reference station in situ at the failed reference station. Then, the failed reference station S0 is switched to the virtual reference station for network positioning calculation. The core of this embodiment is "switching," that is, switching from one real data source to another pre-calculated data source. This can be achieved by generating a virtual reference station in situ "out of nothing" when there is no real observation data from the failed reference station. When there are no real reference stations to switch to (i.e., network reconstruction fails), the virtual reference station can be effectively used for network positioning calculation.

[0082] The embodiments of this application can be applied to ground-based RTK services, such as... Figure 6 As shown, when the service detects a data interruption or abnormal health status of the base station S0, it directly switches to the virtual base station S00 to participate in the network positioning calculation in place of S0. In other words, when the base station participating in the ground-based RTK service network experiences a data interruption or is in poor health and unable to participate in the network, the virtual base station can be used to replace the base station in the network to achieve ground-based RTK service. Specifically, when the service detects a data interruption or abnormal health status of the base station S0, it directly switches to the virtual base station S00 to participate in the network calculation in place of the failed base station S0.

[0083] This improves the availability of ground-based RTK services and ensures overall service availability. After a base station outage or abnormal health check, especially in remote areas, the response time for base station maintenance is often unsatisfactory and costly. Furthermore, if issues cannot be resolved quickly, such as base station replacement or relocation, it can lead to prolonged service disruptions in the area. This application's embodiments provide a solution to improve the availability of ground-based RTK services by reducing base station maintenance response time and costs without affecting service availability. This solution does not rely on increasing base station density or using dual-site systems, thus reducing the costs of base station construction and maintenance.

[0084] The network architecture of the AI ​​model can be a Conditional Generative Adversarial Network (CGAN).

[0085] In some embodiments, the AI ​​model includes a feature encoder and a generator, and step 303 specifically includes:

[0086] The second observation data stream and the feature vector of the feature encoder are input into the generator to predict the observation data, thereby obtaining the third observation data stream of the virtual base station;

[0087] The feature vector is obtained by the feature encoder based on the historical observation data of the first base station and is used to characterize the statistical features of the observation data of the first base station. The generator is used to generate data that takes the second observation data stream as the physical constraint of the virtual base station and conforms to the statistical features of the observation data of the first base station.

[0088] The input to the generator (G) can include:

[0089] Physical constraints, i.e., the second observation data stream V_physical;

[0090] Historical Feature Vector: A latent vector z that encodes statistical features of long-term historical observation data from the first base station S0 (such as RMS of multipath error, mean signal-to-noise ratio, and cycle slip frequency). This feature vector z is extracted using a pre-trained feature encoder (autoencoder).

[0091] Auxiliary information: real-time satellite elevation angle and azimuth angle.

[0092] The generator can be structured using fully connected (dense layers) or temporal convolutional networks (TCNs) as the core structure to adapt to the temporal characteristics of the observed data sequence.

[0093] The generator's output may include: a virtual observation data stream V_ai.

[0094] AI models can be pre-trained using long-term historical data from the first base station S0 (which can include observation data under various space weather conditions) to obtain high-quality feature vectors z.

[0095] In this embodiment, the second observation data stream and the feature vector of the feature encoder can be input into the generator to predict the observation data. The generator can generate data that is physically constrained by the second observation data stream and conforms to the statistical characteristics of the observation data of the first base station, thus obtaining the third observation data stream.

[0096] In some embodiments, the AI ​​model further includes a discriminator. The generator and the discriminator are trained based on a target loss, which includes a physical consistency loss and an adversarial loss. The physical consistency loss is used to characterize the difference between the data generated by the generator and the input data of the AI ​​model on preset key indicators. The adversarial loss is used to characterize whether the data generated by the generator conforms to the statistical characteristics of the observation data of the first base station. The discriminator determines whether the data generated by the generator is physically constrained by the input data of the AI ​​model. If the target loss is less than a preset threshold, the AI ​​model is successfully trained.

[0097] The discriminator (D) can take either real data (which may come from the historical database of the first base station S0) or generated data V_ai, along with conditional information V_physical. Its task is to determine the authenticity of the data, specifically whether the data is reasonable under the given physical conditions V_physical. For example, even if the data appears perfect, if it deviates significantly from the physical constraints V_physical, the discriminator should classify it as "false." The discriminator can be structured using a convolutional neural network (CNN) to extract deep features from the data.

[0098] In some embodiments, the generator and discriminator can be trained using adversarial loss. In some embodiments, the loss function design may include adversarial loss and physical consistency loss. Adversarial loss, denoted by `Loss_adv`, ensures that the generated data distribution approximates the real data distribution, while physical consistency loss, denoted by `Loss_physical`, ensures that the generator generates data that conforms to physical constraints. The objective loss can be represented as `Loss_total = Loss_adv(G, D) + λ * Loss_physical(G)`. The physical consistency loss `Loss_physical` can be obtained by calculating the difference between the generated data `V_ai` and the physical constraint `V_physical` on preset key metrics. For example, the mean squared error (MSE) of the two geometrically uncombined L4 can be calculated, denoted as Loss_physical = MSE( L4_ai, L4_physical ), where L4_ai represents the geometrically uncombined L4 of the generated data V_ai, and L4_physical represents the geometrically uncombined L4 of the physical constraint V_physical.

[0099] In the objective loss, λ is a hyperparameter used to strictly control the strength to which the generator must adhere to physical constraints. A larger value for λ results in more physically reliable generated data, but may sacrifice some learned site characteristics.

[0100] During training, the feature encoder is pre-trained using long-term historical data from the first base station S0 to obtain high-quality feature vectors z.

[0101] Adversarial training extracts a real sample V_real and its corresponding physical constraint V_physical from the dataset (which can be obtained by simulation in historical data).

[0102] The generator G receives V_physical and z, and outputs V_ai.

[0103] Discriminator D compares (V_real, V_physical) and (V_ai, V_physical) respectively to determine the authenticity of the generated data.

[0104] The generator G and discriminator D can be alternately optimized based on backpropagation of the target loss. When the Nash equilibrium is reached, that is, the V_ai generated by the generator G is so that the discriminator D cannot distinguish between true and false, and at the same time, it is highly consistent with V_physical, that is, the AI ​​model is successfully trained.

[0105] If the AI ​​model is successfully trained, the physical space model generates V_physical in real time. The trained generator G immediately starts working, using V_physical and the feature vector z representing the statistical characteristics of the observation data of the first base station as constraints, and instantly outputs the final high-fidelity virtual observation value V_final, which is the third observation data stream, to participate in the network positioning solution.

[0106] The embodiments of this application have broad commercial value.

[0107] It can be widely used in traditional foundation projects, and can integrate network positioning processing methods into edge computing devices or base station gateways, and can provide software and hardware upgrade services for existing base station networks.

[0108] It can provide Virtual Site as a Service (VRSaaS) based on a cloud platform, offering services to regional Cross-Origin Resource Sharing (CORS) operators based on the virtual site's activation duration or data processing volume.

[0109] It can automatically respond to faults, significantly reducing the number of times and costs that need to be spent sending technicians to remote sites for emergency repairs.

[0110] During the network planning phase, the extreme requirements for base station density can be appropriately relaxed, because at critical nodes, virtual stations can partially replace the functions of physical stations, thereby reducing the overall network construction cost.

[0111] "Service availability" is the ultimate competitive barrier for high-precision positioning services. The embodiments of this application can help create a high-availability brand image of "five nines (99.999%)".

[0112] See Figure 7 , Figure 7 This is a schematic diagram of a network positioning processing device provided in an embodiment of this application, applied to a first base station, such as... Figure 7 As shown, the network positioning processing device 300 includes:

[0113] The acquisition module 701 is used to acquire the first observation data stream of M second reference stations when the first reference station fails to network positioning. The distance between the second reference stations and the first reference station is within a preset range, and the second reference stations can participate in network positioning. M is an integer greater than 1.

[0114] The generation module 702 is used to generate a second observation data stream of a virtual reference station based on the first observation data stream and the spatial relationship of the location. The coordinate position of the virtual reference station is the same as the coordinate position of the first reference station. The spatial relationship of the location is the positional relationship between the second reference station and the first reference station in the distance and direction dimensions.

[0115] The prediction module 703 is used to input the second observation data stream into the successfully trained AI model to predict the observation data and obtain the third observation data stream of the virtual base station. The AI ​​model is used to generate data that takes the input data as the physical constraint of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. The third observation data stream is used to replace the observation data stream of the first base station for network positioning.

[0116] Optionally, the generation module 702 is specifically used for:

[0117] Based on the first observation data stream, the first delay information for the satellite signals received by the M second reference stations is determined;

[0118] Based on the first delay information and the spatial relationship, a variogram model is generated. The variogram model is used to characterize the variation of the delay information of the satellite signal received by the reference station with the distance between reference stations.

[0119] Based on the variogram model and the spatial relationship, the weight information of the M second base stations is determined. The weight of the second base station is related to the target spatial location of the second base station. The target spatial location includes the distance and direction of the second base station relative to the first base station.

[0120] The second delay information of the virtual reference station receiving satellite signals is obtained by weighting the first delay information and the weight information. The second observation data stream includes the second delay information.

[0121] Optionally, the generation module 702 is further configured to:

[0122] Based on the first observation data stream, determine the single-difference observation values ​​among the M second reference stations;

[0123] Based on the single-difference observations, the first delay information for the satellite signals received by the M second reference stations is determined.

[0124] Optionally, the AI ​​model includes a feature encoder and a generator, and the prediction module 703 is specifically used for:

[0125] The second observation data stream and the feature vector of the feature encoder are input into the generator to predict the observation data, thereby obtaining the third observation data stream of the virtual base station;

[0126] The feature vector is obtained by the feature encoder based on the historical observation data of the first base station and is used to characterize the statistical features of the observation data of the first base station. The generator is used to generate data that takes the second observation data stream as the physical constraint of the virtual base station and conforms to the statistical features of the observation data of the first base station.

[0127] Optionally, the AI ​​model further includes a discriminator. The generator and the discriminator are trained based on a target loss, which includes a physical consistency loss and an adversarial loss. The physical consistency loss is used to characterize the difference between the data generated by the generator and the input data of the AI ​​model on preset key indicators. The adversarial loss is used to characterize whether the data generated by the generator conforms to the statistical characteristics of the observation data of the first base station. The discriminator judges whether the data generated by the generator is physically constrained by the input data of the AI ​​model. If the target loss is less than a preset threshold, the AI ​​model is successfully trained.

[0128] Optionally, M is 4, and the locations of the M second reference stations satisfy preset conditions; wherein,

[0129] The preset conditions include:

[0130] Centered on the first base station, the M second base stations are located in four different directions from the first base station;

[0131] For any given second reference station, the second reference station is the reference station that is closest to the first reference station among all reference stations in the azimuth of the second reference station.

[0132] The network positioning processing device 700 can implement all the processes implemented in the above-described network positioning processing method embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0133] See Figure 8 The figure shows a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. It is applied to a first base station, such as... Figure 8 As shown, the electronic device 800 includes: a processor 801, a memory 802, a user interface 803, and a bus interface 804.

[0134] Processor 801 is used to read the program from memory 802 and execute the following procedures:

[0135] If the first base station fails to network positioning, the first observation data stream of M second base stations is acquired. The distance between the second base station and the first base station is within a preset range. The second base station can participate in network positioning. M is an integer greater than 1.

[0136] Based on the first observation data stream and the spatial relationship, a second observation data stream of a virtual reference station is generated. The coordinate position of the virtual reference station is the same as that of the first reference station. The spatial relationship is the positional relationship between the second reference station and the first reference station in the distance and direction dimensions.

[0137] The second observation data stream is input into the successfully trained AI model to predict the observation data, thereby obtaining the third observation data stream of the virtual base station. The AI ​​model is used to generate data that takes the input data as the physical constraint of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. The third observation data stream is used to replace the observation data stream of the first base station for network positioning.

[0138] exist Figure 8 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 801 and memory represented by memory 802 together. The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 804 provides an interface. For different user devices, user interface 803 can also be an interface capable of connecting external or internal devices, including but not limited to keypads, displays, speakers, microphones, joysticks, etc.

[0139] The processor 801 is responsible for managing the bus architecture and general processing, while the memory 802 can store the data used by the processor 801 when performing operations.

[0140] In some embodiments, the processor 801 is further configured to:

[0141] Based on the first observation data stream, the first delay information for the satellite signals received by the M second reference stations is determined;

[0142] Based on the first delay information and the spatial relationship, a variogram model is generated. The variogram model is used to characterize the variation of the delay information of the satellite signal received by the reference station with the distance between reference stations.

[0143] Based on the variogram model and the spatial relationship, the weight information of the M second base stations is determined. The weight of the second base station is related to the target spatial location of the second base station. The target spatial location includes the distance and direction of the second base station relative to the first base station.

[0144] The second delay information of the virtual reference station receiving satellite signals is obtained by weighting the first delay information and the weight information. The second observation data stream includes the second delay information.

[0145] In some embodiments, the processor 801 is further configured to:

[0146] Based on the first observation data stream, determine the single-difference observation values ​​among the M second reference stations;

[0147] Based on the single-difference observations, the first delay information for the satellite signals received by the M second reference stations is determined.

[0148] In some embodiments, the AI ​​model includes a feature encoder and a generator, and the processor 801 is further configured to:

[0149] The second observation data stream and the feature vector of the feature encoder are input into the generator to predict the observation data, thereby obtaining the third observation data stream of the virtual base station;

[0150] The feature vector is obtained by the feature encoder based on the historical observation data of the first base station and is used to characterize the statistical features of the observation data of the first base station. The generator is used to generate data that takes the second observation data stream as the physical constraint of the virtual base station and conforms to the statistical features of the observation data of the first base station.

[0151] In some embodiments, the AI ​​model further includes a discriminator. The generator and the discriminator are trained based on a target loss, which includes a physical consistency loss and an adversarial loss. The physical consistency loss is used to characterize the difference between the data generated by the generator and the input data of the AI ​​model on preset key indicators. The adversarial loss is used to characterize whether the data generated by the generator conforms to the statistical characteristics of the observation data of the first base station. The discriminator determines whether the data generated by the generator is physically constrained by the input data of the AI ​​model. If the target loss is less than a preset threshold, the AI ​​model is successfully trained.

[0152] In some embodiments, M is 4, and the locations of the M second reference stations satisfy preset conditions; wherein...

[0153] The preset conditions include:

[0154] Centered on the first base station, the M second base stations are located in four different directions from the first base station;

[0155] For any given second reference station, the second reference station is the reference station that is closest to the first reference station among all reference stations in the azimuth of the second reference station.

[0156] Preferably, the present invention also provides an electronic device 800, including a processor 801, a memory 802, and a computer program stored in the memory 802 and executable on the processor 801. When the computer program is executed by the processor 801, it implements the various processes of the above-described network positioning processing method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0157] This invention also provides a readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the above-described network positioning processing method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here. The readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0158] This application also provides a computer program product, including computer instructions. When executed by a processor, the computer instructions implement the various processes of the above-described network positioning processing method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0159] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0160] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0161] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0162] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0163] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0164] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0165] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A network positioning processing method, characterized in that, Applied to a first reference station, the method includes: If the first base station fails to network positioning, the first observation data stream of M second base stations is acquired. The distance between the second base station and the first base station is within a preset range. The second base station can participate in network positioning. M is an integer greater than 1. Based on the first observation data stream and the spatial relationship, a second observation data stream of a virtual reference station is generated. The coordinate position of the virtual reference station is the same as that of the first reference station. The spatial relationship is the positional relationship between the second reference station and the first reference station in the distance and direction dimensions. The second observation data stream is input into the successfully trained AI model to predict the observation data, thereby obtaining the third observation data stream of the virtual base station. The AI ​​model is used to generate data that takes the input data as the physical constraint of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. The third observation data stream is used to replace the observation data stream of the first base station for network positioning.

2. The method according to claim 1, characterized in that, The process of generating a second observation data stream for a virtual reference station based on the first observation data stream and the spatial relationship between locations includes: Based on the first observation data stream, the first delay information for the satellite signals received by the M second reference stations is determined; Based on the first delay information and the spatial relationship between the locations, a variogram model is generated. The variogram model is used to characterize the variation of the delay information of the satellite signal received by the reference station with the distance between the reference stations. Based on the variogram model and the spatial relationship, the weight information of the M second reference stations is determined. The weight of the second reference station is related to the target spatial position of the second reference station. The target spatial position includes the distance and direction of the second reference station relative to the first reference station. The second delay information of the virtual reference station receiving satellite signals is obtained by weighting the first delay information and the weight information. The second observation data stream includes the second delay information.

3. The method according to claim 2, characterized in that, The determination of the first delay information for the M second reference stations receiving satellite signals based on the first observation data stream includes: Based on the first observation data stream, determine the single-difference observation values ​​among the M second reference stations; Based on the single-difference observations, the first delay information for the satellite signals received by the M second reference stations is determined.

4. The method according to claim 1, characterized in that, The AI ​​model includes a feature encoder and a generator. The step of inputting the second observation data stream into the AI ​​model to predict the observation data and obtain the third observation data stream of the virtual base station includes: The second observation data stream and the feature vector of the feature encoder are input into the generator to predict the observation data, thereby obtaining the third observation data stream of the virtual base station; The feature vector is obtained by the feature encoder based on the historical observation data of the first base station and is used to characterize the statistical features of the observation data of the first base station. The generator is used to generate data that takes the second observation data stream as the physical constraint of the virtual base station and conforms to the statistical features of the observation data of the first base station.

5. The method according to claim 4, characterized in that, The AI ​​model also includes a discriminator. The generator and the discriminator are trained based on target loss, which includes physical consistency loss and adversarial loss. The physical consistency loss is used to characterize the difference between the data generated by the generator and the input data of the AI ​​model on preset key indicators. The adversarial loss is used to characterize whether the data generated by the generator conforms to the statistical characteristics of the observation data of the first base station. The discriminator judges whether the data generated by the generator is physically constrained by the input data of the AI ​​model. If the target loss is less than a preset threshold, the AI ​​model is successfully trained.

6. The method according to claim 1, characterized in that, M is 4, and the locations of the M second reference stations satisfy preset conditions; wherein... The preset conditions include: Centered on the first base station, the M second base stations are located in four different directions from the first base station; For any given second reference station, the second reference station is the reference station that is closest to the first reference station among all reference stations in the azimuth of the second reference station.

7. A network positioning processing device, characterized in that, Applied to a first reference station, the device includes: The acquisition module is used to acquire the first observation data stream of M second reference stations when the first reference station fails to network positioning. The distance between the second reference stations and the first reference station is within a preset range, and the second reference stations can participate in network positioning. M is an integer greater than 1. The generation module is used to generate a second observation data stream of a virtual reference station based on the first observation data stream and the spatial relationship between the virtual reference station and the first reference station. The coordinate position of the virtual reference station is the same as the coordinate position of the first reference station. The spatial relationship between the second reference station and the first reference station is the positional relationship between them in the distance and direction dimensions. The prediction module is used to input the second observation data stream into the successfully trained AI model to predict the observation data and obtain the third observation data stream of the virtual base station. The AI ​​model is used to generate data that takes the input data as the physical constraints of the virtual base station and conforms to the statistical characteristics of the observation data of the first base station. The third observation data stream is used to replace the observation data stream of the first base station for network positioning.

8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the network positioning processing method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the network positioning processing method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the network positioning processing method as described in any one of claims 1-6.

Images