6G base station satellite flash and real-time federated learning combined wireless positioning method and system
The 6G base station star-flash with real-time federated learning enhances UWB positioning accuracy in co-site scenarios by using a collaborative filtering model and iterative learning to address environmental and shared base station issues, improving geometric and error-based evaluation.
Patent Information
- Application Number
- CN202411207545.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-12-31
AI Technical Summary
Existing wireless positioning systems using ultra-wideband (UWB) signals face inaccuracies due to shared base stations, weather, and environmental factors, particularly in co-site scenarios where multiple operators share a single base station, leading to issues like multipath effects, non-line-of-sight propagation, and clock synchronization errors.
A 6G base station star-flash combined with real-time federated learning, employing a collaborative filtering model to predict feature aggregation, data tagging based on indoor/outdoor locations, and using geometric accuracy, mean square error, and root mean square error algorithms to construct an accuracy evaluation model, with iterative updates through real-time federated learning.
Improves positioning accuracy by separately evaluating indoor and outdoor locations, minimizing errors through real-time data learning and updating, addressing factors like multipath effects, non-line-of-sight propagation, and clock synchronization.
Smart Images
Figure CN119233393A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wireless star-flash technology, specifically to a wireless positioning method and system combining 6G base station star-flash with real-time federated learning. Background Technology
[0002] Ultra-wideband (UWB) is one of the major breakthroughs in wireless communication in recent years. Because UWB signals are highly sparse in the time domain, it's natural to consider introducing signal acquisition control (CS) into UWB systems to improve signal acquisition performance. A major breakthrough in UWB technology was the invention of micropower pulse radar (MIR). It was the first technology to operate UWB signals with inexpensive hardware at very low power. This addresses the problem of inaccurate location assessment in signal receiver positioning when multiple operators share a single base station, where the location is easily affected by weather and environmental factors. Summary of the Invention
[0003] The purpose of this invention is to provide a 6G base station star-flash combined real-time federated learning wireless positioning method and system, which improves the accuracy of wireless positioning by combining 6G base station star-flash combined real-time federated learning.
[0004] This invention is achieved through the following technical solution:
[0005] The first aspect of this application provides a 6G base station star-flash combined with real-time federated learning wireless positioning method for wireless positioning in co-site application scenarios formed by multiple operators sharing base stations, including the following steps:
[0006] For real-time UWB positioning data in co-location application scenarios, a collaborative filtering model is used to predict feature aggregation. The prediction results are then sorted to obtain the key factors affecting UWB positioning accuracy.
[0007] Data aggregation of real-time UWB positioning data is performed with key factors affecting UWB positioning accuracy as the center point, and the aggregated data is marked as indoor or outdoor positioning tags according to the location where the log data is generated.
[0008] The aggregated data is evaluated for accuracy at different locations based on whether the location tag is indoor or outdoor: the geometric accuracy factor positioning performance is used as the evaluation index for outdoor locations in co-located application scenarios, and the mean square error and root mean square error algorithms are used as the evaluation index for indoor locations in co-located application scenarios, thus constructing an accuracy evaluation model.
[0009] The accuracy evaluation model is learned by using real-time UWB positioning data through a real-time federated learning process, and the calculation results are updated through model learning and iterative updates.
[0010] To optimize the above technical solution, the specific measures also include:
[0011] The method of using a collaborative filtering model to predict feature aggregation for real-time UWB positioning data in co-location application scenarios specifically includes:
[0012] Using the collaborative filtering model, the calculation formula is as follows:
[0013]
[0014] Where RMSSTD represents the combined standard deviation of all variables in the population, S i Let p represent the sum of the standard deviations of the i-th variable across all groups, where p is the number of variables;
[0015] The greater the difference between groups after clustering, the higher the dissimilarity between groups and the better the clustering effect.
[0016] The formula for calculating the differences between groups after clustering is as follows:
[0017]
[0018] Where R-Square represents the magnitude of the difference between groups after clustering, W represents the degree of difference within each group after clustering, B represents the degree of difference between each group after clustering, and T represents the total degree of difference of all data objects after clustering, and T = W + B; the formula for calculating T is as follows:
[0019]
[0020] Where p represents the number of indicators, s represents the s-th indicator, n represents the number of group members, i represents the ith group member, and Xis represents the clustering result of each data object. This represents the overall average.
[0021] The geometric accuracy factor positioning performance is expressed as an evaluation index as follows:
[0022]
[0023] In the formula, GDOP represents the geometric precision factor positioning performance, G represents the coefficient matrix of the linear equation system established by the distance from the actual location of the base station to the location of the base station positioning tag, and tr() is the trace of the matrix.
[0024] The use of mean square error and root mean square error as evaluation metrics is expressed as follows:
[0025]
[0026] In the formula, MSE represents the mean squared error, N is the dimension of the sample space, and i represents the dimension of the i-th sample space. Let H be the estimated position coordinates of the target in the N-dimensional sample space, and let H be the true position coordinates of the target in the N-dimensional sample space.
[0027] Suppose there are N location tags used for positioning, and the actual position of location tag i is MS(x). i ,y i ,z i The estimated location of the location tag is obtained from the positioning algorithm. The root mean square error of the location estimated by the positioning tag is expressed as follows:
[0028]
[0029] In the formula, RMSE represents the root mean square error.
[0030] The aforementioned process of learning the accuracy evaluation model through real-time UWB positioning data via real-time federated learning, and updating the calculation results through model learning and iterative updates, specifically involves the terminal using local images or videos to perform local training on the global model of the accuracy evaluation model downloaded from the server during training iterations, and reporting the intermediate training results of the iterations to the server through the 6G uplink channel.
[0031] The second aspect of this application provides a 6G base station star-flash combined with real-time federated learning wireless positioning system, comprising:
[0032] The positioning factor acquisition module is used to predict and aggregate features of real-time UWB positioning data in co-site application scenarios using a collaborative filtering model, sort the prediction results, and obtain the key factors affecting UWB positioning accuracy.
[0033] The data tagging module is used to aggregate real-time UWB positioning data with key factors affecting UWB positioning accuracy as the center point, and to tag the aggregated data as indoor or outdoor positioning tags according to the location where the log data is generated.
[0034] The accuracy evaluation model construction module is used to evaluate the accuracy of aggregated data according to whether the location label is indoor or outdoor. In the co-site application scenario, the geometric accuracy factor positioning performance is used as the evaluation index for the outdoor side, and the mean square error and root mean square error algorithms are used as the evaluation index for the indoor side in the co-site application scenario, thus constructing an accuracy evaluation model.
[0035] The real-time federated learning module is used to learn the accuracy evaluation model using real-time UWB positioning data through a real-time federated learning process, and to update the calculation results through model learning and iterative updates.
[0036] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the 6G base station star-flash combined real-time federated learning wireless positioning method as described above.
[0037] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that causes a computer to execute the 6G base station star-flash combined real-time federated learning wireless positioning method as described above.
[0038] Compared with the prior art, the beneficial effects of the present invention are:
[0039] This invention addresses the issue of shared or multiple base stations in UWB signal reconstruction. Specifically, for multiple operators sharing a single base station, it employs a collaborative filtering model to predict feature aggregation. Data aggregation is centered on factors that influence UWB positioning accuracy based on recommended values, and indoor and outdoor locations are marked according to the log data. These factors primarily include: multipath effects, non-line-of-sight propagation, multi-operation interference, the number and location of reference base stations, clock synchronization errors, and atmospheric accuracy degradation.
[0040] This invention estimates the location accuracy of aggregated data separately for indoor and outdoor locations. In co-site application scenarios, the measurement distance and coordinates between outdoor base stations and tags are evaluated using the geometrical factor of precision (GDOP) as a positioning performance evaluation index. For indoor locations, the location estimation accuracy of UWB indoor positioning tags is evaluated using mean square error (MSE) and root mean square error (RMSE) algorithms. Real-time data is used to learn UWB positioning through a real-time federated learning process. Through continuous model learning and iteration, the best calculation results are updated to improve positioning accuracy. Attached Figure Description
[0041] Figure 1 : A flowchart illustrating the 6G base station star-flash combined with real-time federated learning wireless positioning method of the present invention.
[0042] Figure 2 : A schematic diagram of the structure of the 6G base station star-flash combined real-time federated learning wireless positioning system of the present invention. Detailed Implementation
[0043] The present invention will be further described in detail below through embodiments, but it should not be construed as limiting the scope of the subject matter of the present invention to the following embodiments. All technologies implemented based on the above content of the present invention fall within the scope of the present invention.
[0044] The following is an explanation of some of the terms used in this invention:
[0045] A co-located site is a site where multiple base stations share the same location. Specifically, it refers to a site where multiple wireless communication base stations or towers share the same physical location. The sites use different frequencies or modulation schemes to avoid mutual interference and ensure stable signal transmission. The advantages of co-located sites include increased coverage and capacity of the wireless communication network, and reduced construction and maintenance costs.
[0046] UWB: Ultra Wide Band (UWB) technology is a wireless carrier communication technology that does not use sinusoidal carriers but instead uses nanosecond-level non-sinusoidal narrow pulses to transmit data, thus occupying a very wide spectrum.
[0047] The StarLight Alliance was officially established on September 22, 2020, and currently has over 140 member organizations. StarLight technology can meet the needs of smart cars, smart terminals, smart manufacturing, and smart homes for high-quality short-range wireless connectivity.
[0048] K-means++ is an algorithm for selecting initial values (or "seeds") for the K-means clustering algorithm. The selection of the K initial centroids has a significant impact on the final clustering results and running time, so it is necessary to select K appropriate centroids.
[0049] The basic idea of the K-means algorithm is still to minimize the sum of squared deviations (K-means), but in its implementation, it adopts an iterative correction method.
[0050] This invention provides a 6G base station star-flash combined with real-time federated learning wireless positioning method in one embodiment, for wireless positioning in co-site application scenarios formed by multiple operators sharing base stations, such as... Figure 1 As shown, it includes the following steps:
[0051] In co-site application scenarios, a collaborative filtering model is used to predict and aggregate features of real-time UWB positioning data. The prediction results are then retrieved, the result set is sorted, and the key factors affecting UWB positioning accuracy are obtained.
[0052] Data aggregation of real-time UWB positioning data is performed with key factors affecting UWB positioning accuracy as the center point, and the aggregated data is marked as indoor or outdoor positioning tags according to the location where the log data is generated.
[0053] The aggregated data is evaluated for accuracy at different locations based on whether the location tag is indoor or outdoor: the geometric accuracy factor positioning performance is used as the evaluation index for outdoor locations in co-located application scenarios, and the mean square error and root mean square error algorithms are used as the evaluation index for indoor locations in co-located application scenarios, thus constructing an accuracy evaluation model.
[0054] The accuracy evaluation model is learned by using real-time UWB positioning data through a real-time federated learning process, and the calculation results are updated through model learning and iterative updates.
[0055] Among the factors affecting UWB positioning accuracy are:
[0056] Multipath effect: Ultra-wideband signals propagating indoors are affected by complex indoor environments (such as walls, windows, and other obstructions), resulting in reflections and refractions, thus creating a multipath effect. This multipath effect causes signals to arrive at the positioning base station antenna via different paths, meaning the first arriving signal component does not arrive in a straight line, exhibiting different temporal and spatial orientations. Consequently, the phase relationship between the signal components arriving at the positioning base station changes, and the different paths also cause varying time delays for each component. Even small delays lead to interference between signal components, causing rapid signal attenuation and energy loss, a decrease in the signal-to-noise ratio, and ultimately, measurement errors and reduced positioning accuracy.
[0057] Non-line-of-sight (NLOS) propagation: There is no direct path between the transmitting and receiving points. Radio signals can only reach the receiving end after reflection or refraction.
[0058] Multiple access interference:
[0059] Because reference base stations need to coordinate positioning, a reference base station must receive signals not only from the node to be located but also from nearby base stations. The superposition of these different signals causes interference, reducing the signal-to-noise ratio and resulting in a significant difference between the measured TOA value and the true value. Solutions include: a) using mutually orthogonal time-hopping sequences, allowing users to transmit signals at different times, thus avoiding interference between users; b) using a matched filter at the receiving end to output the spread spectrum signal within the multiple access channel, and then using effective signal processing methods to ultimately recover the original transmitted signal data. This method can also mitigate the near-far effect to some extent.
[0060] Reference number of base stations:
[0061] Increasing the number of reference base stations can increase the information element redundancy of TOA measurements, reduce positioning errors, and improve positioning accuracy. However, more positioning base stations are not always better. When the number of positioning reference base stations increases to a certain extent, it becomes impossible to fully utilize the distance information provided by other base stations.
[0062] Reference base station location:
[0063] Properly positioning reference base stations can suppress multiple access interference and improve positioning accuracy to a certain extent. The degree to which the relative positions of the reference base stations and the node to be located affect the positioning system can be represented by the Geometric Dilution of Precision (GDOP). In practical applications, GDOP can be used as a selection criterion when choosing a base station from a large number of reference base stations. Literature review shows that uniformly distributed base stations provide better positioning performance than randomly distributed reference base stations.
[0064] Clock synchronization error: The clock synchronization methods used in TOA-based and TDOA-based ultra-wideband positioning algorithms are different. TOA-based positioning requires strict clock synchronization not only between the node to be located and each reference base station, but also between the reference base stations themselves. TDOA-based positioning, on the other hand, only requires clock synchronization between the reference base stations, thus reducing the complexity and cost of the positioning system. Therefore, clock synchronization error is closely related to positioning accuracy. Clock synchronization is affected by many factors.
[0065] Atmospheric accuracy degradation is primarily due to the significant increase in UWB signal arrival time caused by atmospheric error delays when the base station and tag are far apart. Therefore, it is impossible to accurately calculate the distance between the base station and the tag. This problem is addressed by obtaining the measured distance between the tag and the base station using UWB equipment, establishing an observation equation based on this distance, and introducing an atmospheric error parameter proportional to the measured distance into the equation. The position parameter and atmospheric error parameter are then compensated together using triangulation to eliminate errors caused by atmospheric interference during signal propagation. The resulting tag position after atmospheric error elimination, along with the real-time atmospheric error parameters, are then output.
[0066] This invention retrieves results from a large result set using a collaborative filtering model or some relevance models or algorithms, and then sorts the retrieved result set. The sorting stage can be further subdivided into coarse sorting, fine sorting, and re-sorting for more detailed stages. Based on different recommendation mechanisms and utilizing a subset of data from the data source, certain patterns are analyzed or feature aggregation recommendation values are directly predicted. In this way, direct prediction recommendations are more accurate.
[0067] For real-time UWB positioning data in co-located application scenarios, a collaborative filtering model is used to predict feature aggregation, specifically including:
[0068] Using the collaborative filtering model, the calculation formula is as follows:
[0069]
[0070] Where RMSSTD represents the combined standard deviation of all variables in the population, Si represents the sum of the standard deviations of the i-th variable in each population, and p is the number of variables;
[0071] The greater the difference between groups after clustering, the higher the dissimilarity between groups and the better the clustering effect.
[0072] The formula for calculating the differences between groups after clustering is as follows:
[0073]
[0074] Where R-Square represents the magnitude of the difference between groups after clustering, W represents the degree of difference within each group after clustering, B represents the degree of difference between each group after clustering, and T represents the total degree of difference of all data objects after clustering, and T = W + B; the formula for calculating T is as follows:
[0075]
[0076] Where p represents the number of indicators, s represents the s-th indicator, n represents the number of group members, i represents the ith group member, and Xis represents the clustering result of each data object. This represents the overall average.
[0077] The geometric accuracy factor, representing the positioning performance as an evaluation index, is as follows:
[0078]
[0079] In the formula, GDOP represents the geometric precision factor positioning performance, G represents the coefficient matrix of the linear equation system established by the distance from the actual location of the base station to the location of the base station positioning tag, and tr() is the trace of the matrix.
[0080] The Geometric Difference of Precision (GDOP) represents the amplification factor of the distance vector between the positioning tag and the base station caused by UWB ranging errors. A higher GDOP value indicates a higher angular similarity between the base station and the positioning tag, resulting in lower positioning accuracy. Conversely, a lower GDOP value allows the tag to transmit signals omnidirectionally, and base stations, evenly distributed around the edges and using omnidirectional antennas for signal reception, provide better positioning results. GDOP can be used as a base station layout indicator; base stations participating in positioning should aim to minimize the GDOP.
[0081] The mean square error and root mean square error algorithms are used as evaluation metrics, as shown below:
[0082]
[0083] In the formula, MSE represents the mean squared error, N is the dimension of the sample space, and i represents the dimension of the i-th sample space. Let H be the estimated position coordinates of the target in the N-dimensional sample space, and let H be the true position coordinates of the target in the N-dimensional sample space.
[0084] The magnitude of the mean square error (MSE) is directly proportional to the expected value of the square of the difference between the estimated value and the true value. The smaller the MSE value, the higher the accuracy.
[0085] Suppose there are N location tags used for positioning, and the actual position of location tag i is MS(x). i ,y i ,z i The estimated location of the location tag is obtained from the positioning algorithm. The root mean square error of the location estimated by the positioning tag is expressed as follows:
[0086]
[0087] In the formula, RMSE represents the root mean square error.
[0088] If there are outliers with large errors in the estimated location coordinates of the tag, the outliers must be removed before using the root mean square (RMS) to evaluate the performance of the positioning algorithm. Otherwise, the RMS error will be very large, and the evaluation of the positioning algorithm's performance will be inaccurate.
[0089] The accuracy evaluation model is learned through real-time UWB positioning data via a real-time federated learning process. The learning process involves model learning and iterative updates of the computational results. Specifically: In the Nth training iteration, the terminal uses local images / videos to locally train the global model of the accuracy evaluation model downloaded from the server; the intermediate training results of the Nth iteration are reported to the server via the 6G uplink channel; simultaneously, the server sends the N+1th global model and training configuration to the terminal; while the server aggregates the gradients collected in the Nth iteration, the terminal executes the training for the N+1th iteration; the federated aggregation outputs an updated global model for the next iteration training, and this model, along with the updated training configuration, is distributed to the terminal; the training sequence completes the reporting of training results for N-1 iterations and the distribution of the global model and configuration of training parameters for the N+1th iteration during the terminal's Nth iteration training process.
[0090] In another embodiment of the present invention, a 6G base station star-flash combined real-time federated learning wireless positioning system is proposed, such as... Figure 2 As shown, it includes:
[0091] The positioning factor acquisition module is used to predict and aggregate features of real-time UWB positioning data in co-site application scenarios using a collaborative filtering model, retrieve the prediction results, obtain the result set and sort it to obtain the key factors affecting UWB positioning accuracy.
[0092] The data tagging module is used to aggregate real-time UWB positioning data with key factors affecting UWB positioning accuracy as the center point, and to tag the aggregated data as indoor or outdoor positioning tags according to the location where the log data is generated.
[0093] The accuracy evaluation model construction module is used to evaluate the accuracy of aggregated data according to whether the location label is indoor or outdoor. In the co-site application scenario, the geometric accuracy factor positioning performance is used as the evaluation index for the outdoor side, and the mean square error and root mean square error algorithms are used as the evaluation index for the indoor side in the co-site application scenario, thus constructing an accuracy evaluation model.
[0094] The real-time federated learning module is used to learn the accuracy evaluation model using real-time UWB positioning data through a real-time federated learning process, and to update the calculation results through model learning and iterative updates.
[0095] In another embodiment of the present invention, an electronic device is proposed, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned 6G base station star-flash combined real-time federated learning wireless positioning method.
[0096] In another embodiment of the present invention, a computer-readable storage medium is provided, storing a computer program that enables a computer to execute the above-described 6G base station star-flash combined real-time federated learning wireless positioning method.
[0097] In the embodiments disclosed in this application, a computer storage medium may be a tangible medium that may contain or store programs for use by or in conjunction with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of computer storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0098] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Any simple modifications, equivalent substitutions, and improvements made by those skilled in the art to the above embodiments without departing from the scope of the present invention and based on the technical essence of the present invention shall still fall within the protection scope of the present invention.
Claims
1. A 6G base station star flash combined with real-time federated learning wireless positioning method, characterized by: Wireless positioning is performed for the co-site application scenario formed by multiple operators sharing base stations, including the following steps: The collaborative filtering model is used to predict feature aggregation of real-time UWB positioning data in co-site application scenarios, and the prediction results are sorted to obtain the key factors affecting UWB positioning accuracy. The real-time data of UWB positioning is aggregated with the key factors that affect the accuracy of UWB positioning as the center point, and the aggregated data is marked as indoor or outdoor positioning tags according to the location where the log data is generated; The aggregated data is evaluated for accuracy at different locations according to the positioning label of indoor or outdoor. The geometric precision factor positioning performance is used as the evaluation index for the outdoor side in the co-site application scenario, and the mean square error and root mean square error algorithm are used as the evaluation index for the indoor side in the co-site application scenario to build an accuracy evaluation model. The real-time federated learning process uses UWB positioning real-time data to learn the accuracy evaluation model, and the calculation results are updated through model learning and iteration.
2. According to claim 1, the 6G base station star flash combined with real-time federated learning wireless positioning method is characterized by: The collaborative filtering model is used to predict feature aggregation of UWB positioning real-time data in the co-site application scenario, specifically including: Using the collaborative filtering model, the calculation formula is as follows: Among them, RMSSTD represents the comprehensive standard deviation of all variables in the group, Si represents the sum of the standard deviations of the ith variable in each group, and p is the number of variables; The greater the difference between groups after clustering, the higher the dissimilarity between groups and the better the clustering effect.
3. The 6G base station star flash combined with real-time federated learning wireless positioning method according to claim 2 is characterized by: The formula for calculating the difference between groups after clustering is as follows: Among them, R-Square represents the size of the difference between groups after clustering, W represents the degree of difference within each group after clustering, B represents the degree of difference between each group after clustering, T represents the total degree of difference of all data objects after clustering, and T = W + B; the calculation formula of T is as follows: Where p represents p indicators, s represents the sth indicator, n represents n group members, i represents the ith group member, and Xis represents the clustering result of each data object. Represents the overall mean.
4. The 6G base station star flash combined with real-time federated learning wireless positioning method according to claim 1 is characterized by: The geometric precision factor positioning performance is expressed as an evaluation index as follows: Where GDOP represents the geometric dilution of precision positioning performance, and G represents the coefficient matrix of the linear equations established by the distance from the actual position of the base station to the position of the base station positioning tag; tr() is the trace of a matrix.
5. The 6G base station star flash combined with real-time federated learning wireless positioning method according to claim 1 is characterized by: The mean square error and root mean square error algorithm are used as evaluation indicators, which are expressed as follows: In the formula, MSE represents mean square error, N is the dimension of sample space, i represents the i-th sample space dimension, is the estimated position coordinate of the target to be measured in the N-dimensional sample space, and H is the actual position coordinate of the target to be measured in the N-dimensional sample space.
6. The 6G base station star flash combined with real-time federated learning wireless positioning method according to claim 1 is characterized by: Assume that there are N positioning tags for positioning, and the real position of positioning tag i is MS(x i ,y i ,z i ), the estimated position of the positioning tag obtained by the positioning algorithm is The root mean square error of the estimated position of the positioning tag is expressed as follows: In the formula, RMSE stands for root mean square error.
7. The 6G base station star flash combined with real-time federated learning wireless positioning method according to claim 1 is characterized by: The real-time federated learning process is used to learn the accuracy evaluation model through UWB positioning real-time data, and the calculation results are updated through model learning and iteration. Specifically, in the training iteration, the terminal uses local images or videos to locally train the global model of the accuracy evaluation model downloaded from the server, and reports the iterative intermediate training results to the server through the 6G uplink channel.
8. A 6G base station star flash combined with real-time federated learning wireless positioning system, characterized in that: include: The positioning factor acquisition module is used to predict feature aggregation of UWB positioning real-time data in the co-site application scenario using a collaborative filtering model, sort the prediction results, and obtain the key factors affecting UWB positioning accuracy; The data marking module is used to aggregate the real-time data of UWB positioning with the key factors affecting the UWB positioning accuracy as the center point, and mark the aggregated data as indoor or outdoor positioning tags according to the location where the log data is generated; The module for constructing an accuracy evaluation model is used to evaluate the accuracy of different locations of aggregated data according to the positioning label of indoor or outdoor. The outdoor side in the co-site application scenario uses the geometric precision factor positioning performance as the evaluation index, and the indoor side in the co-site application scenario uses the mean square error and root mean square error algorithm as the evaluation index to construct an accuracy evaluation model. The real-time federated learning module is used to learn the accuracy evaluation model using UWB positioning real-time data through the real-time federated learning process, and update the calculation results through model learning and iteration.
9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the 6G base station star flash combined with real-time federated learning wireless positioning method as described in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the 6G base station star flash combined with real-time federated learning wireless positioning method as described in any one of claims 1 to 7.