Location fingerprint positioning method based on WKNN algorithm

Abstract

The invention relates to the technical field of indoor positioning, and provides a location fingerprint positioning method based on big data mining, which comprises the following steps: extracting sampling point data and base station information from a test area; sampling point preprocessing is carried out by adopting a grid division method, and a cell identifier is obtained; constructing a Radio Map based on a preprocessing result; and based on Radio Map, a WKNN algorithm is combined with a failure strategy to obtain a positioning result. According to the neighbor cell screening method provided by the invention, the accuracy and reliability of the reference points in the positioning process are ensured, and in addition, a double Radio Map construction mode is adopted, so that the comprehensiveness and dynamic updating capability of the model are ensured, and the calculation efficiency and precision during positioning are also improved. And the WKNN algorithm is adopted in the positioning stage, so that the positioning precision can be effectively improved. Meanwhile, by setting a positioning failure strategy, the condition of positioning failure can be identified and avoided in time, and the reliability and stability of positioning are improved.

Description

Technical Field

[0001] The present invention relates to the field of indoor positioning technology, and in particular to a position fingerprint positioning method based on a WKNN algorithm. Background Art

[0002] Location fingerprinting technology, as an important indoor positioning method, has garnered widespread attention and rapid development in recent years. Its basic principle is to collect signal strength information from ambient signal sources (such as Wi-Fi, Bluetooth, and cellular networks) to construct a location fingerprint database (Radio Map). During positioning, the signal characteristics of the target location are matched with the fingerprint data in the database to determine the location. However, this technology still faces several challenges. First, the increasing complexity of indoor environments, such as multipath effects and signal attenuation, significantly impacts the stability and consistency of signal fingerprints, making it difficult to further improve positioning accuracy. Second, the era of big data places higher demands on the real-time and dynamic adaptability of positioning systems. Traditional Radio Map construction and updating methods cannot quickly respond to environmental changes and the addition of new data. Furthermore, existing positioning algorithms are prone to positioning failures or significant accuracy degradation in complex scenarios, such as those with sparse reference points and severe signal interference. Despite extensive research in signal processing, fingerprint extraction, and algorithm optimization, improving the system's real-time performance, reliability, and adaptability while maintaining positioning accuracy remains a key challenge in location fingerprinting technology. Summary of the Invention

[0003] In view of this, the present invention proposes a location fingerprint positioning method based on the WKNN algorithm. Through innovative grid division preprocessing, cell identification acquisition, dual Radio Map construction and update, and WKNN algorithm combined with failure strategy, it effectively solves the problems of slow positioning speed, poor real-time performance, low accuracy, and insufficient reliability in the existing technology, and significantly improves the performance of the positioning system.

[0004] To achieve the above objectives, the present invention proposes a location fingerprint positioning method based on the WKNN algorithm, which is characterized by comprising: Extract sampling point data and base station information from the test area; The grid division method is used to pre-process the sampling points and obtain the cell identification; Build Radio Map based on preprocessing results; Based on the Radio Map, the WKNN algorithm is used in combination with the failure strategy to obtain the positioning result.

[0005] Furthermore, the process of preprocessing the sampling points using the grid division method includes: The reference point is selected as the grid origin, and the longitude and latitude of the reference point are gridded relative to the longitude and latitude of the reference point. Based on the gridding results, the positions of all reference points fall on the grid nodes.

[0006] Furthermore, the process of obtaining the cell identity includes: Obtain the primary cell identifier, decompose the primary cell identifier into ENBID and CELLID, search the engineering parameter table for the cell corresponding to ENBID and CELLID, and obtain the geographic coordinates of the base station where the primary cell is located; Obtaining the EARFCN and PCI of the neighboring cell based on the sampling point data, searching for the cell corresponding to the EARFCN and PCI in the engineering parameter table, and calculating the physical distance between the corresponding cell and the primary cell; The cell with the shortest physical distance is taken as the neighboring cell of the primary cell, and the neighboring cell ENBID and CELLID are obtained and converted to obtain the neighboring cell identifier.

[0007] Furthermore, the process of constructing a Radio Map based on the preprocessing results includes: A full radio map and a statistical radio map are constructed respectively, the full radio map is used for model establishment and reference point update, and a statistical mean of RSRP of the same reference point and the same cell in the full radio map is obtained based on the statistical radio map.

[0008] Furthermore, during the reference point update process, if there is no reference point at the position of the new sampling point, a new reference point is created; if there is a reference point at the position of the new sampling point, the new sampling point is merged with the original reference point.

[0009] Furthermore, the process of merging the new sampling point with the original reference point includes: In the case where the cell at the sampling point is a new cell: if the number of cells stored in the reference point is less than 10, the new cell is directly added; if the number of cells stored in the reference point is full of 10, the RSRP of the new cell is compared with the average RSRP of the existing cells at the reference point. If the RSRP of the new cell is smaller, it is directly discarded; otherwise, the cell with the smallest RSRP among the existing cells is removed and the new cell is added; If the cell at the sampling point is an existing cell: if the number of RSRPs stored in the existing cell is less than 10, a new RSRP is directly added; if the RSRP stored in the existing cell is full, the average of the original 10 RSRPs and the RSRP at the sampling point is calculated, and the 10 RSRPs closest to the average are selected for storage.

[0010] Furthermore, the process of obtaining positioning results using the WKNN algorithm includes: Filtering the Radio Map based on the cell identifier to obtain a reference point set; Calculating the Euclidean distance between the reference point in the reference point set and the point to be located in the signal space; Select several reference points with the smallest signal difference norm and perform weighting to obtain the positioning result.

[0011] Furthermore, the process of screening the Radio Map based on the cell identifier includes: obtaining a set of cells to be located and each reference point cell set, screening out all reference points in the reference point cell set that include the set of cells to be located, and obtaining the reference point set.

[0012] Furthermore, the failure strategy includes: If the number of reference points in the reference point set is less than 3, it is determined that the positioning has failed; Calculate the distances between several reference points with the smallest signal difference norm. If there is a reference point with a distance greater than 80, it is considered a positioning failure. When the above two conditions are met at the same time, a positioning result is generated, and it is determined whether there is a reference point in the grid where the positioning result is located. If not, it is determined that the positioning has failed.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention pre-processes sampling points through a gridding method, effectively solving the problems of slow positioning and poor real-time performance caused by excessive sampling and reference points in outdoor environments. This gridding method quickly determines which points are close to each other, allowing for the merging of adjacent reference points, reducing redundant calculations and improving positioning efficiency. Furthermore, it removes redundant decimal places in the latitude and longitude values ​​of sampling points, avoiding redundant data interference that does not contribute to positioning accuracy, further improving positioning precision.

[0014] Secondly, the neighboring cell screening method proposed in the present invention ensures the accuracy and reliability of the reference points during the positioning process, avoids positioning deviations caused by erroneous cell information, provides a high-quality data basis for subsequent positioning calculations, and significantly improves the accuracy and stability of positioning.

[0015] Thirdly, this invention utilizes a dual Radio Map construction approach, ensuring the comprehensiveness and dynamic update capabilities of the model while also improving computational efficiency and accuracy during positioning. During the reference point update process, detailed rules for merging new sampling points with existing reference points are designed. This allows both new and existing cells to be rationally screened and updated based on the statistical characteristics of RSRP, ensuring that the Radio Map always maintains the latest and most valid data, thereby enhancing the adaptability and accuracy of the positioning system.

[0016] Finally, the present invention utilizes the WKNN algorithm combined with a failure strategy during the positioning phase. The WKNN algorithm calculates the Euclidean distance in signal space between a reference point and the point to be located and selects several reference points with the smallest signal difference norm for weighted positioning, effectively improving positioning accuracy. Furthermore, the implementation of three positioning failure strategies enables timely identification and avoidance of positioning failures, improving positioning reliability and stability and reducing the probability of mispositioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Various other advantages and benefits will become apparent to those skilled in the art by reading the detailed description of the preferred embodiment below. The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention. In the accompanying drawings: Figure 1 This is a schematic diagram of the operating principle framework of position fingerprint positioning in an embodiment of the present invention; Figure 2 Schematic diagram of sampling point density analysis in a test area according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a positioning failure situation in an embodiment of the present invention; Figure 4 This is a flow chart of the location fingerprint positioning method proposed in the present invention; Figure 5 Schematic diagram of the Radio Map of the experimental area in an embodiment of the present invention; Figure 6 Schematic diagram of test points in the experimental area in an embodiment of the present invention; Figure 7 1CDF curve schematic diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, unless there is a conflict, the embodiments of the present disclosure and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0019] Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art. Although only preferred methods and materials are described herein, any methods and materials similar or equivalent to those described herein may also be used in the practice or testing of the present invention. All documents mentioned in this specification are incorporated by reference to disclose and describe the methods and / or materials associated with the documents. In the event of any conflict with any incorporated document, the contents of this specification shall prevail.

[0020] This embodiment proposes a location fingerprint positioning method based on the WKNN algorithm, such as Figure 1 As shown in Figure 2, the fingerprint positioning process can be roughly divided into an offline phase and an online phase. The offline phase is the phase of establishing the Radio Map, and the online phase is the phase of loading the Radio Map and using it for positioning.

[0021] Offline phase - establishing a Radio Map (1) Data collection and sampling point density analysis 1) Data acquisition module The data collection module includes two parts: extracting DT and CQT sampling points from the live network platform and extracting base station operating parameters. The module extracts the current latitude and longitude, the downlink signal strength of the sampling point's broadcast channel, the sampling point's home cell name, the home cell's ECRFCN, and PCI. It also extracts basic LTE network base station information from LTE operating parameters, including the base station's latitude and longitude, cell name, cell ECRFCN, and cell PCI.

[0022] 2) Sampling point density analysis Table 1

[0023] In this embodiment, an area in Jiangbei was selected as the test area to observe the density of sampling points. The area of ​​this area is approximately 16 km*24 km.

[0024] The area is divided into grids with a side length of 40 meters. The number of sampling points in each grid is calculated and plotted to obtain the left figure. The number of samples in the grid is represented by different colors. The closer the color is to yellow, the greater the density of sampling points.

[0025] Depend on Figure 2 It can be seen that most areas have no sampling points, and the number of sampling points in the grids with sampling points is also relatively small. Only a few scattered areas are yellow, that is, the sampling point density is very high.

[0026] (1) Sampling point preprocessing The data preprocessing module includes two parts: gridding of sampling point locations and labeling of sampling point cells.

[0027] Since the sampling points are accurate to 10⁻⁷ (longitude / latitude), which translates to a distance of approximately 0.01 meters, civilian GPS accuracy is only 3 meters, far below 0.01 meters. Therefore, the reported GPS data contains multiple redundant bits. Furthermore, due to the large outdoor space and numerous sampling points, if the RadioMap is used directly without processing, the excessive number of reference points will inevitably result in slow positioning and poor real-time performance. Therefore, adjacent reference points must be merged. This merging requires first determining which points are close to each other. Direct distance calculation is computationally expensive, so a gridding approach is used. Sampling points that fall on the same grid node are considered sufficiently close to be merged. The specific method is as follows: First, a reference point is selected as the grid origin. In principle, any reference point can be chosen. The algorithm selects the reference point (45°N, 126°E). The longitude and latitude of the reference point are gridded relative to the longitude and latitude of the reference point, using a 1m×1m grid size. A 1m grid corresponds to a longitude deviation of 1.3×10⁻⁵ and a latitude deviation of 9×10⁻⁶.

[0028]

[0029] After the above processing, the positions of all reference points are on the nodes of the grid, which is convenient for filtering, merging, and updating reference points.

[0030] Next, obtain the cell identifier (ECI). The ECI is a decimal number consisting of ENBId and CellId. It uniquely represents a cell. The calculation formula is as follows:

[0031] In the engineering parameter table, ECI is not given, but the cell's CGI (in the form of "aaa-bb-cccccc-ddd") is given. There is also a relationship between CGI, ENBId, and CellId. The last two fields of CGI are ENBId and CellId, respectively. The above formula is still used to convert them into ECI.

[0032] It is relatively simple to obtain the ECI of the main cell, and it can be obtained directly from the ECI field of the sampling point. As for obtaining the ECI of the neighboring cell, the sampling point does not provide the ECI of the neighboring cell, only the EARFCN and PCI. It is necessary to find the matching cell in the engineering parameter table according to these two parameters. The main steps are as follows: First, decompose the ECI of the main cell into ENBID and CELLID, find the cell corresponding to ENBID and CELLID in the engineering parameter table, and obtain the geographic coordinates of the base station where the main cell is located; then read the EARFCN and PCI of the neighboring cell from the sampling point data, and find the cell corresponding to EARFCN and PCI in the engineering parameter table. Due to the reuse of the center frequency, multiple cells may be matched through the above method. Next, calculate the physical distance between the matched cell and the main cell. The longitude and latitude conversion distance formula is as follows:

[0033] Where R is the radius of the Earth, A and B are the two reference points for distance calculation, LatA and LatB are the latitudes of points A and B, and LonA and LonB are the longitudes of points A and B. The cell with the shortest distance is the neighboring cell. Obtain the ENBID and CELLID of that cell and convert them to the ECI.

[0034] Description: Reasons for meshing: The sampling point a is accurate to 10-7 (longitude / latitude). 10-7 longitude is converted to a distance of 0.008 meters, and 10-7 latitude is converted to a distance of 0.011 meters. However, the accuracy of civilian GPS is only 3 meters, far from 0.008 meters and 0.011 meters. Therefore, excessive decimal places do not contribute to position accuracy. Excess decimal places can be removed through gridding.

[0035] b. The outdoor environment is large and has many sampling points. If it is used directly as a RadioMap without any processing, the excessive number of reference points will inevitably result in slow positioning speed and poor real-time performance. Therefore, adjacent reference points must be merged. To merge, it is first necessary to determine which points are close to each other. The direct distance calculation method has too much computational overhead, so a gridding method is used. Sampling points that fall on the same grid node are considered close enough to be merged.

[0036] (2) Create a Radio Map There are two ways to establish a Radio Map: initial creation and Radio Map update, both using the same strategy. Initial creation is a special case of an update; it simply initializes the old Radio Map to an empty matrix. The following describes the Radio Map update method. The key concept is to establish two Radio Maps: a full Radio Map for model establishment and updates, storing multiple RSRP values ​​for a cell at a given reference point; and a statistical Radio Map for positioning, taking the statistical mean of the RSRP values ​​for the same reference point and cell in the full Radio Map.

[0037] The following two situations may occur during the reference point update process: the location of the new sampling point does not have a reference point, so a new reference point can be created; or the location of the new sampling point already has a reference point, so the new sampling point needs to be merged with the original reference point. The specific merging method is as follows: For each cell at the new sampling point, the following judgment and merging are performed: If the cell at the sampling point is a new cell: If the number of cells stored in the reference point is less than 10, that is, not full, the new cell is added directly. If the cell storage at the reference point is full, the RSRP of the new cell is compared with the average RSRP of the existing cells at the reference point. If the new cell has the lowest RSRP, it is discarded. Otherwise, the cell with the lowest RSRP among the existing cells is removed and the new cell is added. If the cell at the sampling point is an existing cell: If the number of RSRPs stored in the cell is less than 10, that is, not full, the new RSRP is added directly. If the RSRP storage for the cell is full, the average of the 11 existing RSRPs (10 RSRPs and the RSRP at the sampling point) is calculated, and the 10 RSRPs closest to the average are selected and stored.

[0038] Online stage - WKNN algorithm The preprocessing phase in the online phase is identical to the offline phase. After preprocessing, the test data is matched with the radio map created in the offline phase, and positioning is performed to obtain the result. The online phase mainly includes the following three aspects: reference point matching (coarse matching); KNN positioning (fine positioning); and due to the low density of reference points, three positioning failure detection strategies are added to address extreme situations that may occur during the matching and positioning process.

[0039] Coarse positioning uses the cell identification number to filter the Radio Map. The specific process is as follows: obtain the cell set T of the point to be located, obtain the cell set Ri of each reference point, filter out all reference points whose Ri contains T, and obtain the reference point set P.

[0040] Precise positioning is KNN positioning. For the matched reference point, the Euclidean distance between it and the point to be positioned is calculated:

[0041] Select K reference points with the smallest signal difference 2 norm (set K=3) and perform KNN weighting according to the following formula to obtain the positioning result.

[0042]

[0043] like Figure 3 As shown, this embodiment sets three positioning failure strategies: one for the coarse positioning stage and two for the fine positioning stage. The number of reference points matched in the coarse positioning stage should be greater than or equal to 3. If it is less than 3, positioning fails (KNN fine positioning sets K=3). For the matched reference points, the reference points P1, P2, and P3 with the smallest 2-norm difference between the signal and the point to be positioned are selected. The distances l1, l2, and l3 between the three reference points are calculated respectively. If at least one of the distances is greater than 80m, positioning fails. If both of the above conditions are met, the positioning result can be obtained according to the KNN algorithm with K=3, as shown in the figure. Figure 3 The black dots are matching reference points, the red dots are positioning results, the blue areas are areas with reference points, and the white areas are areas without reference points. Next, we determine whether there are reference points within the 40m×40m grid where the positioning result is located. If not, positioning fails. In the figure, there are no reference points within the grid where the positioning result is located, so positioning fails.

[0044] The implementation process of the WKNN algorithm proposed in this embodiment is as follows: Figure 4 As shown in the figure, the algorithm combined with the failure strategy shows significant technical effects in positioning accuracy, positioning efficiency, reliability and adaptability, providing an efficient, accurate and reliable solution for the field of indoor positioning technology, which has broad application prospects and important practical significance.

[0045] Algorithm correctness testing Select an area to test the correctness of the algorithm.

[0046] (1) The area range is as follows: Longitude: 126.52 ~ 126.56; Latitude: 45.86 ~ 45.88; About 3km*2km.

[0047] (2) The constructed Radio Map is as follows Figure 5 As shown; (3) Regional test points such as Figure 6 As shown; (4) CDF curve is as follows Figure 7As shown in the figure, the allpoint line is the overall CDF curve, and the remaining ones are the CDF curves of the test points within the corresponding reference point density grid. It can be seen that the higher the density of reference points near the point to be located, the higher the positioning accuracy. The backfill ratio is 52.5%, and the positioning error of 78.03% of the points is less than 40m.

[0048] Figure 7 The allpoint line in the middle is the overall CDF curve, and the others are the CDF curves for the test points within the corresponding reference point density grid. It can be seen that a higher density of reference points near the point to be located increases positioning accuracy. The backfill ratio is 52.5%, and 78.03% of the points have positioning errors less than 40 meters.

[0049] In summary, experimental tests demonstrate that within the selected area, this algorithm achieves a backfill rate of 52.5%, with 78.03% of points achieving positioning errors less than 40 meters. Furthermore, the higher the density of reference points near the point to be located, the higher the positioning accuracy. This demonstrates that this algorithm significantly improves positioning accuracy and can meet the demand for high-precision positioning in practical applications. Furthermore, thanks to the aforementioned optimization measures, positioning efficiency has also been significantly improved, enabling real-time, rapid positioning services suitable for a variety of complex indoor and outdoor environments.

[0050] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A location fingerprint positioning method based on the WKNN algorithm, characterized in that: include: Extract sampling point data and base station information from the test area; The grid division method is used to pre-process the sampling points and obtain the cell identification; Build Radio Map based on preprocessing results; Based on the Radio Map, the WKNN algorithm is used in combination with the failure strategy to obtain the positioning result.

2. The position fingerprint positioning method based on the WKNN algorithm according to claim 1 is characterized in that: The process of sampling point preprocessing using the grid division method includes: The reference point is selected as the grid origin, and the longitude and latitude of the reference point are gridded relative to the longitude and latitude of the reference point. Based on the gridding results, the positions of all reference points fall on the grid nodes.

3. The position fingerprint positioning method based on the WKNN algorithm according to claim 1 is characterized in that: The process of obtaining the cell identity includes: Obtain the primary cell identifier, decompose the primary cell identifier into ENBID and CELLID, search the engineering parameter table for the cell corresponding to ENBID and CELLID, and obtain the geographic coordinates of the base station where the primary cell is located; Obtaining the EARFCN and PCI of the neighboring cell based on the sampling point data, searching for the cell corresponding to the EARFCN and PCI in the engineering parameter table, and calculating the physical distance between the corresponding cell and the primary cell; The cell with the shortest physical distance is taken as the neighboring cell of the primary cell, and the neighboring cell ENBID and CELLID are obtained and converted to obtain the neighboring cell identifier.

4. The position fingerprint positioning method based on the WKNN algorithm according to claim 1 is characterized in that: The process of building a Radio Map based on the preprocessing results includes: A full radio map and a statistical radio map are constructed respectively, the full radio map is used for model establishment and reference point update, and a statistical mean of RSRP of the same reference point and the same cell in the full radio map is obtained based on the statistical radio map.

5. The position fingerprint positioning method based on the WKNN algorithm according to claim 4 is characterized in that: During the reference point update process, if there was no reference point at the location of the new sampling point, a new reference point is created; if there was a reference point at the location of the new sampling point, the new sampling point is merged with the original reference point.

6. The position fingerprint positioning method based on the WKNN algorithm according to claim 5 is characterized in that: The process of merging the new sample point with the original reference point includes: In the case where the cell at the sampling point is a new cell: if the number of cells stored in the reference point is less than 10, the new cell is directly added; if the number of cells stored in the reference point is full of 10, the RSRP of the new cell is compared with the average RSRP of the existing cells at the reference point. If the RSRP of the new cell is smaller, it is directly discarded; otherwise, the cell with the smallest RSRP among the existing cells is removed and the new cell is added; If the cell at the sampling point is an existing cell: if the number of RSRPs stored in the existing cell is less than 10, a new RSRP is directly added; if the RSRP stored in the existing cell is full, the average of the original 10 RSRPs and the RSRP at the sampling point is calculated, and the 10 RSRPs closest to the average are selected for storage.

7. The position fingerprint positioning method based on the WKNN algorithm according to claim 1 is characterized in that: The process of obtaining positioning results using the WKNN algorithm includes: Filtering the Radio Map based on the cell identifier to obtain a reference point set; Calculating the Euclidean distance between the reference point in the reference point set and the point to be located in the signal space; Select several reference points with the smallest signal difference norm and perform weighting to obtain the positioning result.

8. The position fingerprint positioning method based on the WKNN algorithm according to claim 7 is characterized in that: The process of screening the Radio Map based on the cell identifier includes: obtaining a set of cells to be located and each reference point cell set, screening out all reference points in the reference point cell set that include the set of cells to be located, and obtaining the reference point set.

9. The position fingerprint positioning method based on the WKNN algorithm according to claim 7, characterized in that: The failure strategies include: If the number of reference points in the reference point set is less than 3, it is determined that the positioning has failed; Calculate the distances between several reference points with the smallest signal difference norm. If there is a reference point with a distance greater than 80, it is considered a positioning failure. When the above two conditions are met at the same time, a positioning result is generated, and it is determined whether there is a reference point in the grid where the positioning result is located. If not, it is determined that the positioning has failed.

Images