A UWB single base station hybrid positioning method based on multipath identification

Abstract

The application discloses a UWB single base station mixed positioning method based on multi-path identification, and belongs to the field of UWB signal processing and positioning. The method judges whether it is a direct path according to channel information characteristics, and carries out target positioning on the direct path and multi-path environment respectively. The application comprises the following steps: constructing multiple characteristics based on the UWB channel information of the direct path and the multi-path, using the characteristics of CIR estimation and the weighted K nearest neighbor method to identify the multi-path in the environment; a positioning method for estimating signal parameters by using the MUSIC algorithm is proposed for the direct path, the algorithm can be used to jointly estimate the angle of arrival and the delay of the direct path for positioning, and a positioning method based on an adaptive genetic algorithm is proposed for the multi-path environment. The application can identify the direct path or the multi-path environment according to the CIR estimation characteristics in a complex unknown UWB positioning environment, and high-precision positioning can be realized by using different positioning methods.

Description

Technical Field

[0001] This invention relates to a UWB single-base station hybrid positioning method based on multipath identification, belonging to the field of UWB signal processing and positioning. Background Technology

[0002] In practical production and daily life applications, UWB boasts numerous advantages such as strong penetration, low power consumption, good anti-interference effect, large spatial capacity, and precise positioning, making it widely used in industries such as smart construction, elderly care and medical care, and logistics and transportation. However, in real-world, unknown observation environments, the changing motion state of the base stations themselves significantly increases the difficulty of multi-station positioning technology. Distance-based positioning accuracy is largely influenced by estimated parameters, and the accuracy of parameter estimation is often affected by the line-of-sight (LOS) path of signal propagation, limiting its application in real-world scenarios. Therefore, achieving high-precision positioning hinges on effective system path identification and designing more effective positioning algorithms based on the identification results.

[0003] In real-world environments, obstacles such as buildings and trees cause signals to reflect or refract between base stations and mobile stations, resulting in NLOS (Normally Inertial Occurrence) environments and significant deviations in measured propagation time and distance. In such cases, distance-based positioning methods will suffer from substantial positioning errors. Furthermore, current research on positioning in NLOS environments largely utilizes multiple base stations, with limited research on single-base station positioning. Therefore, researching single-base station UWB hybrid positioning in complex and unknown environments is highly significant. Summary of the Invention

[0004] This invention addresses the problem of accurate positioning in complex and unknown positioning environments by proposing a UWB single-base station hybrid positioning method based on multipath identification.

[0005] The technical solution adopted in this invention is as follows:

[0006] A UWB single-base station hybrid positioning method based on multipath identification includes the following process:

[0007] Step 1: Establish a system model, which includes a base station (BS) with a known location and a mobile terminal (MS) with an unknown location. Scatterers are randomly distributed between the BS and the MS, and the path between the BS and the MS includes LOS and NLOS paths. After the model is established, CIR feature values ​​are extracted.

[0008] Step 2: Real-time classification and identification are performed using the extracted features and the weighted K-nearest neighbor method to determine whether it is a LOS environment or an NLOS environment;

[0009] Step 3: If the identification result of Step 2 is a LOS environment, the MUSIC algorithm is used to solve for the target angle of arrival and the estimated value of arrival time, which is used for single base station angle-delay joint positioning to solve for the target location information; if the identification result of Step 2 is an NLOS environment, the adaptive genetic algorithm is used to obtain the single base station positioning result under the NLOS environment through multiple selections, crossovers and mutations.

[0010] Step 4: If the target moves through different environments (LOS and NLOS) within a set time period, then use Kalman filtering to combine and filter the LOS and NLOS positioning results at different times to supplement the accuracy of each result.

[0011] Furthermore, the specific implementation process of step 2 is as follows:

[0012] The features extracted in step 1 are used as test set data, and each feature is used as a test sample point. The test set data and the known training set data are weighted and normalized respectively. Then, the absolute distance between each test sample point and the training set data is calculated and arranged. K training points with the smallest distance are selected. The category with the most occurrence of the selected K training points is used as the predicted channel state of the test sample point. The value of K is adjusted with the size of the dataset. The larger the training set and test set, the larger the value of K should be. At the same time, the appropriate value of K is selected considering the amount of computation and real-time performance.

[0013] The specific process of weighted normalization is as follows: The feature parameters are normalized, and then each feature parameter is assigned a different weight based on its correlation with the classification category label. The Relief algorithm is used to calculate the weights: Ai weights. R(A1), R(A2), ..., R(A n ) are the test set features A1, A2, ... A n The correlation with the classification label, where n is the total number of feature parameters, 1≤i≤n.

[0014] Furthermore, in step 3, if the identification result in step 2 is an NLOS environment, then an adaptive genetic algorithm is used to obtain the single-base station positioning result under the NLOS environment through multiple selections, crossovers, and mutations. The specific implementation process is as follows:

[0015] (301) Based on the geometric relationship between the mobile terminal MS, the base station BS and the scatterer and the requirement to minimize the measurement error, an optimization model is obtained. Multiple randomly generated individuals are initialized and their individual fitness is calculated according to the optimization model. Here, an individual represents the possible location of the randomly generated hypothetical target MS'.

[0016] (302) The selection operator, crossover operator and mutation operator are applied to the population respectively, and the population is selected, crossed over and mutated to obtain the next generation population;

[0017] The adaptive crossover probability adjustment function is as follows:

[0018]

[0019] In the formula, p c max It is the crossover probability p c The maximum value of p c min It is the crossover probability p c The minimum value, T is the total number of iterations of the population, t is the number of iterations of the current population, and f is the fitness of an individual in the current population. max It is the individual's maximum fitness, f avg It is the average fitness of an individual, and the crossover probability p. c In the interval [p] c min， p c max The internal components are adaptively adjusted and changed;

[0020] When performing mutation operator calculations, the adaptive mutation probability adjustment function is:

[0021]

[0022] In the formula, p m max Let p be the mutation probability. m The maximum value of p m min For p m The minimum value, the mutation probability p m In the interval [p] m min p m max The internal components are adaptively adjusted and changed;

[0023] (303) Repeat step (302) a set number of times and output the individual with the highest fitness obtained during the evolution process as the optimal solution to obtain the final localization result in the NLOS environment.

[0024] The main advantages of this invention are as follows:

[0025] (1) Based on the actual environment and UWB equipment, this invention identifies the target CIR feature information. According to the LOS / NLOS identification results, different positioning algorithms are used to realize the target position estimation function, ensuring the convergence of the positioning method. This not only improves the positioning accuracy of the target node, but also reduces the positioning error of the system and improves the system performance.

[0026] (2) Interference from simultaneously operating radar systems is an increasingly serious problem in some radio frequency bands. UWB single-base station positioning benefits from coordinated transmitting and receiving components, which can better resist interference and obtain more accurate target information. Attached Figure Description

[0027] Figure 1 This is a system model diagram of the present invention;

[0028] Figure 2 This is a flowchart of the UWB single-base station hybrid positioning method based on multipath identification according to the present invention. Detailed Implementation

[0029] The UWB single-base station hybrid positioning method proposed in this invention fully utilizes the property that CIR has significantly different statistical distribution characteristics under LOS and NLOS conditions. It employs a weighted K-nearest neighbor algorithm, MUSIC positioning, and an adaptive genetic algorithm to obtain relatively accurate positioning results. To better illustrate the purpose and advantages of this invention, the following is a detailed explanation in conjunction with the appendix. Figure 1 , 2 The present invention will be described in detail.

[0030] 1. Establish a system model, including a base station (BS) with a known location and a mobile terminal (MS) with an unknown location, and randomly distribute some scatterers between the BS and the MS, including LOS paths and NLOS paths between the BS and the MS; after the model is established, extract CIR feature values.

[0031] Figure 1 The model of the positioning system is shown. Dashed lines represent the LOS path, and solid lines represent the NLOS path. Position-related parameters within the channels are also displayed. The figure uses two scatterers as an example. 1a d 1b Let d be the distance from BS to the A-th and B-th scatterers. 2a d 2b Let d be the distance from the A-th and B-th scatterers to the MS. los The direct radius distance from BS to MS. It is the emission angle, θ0, θ1...θ L It is the angle of arrival.

[0032] System model in NLOS environment based on Figure 1 Geometric relationships are extended to i = A, B, ..., L, d nlos，i＝ d 1i+ d 2i+ ε i , ε i Let the measurement error be the objective function. The position estimation problem of the MS can be solved by minimizing the measurement error and solving an optimization problem. We set the objective function as... Given the propagation distance d of the LOS path los The propagation distance d is less than the NLOS path. nlos，i That is, the optimization model is minL(p)std los ≤min{d nlos，i}, i = A, B, ..., L.

[0033] After establishing the model, the CIR under LOS and NLOS conditions is obtained to prepare for feature extraction. Statistical distribution features that show significant differences under LOS and NLOS conditions are extracted from the CIR, such as the total energy of the received signal, kurtosis, skewness, average delay distribution / root mean square (RMS) delay distribution, rise time, and kurtosis.

[0034] 2. Real-time classification and identification are performed using extracted features and the weighted K-nearest neighbor method to determine whether it is a LOS environment or an NLOS environment;

[0035] The Weighted K-Nearest Neighbors (WKNN) algorithm is an optimized algorithm that weights the features used in the K-Nearest Neighbors method. Using the WKNN algorithm requires four basic elements: determining the feature weights, the distance metric, the classification decision rule, the selection of the K value, and the determination of the dataset size.

[0036] The features extracted in step 1 are used as test set data, and each feature is used as a test sample point. The test set data and the known training set data are weighted and normalized respectively. Then, the absolute distance between each test sample point and the training set data is calculated and arranged. K training points with the smallest distance are selected. The category with the most occurrence of the selected K training points is used as the predicted channel state of the test sample point. The value of K is adjusted with the size of the dataset. The larger the training set and test set, the larger the value of K should be. At the same time, the appropriate value of K is selected considering the amount of computation and real-time performance.

[0037] The specific process of weighted normalization is as follows: The feature parameters are normalized, and then each feature parameter is assigned a different weight based on its correlation with the classification category label. The Relief algorithm is used to calculate the weights: A1 weights. R(A1), R(A2), ..., R(A n ) are the test set features A1, A2, ... A n The correlation with the classification label, where n is the total number of feature parameters, and the weights of A2…An are calculated in the same way as A1.

[0038] 3. Based on the identification results, different positioning methods are used in different positioning environments. If the identification result in step 2 is a LOS environment, the MUSIC algorithm is used to solve for the target angle of arrival and the estimated value of arrival time, which is used for single base station angle-time delay joint positioning to solve for the target location information. If the identification result in step 2 is a NLOS environment, the adaptive genetic algorithm is used to obtain the single base station positioning result in the NLOS environment through multiple selections, crossovers and mutations.

[0039] If the MUSIC algorithm is used in a LOS environment, the principle of the algorithm is to calculate the second moment of the output signal according to the output model formula, perform eigenvalue decomposition on the second moment, and use the eigenvectors to construct an orthogonal signal subspace and noise subspace to form a spatial spectrum function. The maximum value of the spectrum function is obtained in the spatial spectrum domain, and the angle and time corresponding to the spectral peak are the estimated values ​​of the arrival angle and arrival time of the direct path wave. The two estimated values ​​can be used for joint positioning, which can both estimate the positioning parameters and obtain high-precision positioning results.

[0040] If a single-base station localization algorithm based on an adaptive genetic algorithm is used in an NLOS environment, the specific implementation process is as follows:

[0041] (301) Based on the geometric relationship between the mobile terminal MS, the base station BS and the scatterer and the requirement to minimize the measurement error, the optimization model minL(p)std is obtained. los ≤min{d nlos，i}, i = 1, 2, ..., L, initialize multiple randomly generated individuals and calculate the fitness of each individual according to the optimization model; where, an individual represents a randomly generated possible location of the hypothetical target MS';

[0042] (302) The selection operator, crossover operator and mutation operator are applied to the population respectively, and the population is selected, crossed over and mutated to obtain the next generation population;

[0043] The adaptive crossover probability adjustment function is as follows:

[0044]

[0045] In the formula, p c max It is the crossover probability p c The maximum value of p c min It is the crossover probability p c The minimum value, T is the total number of iterations of the population, t is the number of iterations of the current population, and f is the fitness of an individual in the current population. max It is the individual's maximum fitness, f avg It is the average fitness of an individual, and the crossover probability p. c In the interval [p] c min， p c max The internal components are adaptively adjusted and changed;

[0046] When performing mutation operator calculations, the adaptive mutation probability adjustment function is:

[0047]

[0048] In the formula, p m max Let p be the mutation probability. m The maximum value of p m min For pm The minimum value, the mutation probability p m In the interval [p] m min p m max The internal components are adaptively adjusted and changed;

[0049] (303) Repeat step (302) a set number of times and output the individual with the highest fitness obtained during the evolution process as the optimal solution to obtain the final localization result in the NLOS environment;

[0050] 4. If the target moves through different environments (LOS and NLOS) within a set time period, Kalman filtering is used to combine and filter the LOS and NLOS positioning results at different times to supplement the accuracy of each result.

[0051] This invention proposes a hybrid positioning method for UWB single-base station based on LOS identification, which can improve positioning accuracy in complex and unknown positioning environments. The specific flowchart is as follows: Figure 2 As shown in the diagram, LOS (Location of Spectrum) identification technology is first used to determine the positioning environment. Specifically, the CIR extraction system in schemes 1 and 2 is used, and the weighted K-nearest neighbor method is employed for LOS identification based on feature values. Then, based on the identification results, different positioning methods are used in different positioning environments: the MUSIC algorithm is used in the LOS environment, and the single-base station positioning algorithm based on the adaptive genetic algorithm (as described in scheme 3) is used in the NLOS environment. Finally, if the target terminal MS passes through different LOS and NLOS environments at a given time, Kalman filtering is used to combine the LOS and NLOS positioning results to achieve complementary advantages and improve positioning accuracy.

Claims

1. A UWB single-base station hybrid positioning method based on multipath identification, characterized in that, Includes the following processes: Step 1: Establish a system model, which includes a base station (BS) with a known location and a mobile terminal (MS) with an unknown location. Scatterers are randomly distributed between the BS and the MS, and the path between the BS and the MS includes LOS and NLOS paths. After the model is established, CIR feature values ​​are extracted. Step 2: Real-time classification and identification are performed using the extracted features and the weighted K-nearest neighbor method to determine whether it is a LOS environment or an NLOS environment; Step 3: If the identification result of Step 2 is a LOS environment, the MUSIC algorithm is used to solve for the target angle of arrival and the estimated value of arrival time, which is used for single base station angle-delay joint positioning to solve for the target location information; if the identification result of Step 2 is an NLOS environment, the adaptive genetic algorithm is used to obtain the single base station positioning result under the NLOS environment through multiple selections, crossovers and mutations. Step 4: If the target moves through different environments (LOS and NLOS) within a set time, then use Kalman filtering to combine and filter the LOS and NLOS positioning results at different times to supplement the accuracy of each result. In step 3, if the identification result in step 2 is an NLOS environment, then an adaptive genetic algorithm is used to obtain the single-base station positioning result under the NLOS environment through multiple selections, crossovers, and mutations. The specific implementation process is as follows: (301) Based on the geometric relationship between the mobile terminal MS, the base station BS, and the scatterer, and the requirement to minimize measurement error, an optimization model is obtained. Multiple randomly generated individuals are initialized, and the fitness of each individual is calculated according to the optimization model. Here, an individual represents a randomly generated hypothetical target MS. ’ Possible location; (302) The selection operator, crossover operator and mutation operator are applied to the population respectively, and the population is selected, crossed over and mutated to obtain the next generation population; The adaptive crossover probability adjustment function is as follows: ； In the formula, Crossover probability The maximum value, Crossover probability The minimum value, It is the total number of iterations of the population. This is the current population iteration count. It is the individual fitness of the current population. It is the individual's maximum fitness. It is the average fitness of an individual, and the crossover probability. In the interval Internally, it is adaptively adjusted and changed; When performing mutation operator calculations, the adaptive mutation probability adjustment function is: ； In the formula, Mutation probability The maximum value, for Minimum value, mutation probability In the interval The internal components are adaptively adjusted and changed; (303) Repeat step (302) a set number of times and output the individual with the highest fitness obtained during the evolution process as the optimal solution to obtain the final localization result in the NLOS environment.

2. The UWB single-base station hybrid positioning method based on multipath identification according to claim 1, characterized in that, The specific implementation process of step 2 is as follows: The features extracted in step 1 are used as test set data, and each feature is used as a test sample point. The test set data and the known training set data are weighted and normalized respectively. Then, the absolute distance between each test sample point and the training set data is calculated and arranged. K training points with the smallest distance are selected. The category with the most occurrence of the selected K training points is used as the predicted channel state of the test sample point. The value of K is adjusted with the size of the dataset. The larger the training set and test set, the larger the value of K should be. At the same time, a suitable value of K is selected considering the amount of computation and real-time performance. The specific process of weighted normalization is as follows: The feature parameters are normalized, and then each feature parameter is assigned a different weight based on its correlation with the classification category label. The Relief algorithm is used to calculate the weights: Ai weights. , Features of the test set Relevance to category labels This represents the total number of feature parameters. .

Images