Ultra-wideband ranging gross error elimination method

By using Gaussian process regression model and least squares solution in indoor environments, the problem of ultra-wideband ranging error is solved, the indoor positioning accuracy and robustness are improved, and more accurate indoor positioning is achieved.

CN119064908APending Publication Date: 2024-12-03CHANGCHUN UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411321548.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-23
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In indoor environments, ultra-wideband ranging errors reduce positioning accuracy due to complex electromagnetic environments and building structures, making it difficult to meet the requirements of robust multilateral positioning solutions.

Method used

Using the ultra-wideband ranging error model (GP-Range) based on Gaussian process regression and the least squares position solution method of local optimal ranging correction, the ultra-wideband ranging error model is constructed through training data samples and real coordinates, and predict and measure The distance value is fitted and corrected to reduce the distance measurement error.

Benefits of technology

It improves indoor ultra-wideband positioning accuracy, reduces ranging errors, and enhances positioning robustness in complex electromagnetic environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119064908A_ABST
    Figure CN119064908A_ABST
Patent Text Reader

Abstract

The invention belongs to the field of indoor positioning, and particularly relates to an ultra-wideband ranging gross error elimination method. The invention relates to an indoor ultra-wideband ranging gross error elimination method. The method comprises sight distance ultra-wideband ranging gross error model training based on Gaussian process regression, two-dimensional positioning area ultra-wideband ranging gross error elimination and least square position calculation based on local optimal ranging correction. A Gaussian process is trained by using sample data formed by ultra-wideband ranging errors and real coordinates to obtain a gross error model, and ultra-wideband ranging value prediction is performed on all coordinate samples in a two-dimensional positioning area according to the model. And carrying out Gaussian correction on the new distance measurement value by using a local coordinate sample formed by the label and the corresponding distance measurement base station, and then calculating an optimal estimation position through a least-two standard. According to the invention, the precision of ultra wide band positioning in an indoor sight distance environment is further improved by reducing ultra wide band distance measurement errors caused by an indoor complex electromagnetic environment and a complex building environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of indoor positioning, and specifically to an ultra-wideband ranging gross error elimination method. Background Art

[0002] Ultra WideBand (UWB) technology can achieve positioning by transmitting nanosecond narrow pulses, and has the advantages of high clock resolution, strong penetration, low power consumption, and anti-multipath. UWB technology measures the distance by measuring the flight time of the signal between the tag and the base station, and can replace GNSS positioning technology for multilateral positioning in indoor environments. In an open line-of-sight environment, ultra-wideband positioning technology can generally achieve a positioning accuracy of 10-30 cm. However, the complex architectural environment and electromagnetic environment in the indoor environment seriously affect the accuracy of ultra-wideband ranging and positioning. Specifically, ultra-wideband distance measurement in indoor environments is manifested as a superposition of true distance, ranging errors and Gaussian noise. When performing multilateral positioning coordinate solution, the ranging value errors of all base stations involved in the solution need to be maintained within a small range to meet the robustness of the system.

[0003] In order to ensure the accuracy of ultra-wideband positioning under the conditions of limited indoor space and complex electromagnetic environment, the present invention proposes an ultra-wideband ranging gross error elimination method to reduce the error of indoor ultra-wideband ranging. The least squares position solution method based on local optimal ranging correction is used to solve the coordinates, thereby improving the positioning accuracy in indoor environments. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the deficiencies in the prior art, the present invention provides a method for eliminating gross errors in ultra-wideband ranging, which solves the problems raised in the above background technology.

[0006] (II) Technical solution

[0007] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0008] A method for eliminating gross errors in ultra-wideband ranging comprises the following steps:

[0009] S1, using the data sample E composed of ranging error and true coordinates d ~(x,y), train the ultra-wideband ranging error mathematical model GP-Range based on Gaussian process regression.

[0010] S2. Use the GP-Range model to predict the ultra-wideband ranging values ​​of all coordinate samples in the two-dimensional positioning area, and obtain the set Range~N(μ,Σ) of the Gaussian distribution obeyed by the ranging values ​​of all coordinate samples.

[0011] S3. Use the least squares position solution method based on local optimal ranging correction to solve the coordinates of moving or static tags in the two-dimensional positioning area.

[0012] Furthermore, the specific steps of S1 are as follows:

[0013] S11. Use the sample data collected by the automatic positioning data collection system to construct training and test data sets. A single data sample consists of UWB ranging error and true coordinates.

[0014] S12. Determine the explicit basis function of the Gaussian process according to the data characteristics between the ranging error and the coordinate samples, wherein the explicit basis function matrix H=1, and the explicit basis function coefficients β are determined by the data samples.

[0015] The relationship between the ranging error and the coordinate sample is decomposed into the following formula

[0016] r err =βH+f(d) (1)

[0017] where r err represents the ranging error; β is the coefficient of the explicit basis function; H is the explicit basis function matrix. Since there is no explicit relationship between the coordinates and the ranging error, let H = 1, and βH jointly describe the ranging gross error; f(d)~GP(0,k(d,d′)) is the latent variable introduced by the Gaussian process.

[0018] S13, according to the physical relationship between the distance measurement and the coordinate sample, the radial basis function is selected as the Gaussian process kernel function, where the Gaussian noise variance σ 2 , kernel function parameter σ l ,σ f Needs to be optimized.

[0019] The relationship between the ranging error and the coordinates is described by the GP-Range model based on Gaussian process regression:

[0020] P(r err |f,D)~N(r err |βd+f,σ 2 I) (2)

[0021] P(f|D)~N(f|0,K(D,D)) (3)

[0022] where K(D,D) is the covariance matrix between samples at different coordinates, and its elements are defined by the radial basis kernel function:

[0023]

[0024] where d i ,d jrepresents two different coordinate samples, σ f is the standard deviation of the ranging error, σ l is the kernel function scale, θ is the hyperparameter of the kernel function k, here refers to σ l ,σ f .

[0025] Furthermore, the specific steps of S2 are as follows:

[0026] S21. Use the GP-Range model to predict a coordinate sample of the two-dimensional positioning area. The ultra-wideband ranging error of the coordinate sample follows a Gaussian distribution. The formulas for calculating the mean and variance are as follows.

[0027] r new =K(d new ,D)(K(D,D)+σ 2 I) -1 (r err -Hβ)+h(d new )+d real (5)

[0028]

[0029] where d real is the true distance corresponding to the predicted coordinate sample.

[0030] S22. Use the GP-Range model to predict the ultra-wideband ranging values ​​of all coordinate samples in the two-dimensional positioning area, and obtain the Gaussian distribution set Range~N(μ,Σ) obeyed by the ultra-wideband ranging values ​​of all coordinate samples.

[0031] Furthermore, the specific steps of S3 are as follows:

[0032] S31, select local coordinate samples that participate in the fitting correction of the ranging value. Set a sliding window of length N in the positioning system, calculate the tag movement speed within the sliding window, and use it to estimate the position (x est ,y est ) and displacement r. Select the estimated position (x est ,y est ) The coordinate sample points within a radius of 10r participate in distance fitting.

[0033] S32, extracting corresponding coordinate sample data from the Gaussian distribution set Range~N(μ,Σ) obeyed by all coordinate samples, calculating the probability of a local sample point obtaining a new ranging value, and using the normalized probability to perform fitting correction on a new ultra-wideband ranging value.

[0034] The distance from each coordinate sample to the base station is called the true distance dreal The distance measurement value obtained by using UWB distance measurement for each coordinate sample is called the direct measurement value d mea ,The UWB ranging value predicted by the GP model for each coordinate sample is called the estimated ranging value The estimated ranging value is used to correct the direct ranging value of the sampling point on the GPR verification track, and the corrected result is called the fitting value d fit .

[0035] P (i) Indicates that the coordinate sample participating in the distance fitting takes a direct measurement value d mea Normalized probability.

[0036]

[0037] Use the normalized probability to fit and correct a new UWB ranging value.

[0038]

[0039] Where N is the number of coordinate sample points involved in distance fitting.

[0040] S34, performing least squares coordinate solution on the corrected set of ultra-wideband ranging values, and the coordinate X is shown as follows

[0041] X=(A T A) -1 A T B (10)

[0042]

[0043] Furthermore, the automatic positioning data acquisition system described in S11 includes: four ultra-wideband base stations and one ultra-wideband tag are deployed in the indoor space, and the ultra-wideband tag is placed on the vehicle-mounted motion platform. The vehicle-mounted motion platform consists of a fixed track, a motion chassis, a stepper motor and a main control MCU. Ultra-wideband ranging is completed by the ultra-wideband base station and the ultra-wideband tag, and the ranging information is sent by the ultra-wideband tag to the main control MCU through the serial port. The main control MCU encapsulates the ranging information, real coordinates and timestamp into a data frame and sends it to the host computer via Bluetooth. The host computer sends control instructions to the main control MCU via Bluetooth to realize continuous coordinate acquisition, adjust the coordinate sample spacing and tag movement speed.

[0044] (III) Beneficial effects

[0045] Compared with the prior art, the present invention provides a method for eliminating gross errors in ultra-wideband ranging, which has the following beneficial effects:

[0046] The present invention eliminates the gross errors of ranging in the real-time positioning system during ultra-wideband measurement by constructing an ultra-wideband ranging error model of indoor two-dimensional positioning space, and uses the least squares position solution method based on local optimal ranging correction to perform coordinate solution to improve positioning accuracy. The present invention further improves the accuracy of ultra-wideband positioning in indoor line-of-sight environment by reducing the ultra-wideband ranging error caused by complex indoor electromagnetic environment and complex building environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The following is a summary of the present invention;

[0048] Figure 2 A flow chart of a method for eliminating gross errors in ultra-wideband ranging according to the present invention;

[0049] Figure 3 Construct a flow chart for the GP-Range model of the present invention;

[0050] Figure 4 The present invention is a flow chart of the prediction of the Gaussian distribution set of ranging values ​​based on the GP-Range model;

[0051] Figure 5 The present invention is a flow chart of the least squares position solution method based on local optimal ranging correction;

[0052] Figure 6 A schematic diagram of coordinate sample sampling in an embodiment of the present invention;

[0053] Figure 7 The actual scene graph for sampling coordinate samples in the embodiment of the present invention;

[0054] Figure 8 It is the two-dimensional space ultra-wideband ranging error result described by the GP-Range model in an embodiment of the present invention;

[0055] Fig. 9 This is the result of the improved least squares positioning method in the embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] Example 1

[0058] like Figure 2-5As shown, an ultra-wideband ranging gross error elimination method proposed in one embodiment of the present invention includes the following steps:

[0059] S1, using the data sample E composed of ranging error and true coordinates d ~(x,y), train the ultra-wideband ranging error mathematical model GP-Range based on Gaussian process regression.

[0060] S2. Use the GP-Range model to predict the ultra-wideband ranging values ​​of all coordinate samples in the two-dimensional positioning area, and obtain the set Range~N(μ,Σ) of the Gaussian distribution obeyed by the ranging values ​​of all coordinate samples.

[0061] S3. Use the least squares position solution method based on local optimal ranging correction to solve the coordinates of moving or static tags in the two-dimensional positioning area.

[0062] like Figure 2 , Figure 3 and Figure 8 As shown, in some embodiments, the specific steps of S1 are as follows:

[0063] S11. Use the sample data collected by the automatic positioning data collection system to construct training and test data sets. A single data sample consists of UWB ranging error and true coordinates.

[0064] S12. Determine the explicit basis function of the Gaussian process according to the data characteristics between the ranging error and the coordinate samples, wherein the explicit basis function matrix H=1, and the explicit basis function coefficients β are determined by the data samples.

[0065] The relationship between the ranging error and the coordinate sample is decomposed into the following formula

[0066] r err =βH+f(d) (1)

[0067] where r err represents the ranging error; β is the coefficient of the explicit basis function; H is the explicit basis function matrix. Since there is no explicit relationship between the coordinates and the ranging error, let H = 1, and βH jointly describe the ranging gross error; f(d)~GP(0,k(d,d′)) is the latent variable introduced by the Gaussian process.

[0068] S13, according to the physical relationship between the distance measurement and the coordinate sample, the radial basis function is selected as the Gaussian process kernel function, where the Gaussian noise variance σ 2 , kernel function parameter σ l ,σ f Needs to be optimized.

[0069] The relationship between the ranging error and the coordinates is described by the GP-Range model based on Gaussian process regression:

[0070] P(r err |f,D)~N(r err |βd+f,σ 2 I) (2)

[0071] P(f|D)~N(f|0,K(D,D)) (3)

[0072] where K(D,D) is the covariance matrix between samples at different coordinates, and its elements are defined by the radial basis kernel function:

[0073]

[0074] where d i ,d j represents two different coordinate samples, σ f is the standard deviation of the ranging error, σ l is the kernel function scale, θ is the hyperparameter of the kernel function k, here refers to σ l ,σ f .

[0075] like Figure 4 and Fig. 9 As shown, in some embodiments, the specific steps of S2 are as follows:

[0076] S21. Use the GP-Range model to predict a coordinate sample of the two-dimensional positioning area. The ultra-wideband ranging error of the coordinate sample follows a Gaussian distribution. The formulas for calculating the mean and variance are as follows.

[0077] r new =K(d new ,D)(K(D,D)+σ 2 I) -1 (r err -Hβ)+h(d new )+d real (5)

[0078]

[0079] where d real is the true distance corresponding to the predicted coordinate sample.

[0080] S22, use the GP-Range model to predict all coordinate samples in the two-dimensional positioning area, and obtain the Gaussian distribution set Range~N(μ,Σ) obeyed by all coordinate samples. In this embodiment, all coordinate samples in the two-dimensional positioning space are two-dimensional coordinates in the two-dimensional positioning space with an x ​​and y spacing of 1 mm.

[0081] like Figure 5As shown, in some embodiments, the specific steps of S3 are as follows:

[0082] S31, select local coordinate samples that participate in the fitting correction of the ranging value. Set a sliding window of length N in the positioning system, calculate the tag movement speed within the sliding window, and use it to estimate the position (x est ,y est ) and displacement r. Select the estimated position (x est ,y est ) The coordinate sample points within a radius of 10r participate in distance fitting.

[0083] S32, extracting corresponding coordinate sample data from the Gaussian distribution set Range~N(μ,Σ) obeyed by all coordinate samples, calculating the probability of a local sample point obtaining a new ranging value, and using the normalized probability to perform fitting correction on a new ultra-wideband ranging value.

[0084] The distance from each coordinate sample to the base station is called the true distance d real The distance measurement value obtained by using UWB distance measurement for each coordinate sample is called the direct measurement value d mea ,The UWB ranging value predicted by the GP model for each coordinate sample is called the estimated ranging value The estimated ranging value is used to correct the direct ranging value of the sampling point on the GPR verification track, and the corrected result is called the fitting value d fit .

[0085] P (i) Indicates that the coordinate sample participating in the distance fitting takes a direct measurement value d mea Normalized probability.

[0086]

[0087] Use the normalized probability to fit and correct a new UWB ranging value.

[0088]

[0089] Where N is the number of coordinate sample points involved in distance fitting.

[0090] S34, performing least squares coordinate solution on the corrected set of ultra-wideband ranging values, and the coordinate X is shown as follows

[0091] X=(A T A) -1 A T B (10)

[0092]

[0093] A method for eliminating gross errors in ultra-wideband ranging, the automatic positioning data acquisition system described in S11 includes: four ultra-wideband base stations and an ultra-wideband tag are deployed in the indoor space, and the ultra-wideband tag is placed on a vehicle-mounted motion platform. The vehicle-mounted motion platform consists of a fixed track, a motion chassis, a stepper motor and a main control MCU. Ultra-wideband ranging is completed by the ultra-wideband base station and the ultra-wideband tag, and the ranging information is sent by the ultra-wideband tag to the main control MCU through the serial port. The main control MCU encapsulates the ranging information, real coordinates and timestamp into a data frame and sends it to the upper computer via Bluetooth. The upper computer sends control instructions to the main control MCU via Bluetooth to realize continuous coordinate acquisition, and adjust the coordinate sample spacing and tag movement speed.

[0094] In this embodiment, if Figure 6 and Figure 7 As shown, the actual scene includes: indoor scene, four ultra-wideband ranging base stations, an ultra-wideband ranging tag, a moving track, and an automatic positioning data collection system. The indoor scene is a rectangular area of ​​4 meters wide and 9.6 meters long, with three walls and one side without a wall, and there is no obstruction in the rectangular area; the four ultra-wideband ranging base stations are base stations A, B, C, and D. The ultra-wideband ranging tag is placed on the moving platform. During the collection process, the tag first moves at a constant speed along track 1, performs 150 distance measurements on the four ultra-wideband ranging base stations at each preset coordinate sample, and sends the data to the host computer, where the Euclidean distance between each preset coordinate is 5.6 cm; then the tag is combined with the starting point of track 2 and the collection process is repeated.

[0095] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for eliminating gross errors in ultra-wideband ranging, characterized in that: The following steps are included: S1, using the data sample E composed of ranging error and true coordinates d ~(x,y), train the ultra-wideband ranging error mathematical model GP-Range based on Gaussian process regression. S2. Use the GP-Range model to predict the ultra-wideband ranging values ​​of all coordinate samples in the two-dimensional positioning area, and obtain the set Range~N(μ,Σ) of the Gaussian distribution obeyed by the ranging values ​​of all coordinate samples. S3. Use the least squares position solution method based on local optimal ranging correction to solve the coordinates of moving or static tags in the two-dimensional positioning area.

2. The method for eliminating gross errors in ultra-wideband ranging according to claim 1, characterized in that: The specific steps of S1 are as follows: S11. Use the sample data collected by the automatic positioning data collection system to construct training and test data sets. A single data sample consists of UWB ranging error and true coordinates. S12. Determine the explicit basis function of the Gaussian process according to the data characteristics between the ranging error and the coordinate samples, wherein the explicit basis function matrix H=1, and the explicit basis function coefficients β are determined by the data samples. The relationship between the ranging error and the coordinate sample is decomposed into the following formula r err =βH+f(d) (1) where r err represents the ranging error; β is the coefficient of the explicit basis function; H is the explicit basis function matrix. Since there is no explicit relationship between the coordinates and the ranging error, let H = 1, and βH jointly describe the ranging gross error; f(d)~GP(0,k(d,d′)) is the latent variable introduced by the Gaussian process. S13, according to the physical relationship between the distance measurement and the coordinate sample, the radial basis function is selected as the Gaussian process kernel function, where the Gaussian noise variance σ 2 , kernel function parameter σ l ,σ f Needs to be optimized. The relationship between the ranging error and the coordinates is described by the GP-Range model based on Gaussian process regression: P(r err |f,D)~N(r err |βd+f,σ 2 I) (2) P(f|D)~N(f|0,K(D,D)) (3) where K(D,D) is the covariance matrix between samples at different coordinates, and its elements are defined by the radial basis kernel function: where d i ,d j represents two different coordinate samples, σ f is the standard deviation of the ranging error, σ l is the kernel function scale, θ is the hyperparameter of the kernel function k, here refers to σ l ,σ f .

3. The method for eliminating gross errors in ultra-wideband ranging according to claim 1, characterized in that: The specific steps of S2 are as follows: S21. Use the GP-Range model to predict a coordinate sample of the two-dimensional positioning area. The ultra-wideband ranging value of the coordinate sample obeys the Gaussian distribution. The formulas for calculating the mean and variance are as follows. r new =K(d new ,D)(K(D,D)+σ 2 I) -1 (r err -Hβ)+h(d new )+d real (5) where d real is the true distance corresponding to the predicted coordinate sample. S22. Use the GP-Range model to predict the ultra-wideband ranging values ​​of all coordinate samples in the two-dimensional positioning area, and obtain the Gaussian distribution set Range~N(μ,Σ) obeyed by the ultra-wideband ranging values ​​of all coordinate samples.

4. The method for eliminating gross errors in ultra-wideband ranging according to claim 1, characterized in that: The specific steps of S3 are as follows: S31, select local coordinate samples that participate in the fitting correction of the ranging value. Set a sliding window of length N in the positioning system, calculate the tag movement speed within the sliding window, and use it to estimate the position (x est ,y est ) and displacement r. Select the estimated position (x est ,y est ) The coordinate sample points within a radius of 10r participate in distance fitting. S32, extracting corresponding coordinate sample data from the Gaussian distribution set Range~N(μ,Σ) obeyed by all coordinate samples, calculating the probability of a local sample point obtaining a new ranging value, and fitting the ultra-wideband ranging value using the normalized probability. The distance from each coordinate sample to the base station is called the true distance d real The distance measurement value obtained by using UWB distance measurement for each coordinate sample is called the direct measurement value d mea ,The UWB ranging value predicted by the GP-Range model for each coordinate sample is called the estimated ranging value The estimated ranging value is used to correct the direct ranging value of the sampling point on the GPR verification track, and the corrected result is called the fitting value d fit . P (i) Indicates that the coordinate sample participating in the distance fitting takes a direct measurement value d mea Normalized probability. Use the normalized probability to fit the UWB ranging value. Where N is the number of coordinate sample points involved in distance fitting. S33, performing least squares coordinate solution on the corrected set of ultra-wideband ranging values, and the coordinate X is shown as follows X=(A T A) -1 A T B (10)