A 5g outdoor positioning method based on rss

Abstract

The application discloses a 5G outdoor positioning method based on RSS, which is characterized by a cooperative positioning method constructed by a master base station, a neighbor base station, a mobile terminal and a background server. The method uses the 5G base station deployed, analyzes the RSS (RSRP, RSRQ) information of the master base station and the neighbor base station through the MR data reported by the user, calculates the distance from each source point to the master base station and the neighbor base station, processes the distance ratio function, obtains the logarithmic form of the estimated distance ratio, uses the least square method to obtain the preliminary positioning coordinates, and then uses the MAP-LM algorithm to bring the preliminary coordinates into the iterative equation. If the condition is met, the iteration is exited, and the final positioning coordinates are obtained. Compared with the prior art, the application has higher robustness and positioning accuracy, does not need to lay a large number of hardware devices, is easy to popularize, and has great advantages and commercial prospects in the complex outdoor positioning application scene and NLOS scene.

Description

Technical Field

[0001] This invention relates to the field of positioning, navigation and monitoring technology, specifically a 5G outdoor positioning method based on RSS. Background Technology

[0002] The world's information technology is developing rapidly. Outdoor positioning, represented by technologies such as BeiDou navigation and GPS, and indoor positioning, represented by technologies such as 5G fingerprint databases, TDOA, and AOA, are collectively referred to as wireless positioning technologies. While BeiDou navigation and GPS positioning technologies are becoming increasingly mature, a series of problems have also emerged. One of these problems is the inability to obtain accurate outdoor positioning in environments where users cannot use these two methods. This is because users generally turn off GPS and other positioning functions on their phones when not needed, both because they don't require it and because it consumes a lot of power. During the pandemic, many people did not turn on GPS, making accurate positioning impossible. Therefore, the demand for 5G outdoor positioning has become increasingly strong, as people currently keep their mobile networks on and do not turn them off, allowing 5G positioning to obtain a user's location information.

[0003] Fingerprint database technology, radio frequency identification (RFID), millimeter wave technology, Wi-Fi technology, and UWB and LoRa are all used for outdoor positioning. However, their coverage is not wide, their applicable scenarios are not complex, and their testing range is small, so they cannot represent complex real-world environments, or they require large deployment volumes, making large-scale commercial use impossible. Wireless communication technology, on the other hand, can be commercially deployed, provides high-bandwidth transmission, and is indispensable in people's lives. Before fourth-generation wireless communication technology, its outdoor positioning accuracy was poor and could not meet the needs of precise positioning. With the advent of fifth-generation mobile communication technology (5G), it provides greater bandwidth capacity, faster transmission speeds, and higher outdoor positioning accuracy than ever before.

[0004] In outdoor positioning, machine learning and deep learning are commonly used methods. However, for path loss calculations, current methods are primarily based on simulations. This is due to two main reasons: firstly, real-world outdoor environments are subject to numerous interference factors and drastic environmental changes, and secondly, the pre-deployed 4G base stations cannot meet positioning requirements. Thirdly, collecting real-world data is difficult because this data contains user privacy information. The advent of 5G technology has solved these problems. However, fingerprint database technology performs poorly outdoors, mainly because the outdoor environment is highly variable and covers a large area. Path loss calculations can handle the diverse and extensive outdoor environment.

[0005] In summary, environmental factors have a significant impact on outdoor wireless signals, and current technologies such as GPS cannot provide accurate positioning for situations like NLOS. Furthermore, mobile terminals (such as mobile phones) consume a lot of power and are usually turned off when not needed, making it impossible to determine the user's location when positioning is required. Summary of the Invention

[0006] The purpose of this invention is to provide a 5G outdoor positioning method based on RSS, addressing the shortcomings of existing technologies. This method employs a collaborative positioning approach using a master base station, neighboring base stations, a mobile terminal, and a backend server. It utilizes real user MR data for outdoor positioning. By parsing the user-reported MR data, information such as the RSS (RSRP, RSRQ) of the master and neighboring base stations is extracted. The distance from each source point to the master and neighboring base stations is calculated, and then processed using a distance ratio function to obtain the logarithmic form of the estimated distance ratio. Preliminary positioning coordinates are obtained using the least squares method, and the positioning results are optimized using the MAP-LM algorithm. This invention involves multiple samplings of the experimental scenario, storing error parameters and perturbation equations on the backend server. During positioning, the preliminary results obtained from the least squares algorithm are optimized to reduce the impact of environmental factors on positioning accuracy, improving positioning accuracy to a level superior to the 3GPP standard. It effectively protects user privacy information and eliminates the need for pre-measurement of the environment. The method is simple, effective, requires no additional hardware deployment, and does not require modifications to existing hardware, making it easy to promote. It has significant advantages and commercial prospects in high-precision indoor positioning applications.

[0007] The objective of this invention is achieved as follows: A 5G outdoor positioning method based on RSS, characterized by a collaborative positioning method constructed using a primary base station, neighboring base stations, a mobile terminal, and a backend server. Utilizing existing, deployed 5G base stations, it receives MR data from the user's mobile terminal using the MAP-LM algorithm to accurately locate the user. This algorithm uses the MR data reported by the user to accurately locate the user, specifically including the following steps:

[0008] Step 1: Data Preprocessing

[0009] The raw data is compared with the engineering parameters. If an unknown main base station or neighboring base station is found, it is removed. Data from different PCIs belonging to the same base station is removed. Empty data and duplicate data are removed. Data receiving less than three base stations is removed, thus obtaining preprocessed test data.

[0010] Step 2: Calculation of the estimated distance ratio

[0011] Based on the ratio of the distance to the base station to the distance to the mobile terminal, the estimated distance ratio between the two is calculated. Using characteristic parameters such as RSS, its logarithmic form is obtained through the estimated distance ratio formula.

[0012] Step 3: Calculation of preliminary positioning results

[0013] The estimated distance ratio is input into the least squares algorithm. The estimated matrix and error matrix are obtained through the estimated distance ratio, and the preliminary positioning coordinates are obtained.

[0014] Step 4: Obtaining the final positioning coordinates

[0015] The initial positioning results are fed into the MAP-LM algorithm for iteration. Through iteration, the final positioning information is obtained and sent back to the terminal server. The terminal server then sends the mobile terminal's unique identification code and time synchronization information back to the primary base station. The primary base station is then responsible for sending the final positioning results to the mobile terminal, thus providing the user with the precise coordinates of their location.

[0016] The main base station, neighboring base stations, and server terminal are connected via fiber optic cables. Users transmit signals wirelessly to the main base station and neighboring base stations. Without requiring environmental predictions, real user behavior can be directly tested to obtain raw MR data packets, which are then parsed into readable files. The raw data undergoes preprocessing, including: comparing the raw data with engineering parameters; removing any unknown main or neighboring base stations; removing data from different PCIs belonging to the same base station; removing empty or duplicate data; and removing data from fewer than three receiving base stations. These preprocessing steps primarily remove data that does not conform to the algorithm. After calculating the estimated distance ratio, the preprocessed data and the estimated distance ratio are compared using the least squares algorithm to obtain a preliminary positioning result. This result is then iteratively applied to the MAP-LM algorithm to obtain the final positioning result.

[0017] The data format of the MR is shown in Table 1 below:

[0018] Table 1 Data Format of MR

[0019]

[0020]

[0021] The estimated distance ratio is input into the least squares algorithm to obtain a one-dimensional column vector, which contains the coordinates of the target point and the path loss exponent n; the path loss consists of four parts: received signal strength, path loss exponent, logarithmic form of the distance ratio, and the influence of environmental factors. The path loss model is expressed as follows (1):

[0022]

[0023] Where, ρ a This is the average RSS of the receiving source; here, SS-RSRS is used, which is the average received signal power. ρ unknown d is the transmission power of the unknown base station; n is the path loss exponent; d m This is the distance between the receiving source and the base station; σ m This is the shadow fading coefficient, also known as the environmental factor. The environmental factors and distance d vary depending on the location. m The values ​​are all different. m starts from zero and goes up to (M-1), with the reference base station being the 0th one. For real outdoor environments, the antenna height is negligible, so three-dimensional space is not considered. Therefore, the positioning results do not include a Z-axis, i.e., a three-dimensional vertical axis.

[0024] The shadow fading coefficient d m This can be represented by (2) below:

[0025]

[0026] From equation (2) above, it can be seen that for different environments, the road loss index n and the shadow fading coefficient σ m They are all different, but for the overall environment, it is impossible to calculate all environmental parameters, and it is impossible to know d accurately. m The size of d is such that if some environmental parameters and the road loss index can be approximated as a distribution, and the calculation does not require knowing d... m Instead of determining the size, we fit it and estimate it.

[0027] The calculation of the least squares method and the formulas used in the calculation are as follows:

[0028] In real-world environments, traditional path loss calculation formulas often fail to meet the specific conditions required. Therefore, this paper derives a formula that eliminates the need for a path loss exponent n and a shadow fading coefficient σ. m The formula, and no need to calculate d. m The derivation process is as follows:

[0029] 1) The mean of RSS is ρ b And calculated by the following formula (3):

[0030]

[0031] Where, σ n d is the shadow fading coefficient; n This represents the ratio of the distances from each anchor base station to the source point.

[0032] 2) ρ a -ρ b Substituting into equation (3), we obtain the following equation (4):

[0033]

[0034] 3) Because in the same receiving source, ρ unknown They are different, so the reference point ρ unknown Consider it as ρ unknown_0 Treating other senders as ρ unknown_n And let Finally, we can derive the following equation (5):

[0035]

[0036] in,

[0037] 3) The estimated distance ratio is calculated as follows:

[0038] First, we can obtain the difference. RSRP Compared to distance The probability density function of the road loss index and shadow fading coefficient cannot be calculated for each region or even for each test point, so it needs to be represented by the probability density function of the following equation (6):

[0039]

[0040] The following equation (7) can be obtained:

[0041]

[0042] In the formula,

[0043] Thus, the maximum likelihood function of the distance ratio can be estimated through maximum likelihood estimation, and then the logarithmic form of the estimated distance ratio can be obtained by taking its partial derivative.

[0044] 4) Substitute the estimated distance ratio obtained above into the least squares method to obtain the final positioning result. The least squares method requires a result matrix, which needs to be assigned a column vector. For estimating the distance ratio, it can be expressed by the following equation (8):

[0045]

[0046] Where, for x m With y m In this context, ξ represents the test point with respect to the Mth neighboring base station, where M > 1. x1 represents the host base station, and x0 represents the x-coordinate of the test point. For ξ, all elements are unknowns. Furthermore, for the base station's transmitted signal, a circle is drawn about the base station's center. Ideally, the test point should lie on this circle, so a new result vector needs to be provided. By combining the distance ratio formula obtained earlier with the estimated distance ratio obtained in the previous section, we can obtain the following system of linear equations represented by equation (8-1):

[0047] Aξ=B (8-1).

[0048] Where A is the estimation matrix represented by the following equation (9):

[0049]

[0050] B is the error matrix represented by the following equation (10):

[0051]

[0052] For the two matrices A and B mentioned above, a system of linear equations about ξ can be obtained. By performing inverse processing on the matrices, the following equation (11) can be obtained:

[0053] ξ^=(A T YA) -1 A T YB (11).

[0054] Where γ is the identity matrix; A T A is the transpose of the matrix; -1 It is the inverse of the matrix.

[0055] Using the least squares method, we can obtain the final system of linear equations expressed in equation (11-1):

[0056] ξ^=(A T A) -1 A T B (11-1).

[0057] 5) The MAP algorithm, based on the least squares method, yields a one-dimensional column vector containing the coordinates of the target point and the path loss exponent n. Let ω = matrix[x, y, n]. transportThe algorithm iterates over ω until a relatively ideal target point is obtained. Iterating over test points requires knowing the position information of previous test points; otherwise, error distance comparison is impossible. However, for user predictions, coordinate information is unavailable, so this method cannot be used. Nevertheless, through extensive data analysis, it was found that the method performs well after a certain number of iterations, although it may get stuck in local minima. However, this method also yields relatively high accuracy, i.e., performing a fixed number of iterations and directly returning the iteration value without iterative judgment.

[0058] For difference RSRP In other words, the MAP algorithm for path loss can be represented by the following equation (12), and its maximum likelihood function based on the path loss index n can be obtained:

[0059]

[0060] The extreme points of the maximum likelihood function are calculated by taking the derivative of the function. These extreme points can be pseudo-extreme points, meaning there is no sign change around the extreme points. By taking the partial derivative of equation (12), we can obtain the following equation (12-1) to represent the difference. RSRP probability distribution:

[0061]

[0062] Here, n belongs to the set of integers M, that is, relative to neighboring base stations. Since the number of base stations is fixed, this distribution is 0 for a single base station.

[0063] Difference RSRP The probability distribution, after matrix processing, yields a diagonal matrix, which is expressed by the following equation (12-2):

[0064]

[0065] Treating ω as the model function and ω^ as the result of the previous iteration, we can obtain the curve fitting formula expressed by the following equation (13):

[0066]

[0067] For equation (13) above, j is a dynamic variable about the base station, and if difference RSRP Relatively peaceful, that is, difference RSRP | CDF 80％ <5, ω can be * It is considered an interference quantity and is not included in subsequent calculations. However, if the number of base stations is large, it cannot be considered a small interference quantity. ω *=ω^(n) transport ×W×ω^(n); for difference RSRP When the value is positive, the following radical expression (13-1) can be used for curve fitting:

[0068]

[0069] When difference RSRP When the value is negative, only the square root form of the following equation (13-2) can be used for curve fitting (but the square root form is more accurate):

[0070]

[0071] 6) As a model function, ω(n) - ω^(n) should be about difference. RSRP | i A one-dimensional column vector of the logarithmic distance ratio is required, and gradient descent processing is needed to approximate it to the point with the minimum error. When performing gradient descent, the direction with the largest descent slope should be selected for iteration to speed up program execution and improve iteration feasibility. The first-order partial derivative of its model function is obtained from the following equation (14):

[0072]

[0073]

[0074] Where θ is the step size function of the iteration, and for ease of calculation, let o T = matrix(ω(n)-ω^(n)) transport J is the Jacobian matrix, which is determined based on the road loss index, estimated distance ratio, and estimated coordinates (iteration). The Jacobian matrix needs to be updated during specific iterations. After solving the model using first-order partial derivatives, the... Called a disturbance The magnitude of the disturbance is determined by the road loss index.

[0075] 7) The update of the Jacobian matrix usually requires more than one iteration before it can be applied to the next calculation. Through the Gauss-Newton model, we know that for ω(n), ω(n+1) is obtained by adding the Jacobian matrix and the perturbation. According to the formula provided by Ψ(n), and let ω(n) = ω, we can obtain the curve fitting equation Ψ(n+1) expressed by the following equation (15):

[0076] Ψ(n+1)=ω T×W×ω+ω^T×W×ω^-2(ω T ×W×ω^-H) (15).

[0077] Where H is the perturbation function for Ψ(n+1), H=(ω-ω T )×W×J×hh T ×J T ×W×J×h, where h is the perturbation function for Ψ(n), can be calculated using the perturbation function provided by the Gauss-Newton model in the following equation (16):

[0078] {J T ×W×J}×h=J T ×W×(ω-ω T (16).

[0079] Different methods use different perturbation functions, but their underlying formulas are the same; only the constraints change. The constraint is the step size θ. From equation (16) above, we can obtain the difference in curve fitting expressed by equation (17):

[0080] Ψ(n)-Ψ(n+1)=2*(ω-ω T )×W×J×hh T ×J T ×W×J×h (17).

[0081] The above two methods, namely gradient descent and Gauss-Newton model, are used to obtain curve fitting formulas. However, if better positioning accuracy can be obtained, since the perturbation function of gradient descent and Gauss-Newton model is not what the algorithm needs, and their perturbation function is not as accurate as the Levengerg-Marquardt model, the following equation (18) is used to represent the perturbation of the LM algorithm:

[0082]

[0083] in, It is the damping coefficient for the disturbance.

[0084] Therefore, we can obtain the following equation (19) which represents the final calculation formula:

[0085]

[0086] in, It is an equation concerning the perturbation and the Jacobian matrix, etc.

[0087] 8) The following is given Components: This is the PDF formula for the road loss index.

[0088] The Jacobian matrix is ​​expressed as follows (20):

[0089]

[0090] For the Jacobian matrix, it needs to be updated at a specific number of iterations to ensure the feasibility of continuing iterations. The Jacobian matrix can be updated at even-numbered iterations and at the first iteration, or when the curve fitting result obtained after the iteration is worse than the original curve fitting result. Updating the Jacobian matrix involves substituting the result obtained in the previous step into equation (20) to perform the update operation. However, it cannot be updated at every iteration, as it is easy to fall into the trap of local minima.

[0091] At the start of the iteration, a coefficient of 7 needs to be determined. If the update factor l calculated during the iteration is less than 7, a good result is considered to have been obtained, and this result can be overwritten on the original result before proceeding to the next iteration. However, there are two methods to judge the result: first, if the difference between two results during the iteration is less than 1m, the iteration can be considered to have ended; or, check if the maximum value of the iteration parameter matrix is ​​less than a certain value, which is preset, i.e., max(J T ×W×(ω-ω T The algorithm is expressed by the following equation (21):

[0092]

[0093]

[0094] For l i If the value is greater than 1, it is considered an effective iteration, which can be considered superior to the previous estimation results. The updates of the remaining parameters are expressed as follows in equation (21-1):

[0095]

[0096] Among them, v i This is the step parameter, with an initial value of 0.

[0097] In other cases, it is necessary to... i Expanding this with other parameters for iteration, as shown in equation (21-2) below:

[0098]

[0099] This yields the final result: the MAP algorithm can be used to obtain the positioning location with maximum accuracy.

[0100] The main base station and neighboring base stations receive signals from user mobile terminals, including the mobile terminal's unique identifier, time information, and information such as RSRP, RSRQ, S1NR, BeamId, PLMN, ScTadv, Arfcn, Pci, MR.Longimde, MR.Latimde, MR.HAOA, and MR.VAOA. This information is then aggregated and sent to the cloud server.

[0101] The preprocessing operation of the raw data, which removes outliers, is divided into four steps: removing null and duplicate values, removing base stations that do not exist in the working parameters, removing data with fewer than 3 receiving base stations, and removing data received from different PCIs belonging to the same base station.

[0102] The ratio of the distance to the base station to the distance to the mobile terminal is calculated, and the estimated distance ratio between the two is obtained by using characteristic parameters such as RSS and the estimated distance ratio formula.

[0103] The least squares algorithm calculates the positioning coordinates for data that has undergone data preprocessing, obtains the estimation matrix and error matrix by estimating the distance ratio, and finally obtains the preliminary positioning coordinates.

[0104] The MAP-LM algorithm iteratively processes the preliminary positioning results, and by inputting the update factor and iteration constraints, the final positioning result can be obtained.

[0105] The final iteration result is output from the server terminal, passes through a base station near the mobile terminal, and is wirelessly transmitted to the mobile terminal via positioning coordinate information.

[0106] Compared with existing technologies, this invention has higher robustness and stable positioning accuracy, effectively protects users' privacy information, and optimizes the positioning results through the MAP-LM algorithm, making the positioning accuracy better than the 3GPP standard. It does not require the deployment of a large number of additional hardware devices or the modification of existing hardware devices, making it easy to promote. It has significant advantages and commercial prospects in complex outdoor positioning application scenarios and NLOS scenarios. Attached Figure Description

[0107] Figure 1 This is a system schematic diagram of the architecture of the present invention;

[0108] Figure 2 This is a schematic diagram of the test environment;

[0109] Figure 3 This is a flowchart of the present invention;

[0110] Figure 4 This is a diagram showing the positioning results of Example 1. Detailed Implementation

[0111] See Figure 1 This invention relates to a 5G outdoor positioning system comprising several neighboring base stations (commercial base stations) 1-6, a main base station 0, a mobile terminal 7, a terminal server 8, and a processing terminal 9. The neighboring base stations 1-6 transmit the user MR data uploaded by the mobile terminal 7 to the terminal server 8. Simultaneously, the mobile terminal continuously receives signals from the main base station 0 and from the neighboring base stations 1-6, and transmits its own MR data to the terminal server 8 via the main base station 0. The main base station 0 is responsible for time synchronization and distinguishing the unique identifier of the mobile terminal, and transmits the MR data uploaded by the mobile terminal. The terminal server 8 receives the MR data transmitted by the main base station 0, unpacks and processes it, including converting the data from a CSV file to an Excel file, performing data preprocessing on the Excel file, and verifying the unique identifier of the mobile terminal and time synchronization information. The processed data is then sent to the processing terminal 9. The processing terminal 9 first processes the received preprocessed data using the estimated distance ratio, then inputs the estimated distance ratio into the least squares algorithm to obtain a preliminary positioning result. The preliminary positioning result is then input into the MAP-LM algorithm for iteration. Through iteration, the final positioning information is obtained and transmitted back to the terminal server. The terminal server then transmits the information back to the primary base station according to the mobile terminal's unique identification code and time synchronization information. The primary base station is responsible for sending the final positioning result to the mobile terminal. Thus, the user obtains the precise coordinates of their location.

[0112] The main base station 0 is the base station furthest from the mobile terminal 7, and has the weakest received signal strength. It is responsible for receiving MR information, the mobile terminal's unique identification code, and time synchronization information from the mobile terminal 7, and sending these data to the terminal server 8. It can also receive location information from the terminal server 8 and send this information to the mobile terminal 7.

[0113] The neighboring base stations 1 to 6 are several base stations that are relatively close to the mobile terminal 7. It is not necessary to receive information from all six base stations, but at least information from two neighboring base stations must be received. Their main function is to transmit signals that are received by the mobile terminal 7 and sent to the main base station 0 as MR data.

[0114] The mobile terminal 7 is a user's mobile phone, iPad, Kindle, or other wearable device that can receive 5G base station signals and can function as a mobile device. It has the function of sending data, and can package information such as received signal strength into an MR data format file locally and transmit it to the main base station 0. It can also receive location information sent from the main base station.

[0115] The terminal server 8 includes a control module, a data storage module, a data processing module, and a forwarding module. The terminal server 8 receives MR data from the main base station 0, sends the data to the data storage module through the control module, unpacks and processes the MR data through the data processing module, and then sends the processed data to the processing terminal 9 through the forwarding module. It also receives the positioning results from the processing terminal 9 and sends them to the main base station 0 through the forwarding module according to the mobile terminal's unique identification code and time synchronization information.

[0116] The processing terminal 9 is a module that calculates the data that has been preprocessed by the terminal server to obtain the final location. It includes all the calculation algorithms and the operating environment, and sends the results to the terminal server 8 through its own forwarding module.

[0117] The following uses a map of a park and a distribution map of base stations as an example to further illustrate the invention in detail.

[0118] Example 1

[0119] See Figure 2 a. A map of a park and a map showing the distribution of base stations, with a scale of 1:50m.

[0120] See Figure 2 b. A map showing the distribution and test locations of base stations in a certain area, with a scale of 1:500m. Data collection simulated real user movement: data was collected while walking in a park using a mobile phone; and in another area, data was collected using a mobile phone mounted in a vehicle. Most base stations are approximately 60 meters high, but outdoor large-scale shadow fading and path loss are negligible due to height limitations, so a two-dimensional coordinate system was used for calculations. In the data collection area, there were no restrictions on vehicle speed or other environmental factors to simulate the most realistic environment.

[0121] In this embodiment, two different environments were selected for positioning. In both environments, the base station could act as either the primary base station or a neighboring base station. The mobile terminal, terminal server, and processing terminal are all the same device. The primary base station is the one furthest from the mobile terminal. It is responsible for receiving MR information, the mobile terminal's unique identifier, and time synchronization information from the mobile terminal and sending this data to the terminal server. The neighboring base stations are several base stations closer to the mobile terminal. They primarily transmit signals that are received by the mobile terminal and sent as MR data to the primary base station. The mobile terminal is a wearable or mobile device such as a smartphone, capable of receiving location information from the primary base station and sending its own generated MR data packets. The terminal server receives the MR data from the primary base station, sends it to the data storage module via the control module, unpacks and processes the MR data via the data processing module, and then sends the processed data to the processing terminal via the forwarding module. The processing terminal is a module that calculates the data preprocessed by the terminal server to obtain the final location. It includes all the calculation algorithms and the operating environment, and sends the results to the terminal server through its own forwarding module. This processing terminal needs to unpack the received and processed data to obtain a readable file. It uses the RSRP or RSRQ information of each neighboring base station and its respective master base station in the data to obtain the estimated distance ratio through the distance ratio estimation formula. It selects the most suitable estimated distance ratio through a selection function. It uses the TOA triangulation method to input the latitude and longitude or two-dimensional plane coordinate information of each base station and the estimated distance ratio into the estimation matrix and error matrix. It calculates the preliminary result through the least squares algorithm. It obtains the preliminary iteration value by substituting the preliminary result into the MAP-LM algorithm. According to the update factor and decision condition of the iteration, it is determined whether this result is the optimal point. If not, and the condition for continuing the iteration is met, this result is updated and the preliminary result is replaced to participate in the next iteration. This process continues until the update factor or the iteration judgment condition is met. Then, this result is output as the final result to obtain the final location coordinates.

[0122] See Figure 3 The specific process of 5G outdoor positioning is as follows:

[0123] a) Plan a route for the area to be located on a two-dimensional plane, and record the base station names, PCI and latitude and longitude coordinates that may be received along the route for use as engineering parameters.

[0124] b) Input the collected data into the computer for data preprocessing, remove unusable and interfering data, save the other data, and also save the latitude and longitude information of the test points. Then, error analysis can be performed with the final positioning results.

[0125] c) Analyzing the obtained preprocessed data, we can obtain the RSS information of the primary base station and the neighboring base stations for a single data point. This RSS information can be converted into the required format. Taking RSRP and RSRQ as examples, it can be expressed as RSSI using the following equation (a):

[0126]

[0127] RSSI is the received signal strength, N is the amplitude frequency, and the main frequency is several times the amplitude frequency. The value of N can be calculated from the RSSI provided by the main base station. N can be an integer, or each base station will provide the value of N, from which the values ​​of other parameters can be obtained. Then, the difference between the RSS and the main base station's RSS needs to be calculated using the following equation (b): DRSS is obtained by subtracting the neighboring base station's RSS from the main base station's RSS.

[0128] DRSS = RSRP N -RSRP1 (b).

[0129] Furthermore, another expression for DRSS can be obtained as shown in equation (c) below:

[0130]

[0131] in, It is a noise parameter, from which difference is obtained. RSRP .

[0132] d) The difference obtained in step c) RSRP Substituting this into equation (d) below expresses the distance ratio and its probability density function:

[0133]

[0134] in,

[0135] After a series of integrals and partial derivatives, the logarithmic form of the estimated distance ratio can be obtained as shown in equation (e) below:

[0136]

[0137] In the formula,

[0138] Thus, the maximum likelihood function of the distance ratio can be estimated through maximum likelihood estimation, and then the logarithmic form of the estimated distance ratio can be obtained by taking its partial derivative. Equation (e) above is a cubic equation in one variable, which theoretically can yield three different solutions, generally in the following three forms: one real root and two imaginary roots; one imaginary root and two real roots; and all three roots are real roots.

[0139] e) For the three roots obtained from equation (e) above, we need to select the best one. The criterion is to take the real root if there is one. If there are more than two real roots, we need to substitute them into the selection function for selection. The logic is to select the value with the distance ratio between [0 and 1] and not to be negative. If these estimated distance ratios are all numbers between [0 and 1] and cannot be distinguished, we need to use these results as conditions for subsequent processes.

[0140] f) Substituting the obtained estimated distance ratios into the two matrices in equations (f) to (g) below, we can obtain a result matrix (result vector) represented by equation (h):

[0141]

[0142]

[0143]

[0144] Substituting matrices A and B into equation (h), we obtain the result vector represented by equation (i) below:

[0145] ξ^=(A T YA) -1 A T YB (i);

[0146] Where Y is the identity matrix; A T A is the transpose of the matrix; -1 It is the inverse of the matrix.

[0147] The estimated position matrix can be obtained from equation (i): Two-step least squares estimation can be derived from the first two formulas into the following formula (i-1) regarding the unknown location information:

[0148]

[0149] in,

[0150] At this point, we can obtain preliminary positioning results.

[0151] g) The preliminary result obtained in step f) has poor accuracy. While it may be sufficient for some situations, it is not applicable to all outdoor conditions. Therefore, the preliminary result needs to be iteratively processed using the MP-LM algorithm to obtain the best positioning accuracy. The obtained result is expressed as follows:

[0152] ω = matrix[x, y, n] transport (j).

[0153] This includes the coordinates of the target point and the road loss index n, for difference RSRP In other words, we can obtain the following equation (k) to represent its maximum likelihood function based on the road loss index n:

[0154]

[0155] Equation (k) above is the MAP algorithm formula for path loss. It requires differentiating the maximum likelihood function to calculate its extreme points. These extreme points can be pseudo-extreme points, meaning there is no sign change around them. By taking the partial derivative of equation (k), we can obtain the following equation (k-1) regarding difference. RSRP probability distribution:

[0156]

[0157] Here, n belongs to the set of integers M, that is, relative to neighboring base stations. Since the number of base stations is fixed, this distribution is 0 for a single base station. This probability distribution is processed by a matrix to obtain a diagonal matrix, which is expressed by the following equation (I):

[0158]

[0159] Treating ω as the model function and ω^ as the result of the previous iteration, we can obtain the curve fitting formulas expressed by the following equations (m) to (n):

[0160]

[0161]

[0162] The first-order partial derivative of its model function is obtained from the following equation (o):

[0163]

[0164] Where θ is the step size function of the iteration.

[0165] For ease of calculation, let o T = matrix(ω(n)-ω^(n)) transport J is the Jacobian matrix. Substituting the results obtained from equations (1) and (n) into the curve fitting difference expressed in equation (p) below:

[0166] Ψ(n)-Ψ(n+1)=2*(ω-ω T )×W×J×hh T ×J T ×W×J×h(p).

[0167] Simultaneously, the perturbation function of the MAP-LM algorithm is introduced in equation (q):

[0168]

[0169] in, It is the damping coefficient for the disturbance.

[0170] Therefore, we can obtain the following formula (r) to represent the final calculation formula:

[0171]

[0172] in, These are equations concerning the perturbation and the Jacobian matrix, among other parametric matrices. The following is given... Components:

[0173] This is the PDF formula for the road damage index.

[0174] The Jacobian matrix is ​​expressed as follows (s):

[0175]

[0176] At the start of the iteration, a coefficient of 7 needs to be determined. If the update factor l calculated during the iteration is less than 7, a good result is considered to have been obtained, and this result can be overwritten on the original result before proceeding to the next iteration. However, there are two methods to judge the result: first, if the difference between two results during the iteration is less than 1m, the iteration can be considered to have ended; or, check if the maximum value of the iteration parameter matrix is ​​less than a certain value, which is preset, i.e., max(J T ×W×(ω-ω T The algorithm steps are expressed as follows (t):

[0177]

[0178] For l i If the value is greater than 1, it is considered an effective iteration, which can be considered superior to the previous estimation results. The updates of the remaining parameters are expressed as follows in equation (t-1):

[0179]

[0180] Among them, v i This is the step parameter, with an initial value of 0.

[0181] In other cases, it is necessary to... iExpanding this with other parameters for iteration, as shown in equation (t-2) below:

[0182]

[0183] See Figure 4 a. Using a certain region as the test environment, the MAP-LM algorithm yields an 80% cumulative distribution function of accuracy of 9.88 meters and a 50% cumulative distribution function of 1.47 meters.

[0184] See Figure 4 b. The traditional least squares algorithm yields a cumulative distribution function of 79.83 meters for 80% of the accuracy error and 59.54 meters for 50% of the accuracy error in a certain region.

[0185] See Figure 4 c. Using the MAP-LM algorithm in a test environment of a certain park, the 80% cumulative distribution function of the obtained accuracy is 13.68 meters; the 50% cumulative distribution function is 1.82 meters.

[0186] See Figure 4 d. The traditional least squares algorithm yields a cumulative distribution function of 30.14 meters for 80% accuracy error and 14.65 meters for 50% accuracy error in a certain park.

[0187] See Figure 4 It is evident that the RSS-based outdoor positioning algorithm of this invention can effectively reduce the workload of pre-collecting parameters and setting up the test environment, and can effectively and accurately locate mobile terminals, improving positioning accuracy and robustness, which is superior to existing positioning technologies.

[0188] This invention utilizes real user MR data for outdoor positioning and optimizes the positioning results using the MAP-LM algorithm, achieving positioning accuracy superior to the 3GPP standard. The MAP-LM algorithm does not rely on GPS or other positioning technologies, effectively protecting user privacy and eliminating the need for pre-measurement of the environment. By parsing the RSS (RSRP, RSRQ) information of the primary and neighboring base stations from the user-reported MR data, the distance from each source point to the primary and neighboring base stations is calculated. This distance is then processed using a distance ratio function to obtain the logarithmic form of the estimated distance ratio. Preliminary positioning coordinates are obtained using the least squares method. The proposed MAP-LM algorithm is then used to input these preliminary coordinates into an iterative equation. If certain conditions are met, the iteration exits, yielding the final positioning coordinates. Compared to existing technologies, this invention offers higher robustness and positioning accuracy, requires no extensive additional hardware deployment, is easy to promote, and has significant advantages and commercial prospects in complex outdoor positioning applications and NLOS scenarios.

[0189] The above is merely a further description of the present invention and is not intended to limit the present invention. Equivalent implementations that do not depart from the spirit and scope of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A 5G outdoor positioning method based on RSS, characterized in that, A collaborative positioning method employing a primary, neighboring base stations, mobile terminals, and a backend server utilizes deployed 5G base stations to receive MR data from user mobile terminals via the MAP-LM algorithm for precise user positioning. The method includes the following steps: Step 1: Data Preprocessing Compare the raw data with the engineering parameters. If any unknown main base station or neighboring base station is found, remove it. Remove data from different PCIs belonging to the same base station. Remove empty data and duplicate data. Data from fewer than three base stations is discarded to obtain preprocessed test data. Step 2: Calculation of the estimated distance ratio Based on the ratio of the distance to the base station to the distance to the mobile terminal, the estimated distance ratio between the two is calculated. Using the RSS feature parameter, its logarithmic form is obtained through the estimated distance ratio formula. Step 3: Calculation of preliminary positioning results The estimated distance ratios are input to a least squares algorithm to obtain an estimated matrix and an error matrix to obtain a preliminary positioning result, the estimated matrix is expressed by the following equation (9): （9）； The error matrix It can be expressed by the following equation (10): （10）； in, , The coordinates of the anchor base station; , The coordinates of the reference point; This represents the estimated distance ratio between the anchor base station and the source point. Step 4: Obtaining the final positioning coordinates The initial positioning results are input into the MAP-LM algorithm for iteration. The final positioning information is obtained through iteration and transmitted back to the terminal server. The terminal server then transmits the information back to the main base station according to the mobile terminal's unique identification code and time synchronization information. The main base station is then responsible for sending the final positioning results to the mobile terminal. Thus, the user obtains the precise coordinates of their location. The estimated distance ratio is calculated as follows: 1) Expressed using the following equation (6) Compared to distance The probability density function: （6）； In the formula, For about The probability density function of DRSRP; The difference between RSRP; 2) The maximum likelihood function of the distance ratio is estimated using maximum likelihood estimation. The maximum likelihood function is expressed by the following equation (7): ； In the formula, ; It is the shading attenuation coefficient of PLE; The difference between RSRP; These are noise parameters; It is the mean of the PLE distribution; 3) By taking the partial derivative of the above maximum likelihood function, we can obtain the logarithmic form of the estimated distance ratio.

2. The 5G outdoor positioning method based on RSS according to claim 1, characterized in that, The collaborative positioning method constructed by the main base station, neighboring base stations, mobile terminal and back-end server uses the existing deployed 5G base stations to accurately locate outdoor users. The main base station, neighboring base stations and server terminal are connected by optical fiber wired connection, and the user transmits signals to the main base station and neighboring base stations wirelessly. The main base station and neighboring base stations receive signals from the user's mobile terminal, including: the mobile terminal's unique identifier, time information, and RSRP, RSRQ, SINR, BeamId, PLMN, ScTadv, Arfcn, Pci, MR.Longitude, MR.Latitude, MR.HAOA, and MR.VAOA information, and then aggregate and send this information to the cloud server.

3. The 5G outdoor positioning method based on RSS according to claim 1, characterized in that, The preprocessing of the raw data involves removing outliers, specifically: removing null and duplicate values, removing base stations that do not exist in the parameters, removing data with fewer than 3 base stations, and removing data received from different PCIs belonging to the same base station. The raw data is the original MR data packet obtained from real user behavior tests, which is then parsed into a readable file.

4. The 5G outdoor positioning method based on RSS according to claim 1, characterized in that, The estimated distance ratio input least squares algorithm yields a one-dimensional column vector containing the coordinates of the target point and the road loss index n.

5. The 5G outdoor positioning method based on RSS according to claim 1, characterized in that, The preliminary positioning result is input into the MAP-LM algorithm to obtain the result matrix. The result matrix is ​​then iteratively processed, and the update factor and iteration constraints are input to obtain the final positioning result. The preliminary positioning result is a system of linear equations represented by the following equation (11-1): （11-1）； in, For estimating the matrix: The error matrix; for Transpose of a matrix; The resulting matrix is ​​expressed by the following equation (19): ； in, It is an equation concerning the perturbation and the Jacobian matrix, among other parameter matrices; The iteration constraints are expressed by the following equations (21-1) to (21-2): （21-1）； （21-2）； in This is the step parameter, initially set to 0; ;

Images