A UWB indoor positioning method and system based on generative fingerprint traversal matching

Abstract

The application discloses a kind of UWB indoor positioning method and system based on generative fingerprint traversal matching, first configure hyperparameter and divide rough virtual grid, then construct rough fingerprint database according to rough virtual grid, then according to index number, rough grid point data is selected from rough fingerprint database in turn, the difference between the measured distance of each base station and the Euclidean distance of corresponding rough grid point is calculated and used as the first distance difference value;The application realizes the indoor positioning function with the target positioning space being divided into three-dimensional virtual grid unit for sub-meter accuracy, and the distance difference can be non-linearized through Huber loss function, thereby reducing the influence of abnormal measurement in the traversal matching process, and the difference between the real position and the virtual grid point can be accurately described through the loss value to determine the target position, not only avoiding the complex workload of manual fingerprint data collection, but also significantly reducing the influence of ranging error on positioning, suitable for being widely promoted and used.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of indoor positioning, and particularly relates to a UWB indoor positioning method and system based on generative fingerprint traversal matching. BACKGROUND

[0002] The rapid development of Internet of Things (IoT) technology is promoting the intelligent upgrading of indoor infrastructure, and various intelligent devices have an increasing demand for real-time location information of users to provide more accurate and personalized services. In recent years, indoor positioning technology has made significant progress, and various positioning technologies based on WiFi, Bluetooth, ultra-wideband (UWB), inertial navigation and the like have emerged. The UWB technology has become one of the mainstream solutions for short-distance high-precision positioning due to its nanosecond non-sine narrow pulse and ultra-wideband characteristics, and has a centimeter-level ranging accuracy and strong anti-interference capability.

[0003] At present, the existing indoor positioning methods are generally affected by the environment and signal propagation characteristics, so that the UWB ranging results are mostly accompanied by ranging errors, especially in the non-line-of-sight (NLOS) propagation situation, obstacles can cause signal attenuation and change the propagation path, thereby introducing positive bias gross errors in the ranging values, which seriously affects the accuracy and reliability of positioning. Moreover, the existing NLOS identification and mitigation methods mostly depend on the accurate detection and correction of abnormal ranging values, which is easily affected by the identification accuracy in a complex dynamic environment, thereby leading to unstable error mitigation effect. Therefore, it is necessary to design a UWB indoor positioning method and system based on generative fingerprint traversal matching. SUMMARY

[0004] The application aims to overcome the shortcomings of the prior art, better effectively solve the problem that the existing indoor positioning methods are generally affected by the environment and signal propagation characteristics, so that the UWB ranging results are mostly accompanied by ranging errors, especially in the non-line-of-sight (NLOS) propagation situation, obstacles can cause signal attenuation and change the propagation path, thereby introducing positive bias gross errors in the ranging values, which seriously affects the accuracy and reliability of positioning, and the existing NLOS identification and mitigation methods mostly depend on the accurate detection and correction of abnormal ranging values, which is easily affected by the identification accuracy in a complex dynamic environment, thereby leading to unstable error mitigation effect, and provide a UWB indoor positioning method and system based on generative fingerprint traversal matching. The application realizes the indoor positioning function with sub-meter accuracy by dividing the target positioning space into three-dimensional virtual grid units, and the Huber loss function can non-linearize the distance difference, thereby reducing the influence of abnormal measurement in the traversal matching process. The loss value can accurately describe the difference between the real position and the virtual grid point and determine the target position, which not only avoids the cumbersome workload of manually collecting fingerprint data, but also significantly reduces the influence of ranging errors on positioning.

[0005] To achieve the above object, the technical scheme adopted by the present application is:

[0006] A UWB indoor positioning method based on generative fingerprint traversal matching, comprising the following steps,

[0007] Step A, configuring hyperparameters and dividing coarse virtual grids;

[0008] Step B, constructing a coarse fingerprint database according to the coarse virtual grids;

[0009] Step C, selecting coarse grid point data from the coarse fingerprint database in order according to the index number, and then calculating the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point as the first distance difference;

[0010] Step D, calculating the Huber loss corresponding to the coarse grid point using the first distance difference, and then taking the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss as the coarse positioning position;

[0011] Step E, setting a cubic region in the three-dimensional space with the coarse positioning position as the center, dividing the cubic region into a plurality of fine virtual grids, and obtaining fine grid points;

[0012] Step F, assigning an index number to each fine grid point and recording the three-dimensional coordinates of the fine grid point, and then calculating the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point as the second distance difference;

[0013] Step G, calculating the Huber loss corresponding to the fine grid point using the second distance difference, and taking the fine grid point with the minimum Huber loss as the fine positioning position, and then taking the fine positioning position as the target position and completing the UWB indoor positioning work.

[0014] The foregoing UWB indoor positioning method based on generative fingerprint traversal matching, step A, configuring hyperparameters and dividing coarse virtual grids, the specific steps are as follows,

[0015] Step A1, configuring hyperparameters, the hyperparameters including Huber loss function threshold δ, coarse grid side length C and fine grid side length F, the Huber loss function threshold δ is specifically valued according to the ranging error level, the coarse grid side length C is specifically valued according to the balance of coarse positioning and fine positioning calculation amount, and the fine grid side length F is specifically valued according to the target positioning accuracy of the user;

[0016] Step A2, dividing coarse virtual grids, the specific steps are as follows,

[0017] Step A21 involves expanding the base station space and using it as the target positioning space TLS. Specifically, the planar boundary of the base station group is expanded to twice its original length in the planar dimension, and the height boundary is expanded from the ground upwards to 1.5 times the height of the highest base station. The side lengths of the target positioning space TLS in the three-dimensional direction can be represented as L. x L y and L z ;

[0018] Step A22: Divide the target positioning space TLS into multiple three-dimensional virtual grids with side length C, and retain sub-grids that are less than a complete grid space after division. Then, take all grid vertices after rasterization as coarse grid points, as shown in formula (1).

[0019]

[0020] Where O represents the total number of coarse grid points. It is a rounding function;

[0021] Step A23: Starting from the grid point with the smallest three-dimensional coordinates in the target positioning space TLS, assign a unique index number to each coarse grid point in dictionary order, and then record the three-dimensional coordinates corresponding to each coarse grid point while numbering them sequentially along the x-axis, y-axis, and z-axis.

[0022] The aforementioned UWB indoor positioning method based on generative fingerprint traversal matching, step B, involves constructing a coarse fingerprint database based on a coarse virtual grid. The specific steps are as follows:

[0023] Step B1: Using each coarse grid point as a reference position, extract the corresponding distance features for each coarse grid point, where the Euclidean distance between the i-th coarse grid point and the k-th base station is... As shown in formula (2),

[0024]

[0025] in, and (x) k ,y k ,z k The coordinates of the i-th coarse grid point and the k-th base station are respectively the three-dimensional coordinates of the i-th coarse grid point and the k-th base station.

[0026] Step B2: After obtaining the Euclidean distances from all coarse grid points to each base station, a coarse fingerprint database can be constructed. The coarse fingerprint database is stored in a structured table format. Each row of the coarse fingerprint database corresponds to the fingerprint information of a coarse grid point. The first column of the coarse fingerprint database is the coarse grid point index number, the second column is the three-dimensional coordinates of the coarse grid point, and each of the other columns of the coarse fingerprint database is the Euclidean distance between the coarse grid point and each base station. All data in the coarse fingerprint database is stored with a precision of three decimal places.

[0027] In the aforementioned UWB indoor positioning method based on generative fingerprint traversal matching, step C involves sequentially selecting coarse grid point data from the coarse fingerprint database according to the index number, then calculating the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point, and using this difference as the first distance difference. The measured distance d of the k-th base station is... k Euclidean distance from the i-th coarse grid point to the corresponding base station The difference r i,k As shown in formula (3),

[0028]

[0029] The aforementioned UWB indoor positioning method based on generative fingerprint traversal matching, in step D, calculates the Huber loss corresponding to the coarse grid point using the first distance difference, and then uses the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss as the coarse positioning position. The specific steps are as follows.

[0030] Step D1: Calculate the Huber loss corresponding to the coarse grid point using the first distance difference. Specifically, assume there are M first distance differences, and then calculate the Huber loss corresponding to the coarse grid point in each first distance difference, as shown in formula (4).

[0031]

[0032] in, The Huber loss is the approximate value for each grid point.

[0033] Step D2: The coarse grid point's 3D coordinates corresponding to the minimum Huber loss are used as the coarse localization position. Specifically, after obtaining the Huber losses corresponding to all coarse grid points, the minimum Huber loss is found by traversing according to the index number, and then the coarse grid point's 3D coordinates corresponding to the minimum Huber loss are selected as the coarse localization position P. c The Huber loss is used to quantitatively describe the similarity between the target location and the coarse grid points.

[0034] The aforementioned UWB indoor positioning method based on generative fingerprint traversal matching, in step E, uses the coarse positioning location as the center and sets a cubic region in three-dimensional space, then divides the cubic region into multiple fine virtual grids to obtain fine grid points. The specific steps are as follows.

[0035] Step E1: Using the coarse positioning position as the center, define a cubic region in three-dimensional space. Specifically, this involves setting the coarse positioning position P... c As the center, define a cube region with side length C in three-dimensional space;

[0036] Step E1 involves dividing the cubic region into multiple fine virtual grids and obtaining fine grid points. Specifically, the cubic region with side length C is divided into multiple fine virtual grids with side length F and fine grid points are obtained. If part of the cubic region exceeds the target positioning space TLS, the excess space is removed. The process of obtaining the fine grid points is shown in formula (5).

[0037]

[0038] Where Q is the total number of fine grid points obtained.

[0039] The aforementioned UWB indoor positioning method based on generative fingerprint traversal matching, step F, involves assigning an index number to each fine grid point and recording its three-dimensional coordinates. Then, the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point is calculated and used as the second distance difference. The specific steps are as follows:

[0040] Step F1: Assign an index number to each fine grid point and record the three-dimensional coordinates of the fine grid point. Specifically, starting from the grid point with the smallest three-dimensional coordinates in the target positioning space TLS, assign a unique index number to each fine grid point in dictionary order, and then record the corresponding three-dimensional coordinates of each fine grid point while numbering them sequentially along the x-axis, y-axis, and z-axis.

[0041] Step F2: Calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point, and use this difference as the second distance difference, where the measured distance d of the k-th base station is... k The difference r between the j-th fine grid point and the Euclidean distance to the corresponding base station j,k As shown in formula (6),

[0042]

[0043] in, Let be the three-dimensional coordinates of the j-th fine grid point.

[0044] The aforementioned UWB indoor positioning method based on generative fingerprint traversal matching, in step G, calculates the Huber loss corresponding to the fine grid point using the second distance difference, and takes the fine grid point with the smallest Huber loss as the fine positioning position. Then, the fine positioning position is taken as the target position, and the UWB indoor positioning operation is completed. Specifically, the calculation of the Huber loss corresponding to the fine grid point using the second distance difference is as follows: assuming there are N total second distance differences, the Huber loss corresponding to the fine grid point in each second distance difference is calculated, as shown in formula (7).

[0045]

[0046] in, This represents the Huber loss corresponding to the fine grid points.

[0047] A UWB indoor positioning system based on generative fingerprint traversal matching includes a coarse virtual grid division module, a coarse fingerprint database construction module, a first distance difference calculation module, a coarse positioning location acquisition module, a fine virtual grid division module, a second distance difference calculation module, and a fine positioning location acquisition module. The coarse virtual grid division module is used to configure hyperparameters and divide coarse virtual grids.

[0048] The coarse fingerprint database construction module is used to construct a coarse fingerprint database based on a coarse virtual grid.

[0049] The first distance difference calculation module is used to select coarse grid point data from the coarse fingerprint database in sequence according to the index number, and then calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point as the first distance difference.

[0050] The coarse positioning position acquisition module is used to calculate the Huber loss corresponding to the coarse grid point using the first distance difference, and then use the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss as the coarse positioning position.

[0051] The fine virtual grid division module is used to take the coarse positioning position as the center and set a cubic region in three-dimensional space, and then divide the cubic region into multiple fine virtual grids to obtain fine grid points.

[0052] The second distance difference calculation module is used to assign an index number to each fine grid point and record the three-dimensional coordinates of the fine grid point, and then calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point as the second distance difference;

[0053] The fine positioning location acquisition module is used to calculate the Huber loss corresponding to the fine grid point using the second distance difference, and take the fine grid point with the smallest Huber loss as the fine positioning location, and then take the fine positioning location as the target location to complete the UWB indoor positioning operation.

[0054] The beneficial effects of this invention are as follows: The present invention provides a UWB indoor positioning method and system based on generative fingerprint traversal matching. First, hyperparameters are configured and a coarse virtual grid is divided. Then, a coarse fingerprint database is constructed based on the coarse virtual grid. Next, coarse grid point data is selected sequentially from the coarse fingerprint database according to the index number. Then, the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point is calculated and used as the first distance difference. Subsequently, the Huber loss corresponding to the coarse grid point is calculated using the first distance difference. Then, the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss are used as the coarse positioning position. Then, the coarse positioning position is used as the center and a cubic region is set in three-dimensional space. The cubic region is then divided into multiple fine virtual grids to obtain fine grid points. Next, an index number is assigned to each fine grid point and the three-dimensional coordinates of the fine grid point are recorded. Then, the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point is calculated and used as the second distance difference. Finally, the Huber loss corresponding to the fine grid point is calculated using the second distance difference, and the Huber loss is minimized. Small, fine-grained grid points are used as fine-grained positioning locations, and these locations are then used as the target locations to complete UWB indoor positioning operations. This effectively enables the UWB indoor positioning method and system to achieve sub-meter-level accuracy indoor positioning by dividing the target positioning space into three-dimensional virtual grid units. Furthermore, by constructing a robust mapping between location-related features and spatial coordinates, the dependence on NLOS recognition or error modeling is eliminated, resulting in better adaptability and accuracy. The Huber loss function nonlinearizes the distance difference, thereby mitigating the impact of abnormal measurements during the traversal matching process. Simultaneously, the loss value can more accurately describe the difference between the real location and the virtual grid points, achieving precise matching and determining the target location. This not only avoids the cumbersome workload of manually collecting fingerprint data and training recognition models but also significantly reduces the impact of ranging errors on positioning. This invention can meet the requirements of real-time performance and efficiency in practical scenarios and can be widely applied in smart cities, autonomous driving, and emergency rescue fields. This provides important technical support and theoretical basis for high-precision positioning in complex indoor environments. Attached Figure Description

[0055] Figure 1 This is an overall flowchart of a UWB indoor positioning method based on generative fingerprint traversal matching according to the present invention.

[0056] Figure 2 This is a schematic diagram of configuring hyperparameters and dividing a coarse virtual grid according to the present invention;

[0057] Figure 3 This is a schematic diagram of the coarse positioning position of the present invention;

[0058] Figure 4 This is a schematic diagram of the fine virtual grid division of the present invention;

[0059] Figure 5 This is a schematic diagram of the fine positioning position of the present invention;

[0060] Figure 6 This is a schematic diagram of a laboratory corridor environment in an embodiment of the present invention;

[0061] Figure 7 This is a schematic diagram of the laboratory corridor environment surrounding the corridor in an embodiment of the present invention;

[0062] Figure 8 This is a schematic diagram comparing the positioning trajectories of various algorithms in the embodiments of the present invention;

[0063] Figure 9 This is a schematic diagram comparing the CDF curves of various algorithms in the embodiments of the present invention. Detailed Implementation

[0064] The present invention will now be further described with reference to the accompanying drawings.

[0065] like Figure 1 As shown, the present invention provides a UWB indoor positioning method based on generative fingerprint traversal matching, comprising the following steps:

[0066] like Figure 2 As shown, step A involves configuring hyperparameters and dividing a coarse virtual grid. The specific steps are as follows:

[0067] Step A1: Configure hyperparameters, which include Huber loss function threshold δ, coarse grid side length C, and fine grid side length F. The Huber loss function threshold δ is specifically determined based on the ranging error level. The coarse grid side length C is specifically determined based on the balance calculation between coarse and fine positioning. The fine grid side length F is specifically determined based on the user's target positioning accuracy.

[0068] Step A2: Divide the grid into a rough virtual grid. The specific steps are as follows:

[0069] Step A21 involves expanding the base station space and using it as the target positioning space TLS. Specifically, the planar boundary of the base station group is expanded to twice its original length in the planar dimension, and the height boundary is expanded from the ground upwards to 1.5 times the height of the highest base station. The side lengths of the target positioning space TLS in the three-dimensional direction can be represented as L. x L y and L z ;

[0070] Step A22: Divide the target positioning space TLS into multiple three-dimensional virtual grids with side length C, and retain sub-grids that are less than a complete grid space after division. Then, take all grid vertices after rasterization as coarse grid points, as shown in formula (1).

[0071]

[0072] Where O represents the total number of coarse grid points. It is a rounding function;

[0073] Step A23: Starting from the grid point with the smallest three-dimensional coordinates in the target positioning space TLS, assign a unique index number to each coarse grid point in dictionary order, and then record the three-dimensional coordinates corresponding to each coarse grid point while numbering them sequentially along the x-axis, y-axis, and z-axis.

[0074] Step B involves constructing a coarse fingerprint database based on the coarse virtual grid. The specific steps are as follows:

[0075] Step B1: Using each coarse grid point as a reference position, extract the corresponding distance features for each coarse grid point, where the Euclidean distance between the i-th coarse grid point and the k-th base station is... As shown in formula (2),

[0076]

[0077] in, and (x) k ,y k ,z k The coordinates of the i-th coarse grid point and the k-th base station are respectively the three-dimensional coordinates of the i-th coarse grid point and the k-th base station.

[0078] Step B2: After obtaining the Euclidean distances from all coarse grid points to each base station, a coarse fingerprint database can be constructed. The coarse fingerprint database is stored in a structured table format. Each row of the coarse fingerprint database corresponds to the fingerprint information of a coarse grid point. The first column of the coarse fingerprint database is the coarse grid point index number, the second column is the three-dimensional coordinates of the coarse grid point, and each of the other columns of the coarse fingerprint database is the Euclidean distance between the coarse grid point and each base station. All data in the coarse fingerprint database is stored with a precision of three decimal places.

[0079] Step C: Select coarse grid point data from the coarse fingerprint database sequentially according to the index number, then calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point, and use this as the first distance difference, where the measured distance d of the k-th base station is... kEuclidean distance from the i-th coarse grid point to the corresponding base station The difference r i,k As shown in formula (3),

[0080]

[0081] like Figure 3 As shown, in step D, the Huber loss corresponding to the coarse grid point is calculated using the first distance difference, and then the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss are used as the coarse positioning position. The specific steps are as follows.

[0082] Step D1: Calculate the Huber loss corresponding to the coarse grid point using the first distance difference. Specifically, assume there are M first distance differences, and then calculate the Huber loss corresponding to the coarse grid point in each first distance difference, as shown in formula (4).

[0083]

[0084] in, The Huber loss is the approximate value for each grid point.

[0085] Step D2: The coarse grid point's 3D coordinates corresponding to the minimum Huber loss are used as the coarse localization position. Specifically, after obtaining the Huber losses corresponding to all coarse grid points, the minimum Huber loss is found by traversing according to the index number, and then the coarse grid point's 3D coordinates corresponding to the minimum Huber loss are selected as the coarse localization position P. c The Huber loss is used to quantitatively describe the similarity between the target location and the coarse grid points.

[0086] like Figure 4 As shown, in step E, the coarse positioning position is used as the center, and a cubic region is defined in three-dimensional space. Then, the cubic region is divided into multiple fine virtual grids to obtain fine grid points. The specific steps are as follows.

[0087] Step E1: Using the coarse positioning position as the center, define a cubic region in three-dimensional space. Specifically, this involves setting the coarse positioning position P... c As the center, define a cube region with side length C in three-dimensional space;

[0088] Step E1 involves dividing the cubic region into multiple fine virtual grids and obtaining fine grid points. Specifically, the cubic region with side length C is divided into multiple fine virtual grids with side length F and fine grid points are obtained. If part of the cubic region exceeds the target positioning space TLS, the excess space is removed. The process of obtaining the fine grid points is shown in formula (5).

[0089]

[0090] Where Q is the total number of fine grid points obtained.

[0091] Step F involves assigning an index number to each fine grid point and recording its 3D coordinates. Then, the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point is calculated and used as the second distance difference. The specific steps are as follows:

[0092] Step F1: Assign an index number to each fine grid point and record the three-dimensional coordinates of the fine grid point. Specifically, starting from the grid point with the smallest three-dimensional coordinates in the target positioning space TLS, assign a unique index number to each fine grid point in dictionary order, and then record the corresponding three-dimensional coordinates of each fine grid point while numbering them sequentially along the x-axis, y-axis, and z-axis.

[0093] Step F2: Calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point, and use this difference as the second distance difference, where the measured distance d of the k-th base station is... k The difference r between the j-th fine grid point and the Euclidean distance to the corresponding base station j,k As shown in formula (6),

[0094]

[0095] in, Let be the three-dimensional coordinates of the j-th fine grid point.

[0096] like Figure 5 As shown, in step G, the Huber loss corresponding to the fine grid point is calculated using the second distance difference, and the fine grid point with the smallest Huber loss is taken as the fine positioning position. Then, the fine positioning position is taken as the target position, and the UWB indoor positioning operation is completed. Specifically, the Huber loss corresponding to the fine grid point is calculated using the second distance difference. Assuming there are N total second distance differences, the Huber loss corresponding to the fine grid point in each second distance difference is calculated, as shown in formula (7).

[0097]

[0098] in, This represents the Huber loss corresponding to the fine grid points.

[0099] A UWB indoor positioning system based on generative fingerprint traversal matching includes a coarse virtual grid division module, a coarse fingerprint database construction module, a first distance difference calculation module, a coarse positioning location acquisition module, a fine virtual grid division module, a second distance difference calculation module, and a fine positioning location acquisition module. The coarse virtual grid division module is used to configure hyperparameters and divide coarse virtual grids.

[0100] The coarse fingerprint database construction module is used to construct a coarse fingerprint database based on a coarse virtual grid.

[0101] The first distance difference calculation module is used to select coarse grid point data from the coarse fingerprint database in sequence according to the index number, and then calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point as the first distance difference.

[0102] The coarse positioning position acquisition module is used to calculate the Huber loss corresponding to the coarse grid point using the first distance difference, and then use the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss as the coarse positioning position.

[0103] The fine virtual grid division module is used to take the coarse positioning position as the center and set a cubic region in three-dimensional space, and then divide the cubic region into multiple fine virtual grids to obtain fine grid points.

[0104] The second distance difference calculation module is used to assign an index number to each fine grid point and record the three-dimensional coordinates of the fine grid point, and then calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point as the second distance difference;

[0105] The fine positioning location acquisition module is used to calculate the Huber loss corresponding to the fine grid point using the second distance difference, and take the fine grid point with the smallest Huber loss as the fine positioning location, and then take the fine positioning location as the target location to complete the UWB indoor positioning operation.

[0106] To better illustrate the effectiveness of the present invention, a specific embodiment of indoor positioning using the method and system of the present invention is described below.

[0107] like Figure 6 and Figure 7 As shown, this embodiment illustrates a laboratory corridor environment with interference sources such as concrete walls, glass railings, and elevators. The nearby laboratory also houses various types of large-scale experimental equipment, which will be subject to additional interference from non-experimental personnel walking randomly during the experiment. Four UWB base stations are deployed in the environment, and experimental personnel carrying UWB tags walk around the corridor atrium at a speed of 1 m / s.

[0108] like Figure 8 and Figure 9 As shown, this embodiment illustrates the localization trajectories and CDF curves of each algorithm; from Figure 8 It can be seen that the positioning trajectories of the KF-LS, KF-Chan, and KF-Taylor algorithms fluctuate significantly, while the positioning trajectory of the GFTML algorithm used in this invention is relatively smooth. Figure 9As can be seen from Table 1, the GFTML algorithm used in this invention has achieved a significant improvement in positioning accuracy. The root mean square error is reduced by 64.0%, 51.8%, and 39.3% respectively compared with the traditional algorithm, demonstrating a clear performance advantage.

[0109] Table 1. Laboratory corridor location results

[0110]

[0111] In summary, the present invention provides a UWB indoor positioning method and system based on generative fingerprint traversal matching. First, hyperparameters are configured and a coarse virtual grid is divided. Then, a coarse fingerprint database is constructed based on the coarse virtual grid. Next, coarse grid point data is selected sequentially from the coarse fingerprint database according to index numbers. The difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point is calculated and used as the first distance difference. Subsequently, the Huber loss corresponding to the coarse grid point is calculated using the first distance difference. The three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss are then used as the coarse positioning position. The coarse positioning position is then used as the center, and a cubic region is defined in three-dimensional space. This cubic region is then divided into multiple fine virtual grids to obtain fine grid points. Next, an index number is assigned to each fine grid point, and the three-dimensional coordinates of the fine grid point are recorded. The difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point is calculated and used as the second distance difference. Finally, the Huber loss corresponding to the fine grid point is calculated using the second distance difference, and the grid point with the minimum Huber loss is selected. Fine-grained grid points are used as fine-grained positioning locations, and these locations are then used as the target locations to complete UWB indoor positioning operations. This effectively realizes the UWB indoor positioning method and system, which can divide the target positioning space into three-dimensional virtual grid units for sub-meter-level indoor positioning accuracy. By constructing a robust mapping between position-related features and spatial coordinates, the dependence on NLOS recognition or error modeling is eliminated, resulting in better adaptability and accuracy. The Huber loss function can nonlinearize the distance difference, thereby mitigating the impact of abnormal measurements during traversal matching. At the same time, the loss value can more accurately describe the difference between the real location and the virtual grid points, thus achieving accurate matching and determining the target location. This not only avoids the cumbersome workload of manually collecting fingerprint data and training recognition models, but also significantly reduces the impact of ranging errors on positioning. This invention can meet the requirements of real-time performance and efficiency in practical scenarios and can be widely applied in smart cities, autonomous driving, and emergency rescue fields. This provides important technical support and theoretical basis for high-precision positioning in complex indoor environments.

[0112] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A UWB indoor positioning method based on generative fingerprint traversal matching, characterized in that: Comprising the following steps, Step A, configuring hyperparameters and dividing coarse virtual grid, the specific steps are as follows, Step A1, configuring hyperparameters, the hyperparameters include Huber loss function threshold δ, coarse grid side length C and fine grid side length F, the Huber loss function threshold δ is specifically valued according to the ranging error level, the coarse grid side length C is specifically valued according to the balance of coarse positioning and fine positioning calculation amount, and the fine grid side length F is specifically valued according to the target positioning accuracy of the user; Step A2, dividing coarse virtual grid, the specific steps are as follows, Step A21, expanding the base station space and taking it as the target location space TLS, specifically expanding the planar boundary of the base station group to 2 times the original length in the planar dimension, and expanding the height boundary from the ground upwards to 1.5 times the height of the highest base station, the edge length of the target location space TLS in three-dimensional directions can be represented as L x , L y , and L z ; Step A22, dividing the target positioning space TLS into a plurality of three-dimensional virtual grids with a side length of C and retaining the sub-grid of less than one complete grid space after division, and then taking all the grid vertices after gridding as coarse grid points, as shown in formula (1), （1） wherein is the total number of coarse grid points, is a ceiling function; Step A23, starting from the grid point of the smallest three-dimensional coordinate in the target positioning space TLS, each coarse grid point is assigned a unique index number in dictionary order, and the three-dimensional coordinates corresponding to each coarse grid point are recorded while sequentially numbering along the x-axis, y-axis and z-axis; Step B, constructing a coarse fingerprint database according to the coarse virtual grid, the specific steps are as follows, Step B1, taking each coarse grid point as a reference position, respectively extracting the distance feature corresponding to each coarse grid point, wherein the Euclidean distance between the ith coarse grid point and the kth base station As shown in formula (2), （2） wherein, and Xi and Xk are the third dimension coordinates of the ith coarse grid point and the kth base station, respectively. Step B2, after obtaining the Euclidean distance of all coarse grid points to each base station, a coarse fingerprint database can be constructed, the coarse fingerprint database is stored in a structured table form, each row of the coarse fingerprint database corresponds to a coarse grid point fingerprint information, the first column of the coarse fingerprint database is the coarse grid point index number, the second column of the coarse fingerprint database is the coarse grid point three-dimensional coordinates, and each column of the coarse fingerprint database is the Euclidean distance between the coarse grid point and each base station, all data in the coarse fingerprint database are stored with three decimal places accuracy; Step C, selecting the coarse grid point data from the coarse fingerprint database according to the index number in turn, and calculating the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point as the first distance difference value, wherein the difference between the measured distance of the kth base station and the Euclidean distance of the corresponding coarse grid point to the Euclidean distance of the corresponding base station As shown in formula (3),​ （3）； Step D, calculating the Huber loss of the coarse grid point corresponding to the first distance difference value, and taking the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss as the coarse positioning position, the specific steps are as follows, Step D1, calculating the Huber loss of the coarse grid point corresponding to the first distance difference value, specifically setting the total number of first distance difference values as M, and then calculating the Huber loss of the coarse grid point corresponding to each first distance difference value, as shown in formula (4), （4） wherein, is the Huber loss for the coarse grid point; Step D2, taking the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss as the coarse positioning position, specifically, after obtaining the Huber loss corresponding to all coarse grid points, the minimum Huber loss is found by traversing according to the index number, and the three-dimensional coordinates of the coarse grid point corresponding to the minimum Huber loss are selected as the coarse positioning position wherein the Huber loss is used to quantitatively describe the similarity between the target position and the coarse grid point; Step E, taking the coarse positioning position as the center and setting a cubic region in the three-dimensional space, and then dividing the cubic region into a plurality of fine virtual grids and obtaining fine grid points; Step F, assigning index number to each fine grid point and recording the three-dimensional coordinates of the fine grid point, and then calculating the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point as the second distance difference value; Step G, calculating the Huber loss of the fine grid point corresponding to the second distance difference value and taking the fine grid point with the minimum Huber loss as the fine positioning position, and then taking the fine positioning position as the target position and completing the UWB indoor positioning work.

2. The UWB indoor positioning method based on generative fingerprint traversal matching according to claim 1, characterized in that: Step E, setting a cubic region in three-dimensional space with the coarse positioning position as the center, and dividing the cubic region into a plurality of fine virtual grids to obtain fine grid points, the specific steps are as follows, Step E1, set a cubic region in three-dimensional space with the coarse positioning position as the center, specifically set a cubic region with the coarse positioning position as the center and with a side length of C in three-dimensional space; Step E2, dividing the cubic region into a plurality of fine virtual grids to obtain fine grid points, specifically, dividing the cubic region with a side length of C into a plurality of fine virtual grids with a side length of F to obtain fine grid points, wherein if part of the space of the cubic region exceeds the target positioning space TLS, the exceeding space part is removed, and the process of obtaining the fine grid points is shown in formula (5), （5） wherein, is the total number of fine grid points obtained.

3. The UWB indoor positioning method based on generative fingerprint traversal matching according to claim 2, characterized in that: Step F, assigning an index number to each fine grid point and recording the three-dimensional coordinates of the fine grid point, and then calculating the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point as the second distance difference, the specific steps are as follows, Step F1, assigning an index number to each fine grid point and recording the three-dimensional coordinates of the fine grid point, specifically, starting from the grid point with the smallest three-dimensional coordinates in the target positioning space TLS, assigning a unique index number to each fine grid point in dictionary order, and recording the three-dimensional coordinates corresponding to each fine grid point while sequentially numbering along the x-axis, y-axis and z-axis; Step F2, calculate the difference between the measured distance of each base station and the corresponding fine grid point Euclidean distance as the second distance difference value, wherein the measured distance of the kth base station the difference between the Euclidean distance of the jth fine grid point to the corresponding base station As shown in formula (6), （6） wherein, is the three-dimensional coordinate of the jth fine grid point.

4. The UWB indoor positioning method based on generative fingerprint traversal matching according to claim 3, characterized in that: Step G, calculating the Huber loss corresponding to the fine grid point using the second distance difference, and taking the fine grid point with the smallest Huber loss as the fine positioning position, and then taking the fine positioning position as the target position to complete the UWB indoor positioning task, wherein the Huber loss corresponding to the fine grid point is calculated using the second distance difference, and the total number of the second distance difference is N, and the Huber loss corresponding to the fine grid point in each second distance difference is calculated, as shown in formula (7), （7） wherein, is the Huber loss for the fine grid point.

5. A UWB indoor positioning system based on generative fingerprint traversal matching, the specific positioning procedure of the UWB indoor positioning system is based on the UWB indoor positioning method of any one of claims 1-4, characterized in that: The system comprises a coarse virtual grid division module, a coarse fingerprint database construction module, a first distance difference calculation module, a coarse positioning position obtaining module, a fine virtual grid division module, a second distance difference calculation module, and a fine positioning position obtaining module, wherein the coarse virtual grid division module is configured to configure hyperparameters and divide coarse virtual grids; The coarse fingerprint database construction module is configured to construct a coarse fingerprint database according to the coarse virtual grids; The first distance difference calculation module is configured to sequentially select coarse grid point data from the coarse fingerprint database according to the index number, and calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding coarse grid point as the first distance difference; The coarse positioning position obtaining module is configured to calculate the Huber loss corresponding to the coarse grid point using the first distance difference, and take the three-dimensional coordinates of the coarse grid point with the smallest Huber loss as the coarse positioning position; The fine virtual grid division module is configured to set a cubic region in three-dimensional space with the coarse positioning position as the center, and divide the cubic region into a plurality of fine virtual grids to obtain fine grid points; The second distance difference calculation module is configured to assign an index number to each fine grid point and record the three-dimensional coordinates of the fine grid point, and then calculate the difference between the measured distance of each base station and the Euclidean distance of the corresponding fine grid point as the second distance difference; The fine positioning position obtaining module is configured to calculate Huber loss corresponding to each fine grid point by using the second distance difference value, and take the fine grid point with the minimum Huber loss as the fine positioning position, and then take the fine positioning position as the target position and complete the UWB indoor positioning task.

Images