WO2015070613A1

WO2015070613A1 - Wireless positioning method and device

Info

Publication number: WO2015070613A1
Application number: PCT/CN2014/080718
Authority: WO
Inventors: 卢恒惠; 李超
Original assignee: 中兴通讯股份有限公司
Priority date: 2013-11-14
Filing date: 2014-06-25
Publication date: 2015-05-21
Also published as: CN104635206B; JP2017501389A; CN104635206A; JP6300922B2

Abstract

A wireless positioning method and device. The method comprises: respectively collecting the received signal strength of a plurality of known reference points on a point to be detected; according to the received signal strength, estimating a value space of the position of the point to be detected; dividing the estimated value space of the position of the point to be detected into a plurality of grids having the same size; and solving a corresponding Pearson product-moment correlation coefficient, and determining the minimum Pearson product-moment correlation coefficient ρ_L, a grid where a position corresponding to the ρ_L is located being taken as a position space of the point to be detected. By means of the technical solution, wireless positioning of higher accuracy can be achieved at lower calculation complexity.

Description

Method and device for wireless positioning

Technical field

The present invention relates to the field of wireless positioning technologies, and in particular, to a method and apparatus for wireless positioning.

Background technique

Due to the availability of low received signal strength (RSS) information and low cost, RSS-based wireless positioning has received widespread attention and widespread application. RSS-based targeting can often be divided into two broad categories: Location Fingerprint location and triangle positioning. The former needs to establish a database in advance, and the database is updated with the environment. The cost of building and maintaining is relatively high. Therefore, it is currently used in laboratories, buildings, etc., and has not been widely used. The latter calculates the distance between the point to be measured and the known reference point through the path loss model, and then performs triangle positioning based on the known reference point position and estimated distance. This triangle-based positioning scheme is simple and easy to implement, and has been widely used in commercial and scientific research fields. However, since the wireless signal is highly susceptible to environmental changes, the estimation of the unknown path loss model is very difficult, thus affecting the use of the triangle positioning scheme.

In order to solve the problem of triangle-based positioning based on RSS in the case of unknown path loss model, the related research proposes the following solutions: [1] Modeling the joint estimation problem of path loss index and position into a nonlinear optimization problem, and based on Levenberg The -Marquardt algorithm solves this problem; [2] defines distance compatibility based on the root axis, and dynamically estimates the path loss exponent by maximizing distance compatibility, and then uses it to perform triangle algorithm localization. [3] Modeling the joint estimation of path loss index and position to form a nonlinear optimization problem is solved by Gaussian-Seidel algorithm [4] by reducing the dimension of the Jacobian matrix. Simplify the complexity of the Lavenberg-Marquardt implementation in [1]; [5] solve the optimization problem of the path loss exponent and position of the original three variables by converting it into a univariate optimization problem by linearized path loss model processing, and The obtained value is substituted into the [4] scheme as the initial value to further improve the positioning accuracy.

Although the above five solutions can solve the positioning problem under the unknown path loss model, they also have their own shortcomings. [1] Computational complexity and results are limited by the choice of initial values; distance compatibility in [2] is prone to errors under noisy channels, leading to erroneous path loss indices and position estimates; The nonlinear Gaussian-Seidel algorithm in [3] does not guarantee the output of the global optimal solution in the nonconvex optimization problem. The result also depends on the selection of suitable initial values. [4] Although the application of [1] is simplified, it is complicated. The degree is still not low, and inherits the problem that the result of [1] is limited by the initial value; [5] linearizes the path loss model of nonlinearization, and introduces the error by loss of detail, and the complexity is not low.

Summary of the invention

The technical problem to be solved by the embodiments of the present invention is to provide a method and device for wireless positioning, which can achieve higher precision wireless positioning with lower computational complexity.

In order to solve the above technical problems, the following technical solutions are used:

A method for wireless positioning, comprising: collecting, respectively, received signal strengths of a plurality of known reference points on a point to be measured; estimating a value space of a position of the to-be-measured point according to the received signal strength; The point position value space is divided into a plurality of equal-sized grids; the Pearson product moment correlation coefficient corresponding to the plurality of equal-sized grids is solved, and the minimum Pearson product moment correlation coefficient ^ is determined, and the p _{L is} corresponding The grid where the location is located is the location space to be the 'J point.

Optionally, the step of dividing the estimated value space of the to-be-measured point into a plurality of equal-sized grids comprises: dividing the estimated value of the position of the to-be-measured point into equal sizes, and the area is m ² square grid, with the center position of each grid, ) ( / = 1, 2, · · · , m) as the possible values of all positions of the point to be measured, s is selected by the expected number of iterations and positioning The accuracy requirements are determined.

Optionally, the Pearson product moment correlation coefficient _Ρϊ (/= 1, 2 ., ^) corresponding to a plurality of grids of equal size includes:

The Α is calculated by the following formula:

A = - -

" _

I-1 /:1 N j=k _{t N} where ₇ = ^^, β = ^ ^, is the number of received signal strength measurements from the first reference point, ^ is the received signal strength received from the first reference point,

(x, ., .) is the position of the known reference point, i = l, 2 "N ,

N>3.

Optionally, the corresponding position ( _z , J is obtained by:

( _z , y _L ) = arg min p

A wireless positioning device includes a collection module, an estimation module, a division module and a processing module, wherein:

The collection module is configured to: respectively collect received signal strengths of a plurality of known reference points on the points to be measured;

The estimating module is configured to: estimate a value space of the location of the to-be-measured point according to the received signal strength;

The dividing module is configured to: divide the estimated value space of the to-be-measured point into a plurality of grids of equal size;

The processing module is configured to: solve a Pearson product moment correlation coefficient corresponding to a plurality of grids of equal size, and determine a minimum Pearson product moment correlation coefficient 3⁄4, and use the grid where the corresponding position is located as a point to be measured The location space.

Optionally, the dividing module is configured to divide the estimated value space of the to-be-measured point into a plurality of equal-sized grids according to the following manner: dividing the estimated value of the position of the to-be-measured point into a size Equal square grids of area m ² , with the center position of each grid, ) (/ = 1, 2, · · · , m) as the possible values of all positions of the point to be measured, the selection of s is expected The number of iterations and the positioning accuracy requirements are determined.

Optionally, the solving module is configured to solve a corresponding Pearson product moment correlation coefficient by using the following formula (ί = 1, 2, ··· )··

For the number of received signal strength measurements from the first reference point, ^ is the _th received signal strength received from the first reference point, _/· = ι,2,···Λ; β, = log ₁₀

For the position of the known reference point, i = l, 2 "N , Ν > 3

Optionally, the processing module is configured to obtain a position corresponding to p _L ( _z , y _L ) by using the following formula:

Arg mm p

The wireless positioning method and apparatus of the above technical solution can achieve higher precision wireless positioning with lower computational complexity.

BRIEF abstract

FIG. 1 is a flowchart of a method for wireless positioning according to an embodiment of the present invention;

2 is a schematic diagram of a device for wireless positioning according to an embodiment of the present invention;

Figure 3 is a schematic diagram of the results of an embodiment of the present invention.

Preferred embodiment of the invention

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that, in the case of no conflict, the features in the embodiments and the embodiments in the present application may be arbitrarily combined with each other.

In the embodiment of the present invention, RSS information from N (N ≥ 3) known reference points is measured at a point to be measured, and the position of the point to be measured is estimated by using the RSS information. H has no target node at (x, _y); the known position of the reference point is (Χ,., (Ζ· = 1,2 ··Λ ; the number of RSS measurements from the first reference point is, from the first reference point The received _/·s RSS is ^ (· = 1,2,···Λ). According to the path loss model of wireless propagation, ^ can be given by: r _j (dbm) = or(dbm) - 1 (Mog ₁₀ Uix-x, † + (yy _i f) + ω. Wherein, is the power received from the transmitting antenna lm; "For the path loss index, according to the existing research, the value range of the value is usually: 2 ≤ « ≤ 5; represents the shadow fading, usually in the study of wireless transmission models Suppose it obeys a Gaussian distribution, such as 0) Ν(0, σ ² , . , The above /^ expression can be simplified to:

Easy to find ^ and linear correlation. Further, from 2 ≤ « ≤ 5, ^ is linearly negatively correlated. Usually the degree of correlation between two variables can be described by the Pearson product moment correlation coefficient /?. The Pearson product moment correlation coefficient ranges from -1 ≤ /7 ≤ 1, and 7>0 indicates that two variables are positively correlated, that is, one variable increases as the other variable increases; ? < 0 indicates two The variables are negatively correlated, ie one variable decreases as the other variable increases; ? = 0 means that the two variables are not linearly related; ? = ±1 means that the two variables can be described by a good linear equation, ie The values of both variables fall on the same line.

Since the accurate position of the point to be measured is estimated to satisfy the expression ^

+ ω ₃ , that is, the accurate position estimation should make r _] satisfy the linear negative correlation, so the estimation problem of the position of the point to be measured can be transformed into the minimization problem of solving the correlation coefficient of the following Pearson product moments:

N _N is: ( J arg ^? ? , where = ^^, β = ^ ^.

Based on the above model, the embodiment of the present invention proposes a simple and easy algorithm to solve the above minimization problem. The algorithm uses the idea of hierarchical processing to iteratively solve the position of the point to be measured: First, the positioning space is discretized into several positioning areas, and the positioning space is reduced to a certain area by the Pearson product moment correlation coefficient calculation; then in the selected small area Repeat the previous step until the iteration termination condition is met.

The value of ^ and will change in each iteration above.

The embodiment of the invention models the wireless positioning problem based on RSS under the unknown path loss model into the most The optimization problem of the small Pearson product moment correlation coefficient is given, and a simple and easy iterative solution algorithm is given. Compared with the prior art, it can achieve accuracy directly with low complexity without estimating the path loss index. Positioning. Although the algorithm proposed in the embodiment of the present invention does not jointly estimate the path loss model parameters, after obtaining the position estimation, the path loss model parameters can be easily calculated directly from the linear regression. The embodiment of the invention proposes an RSS-based wireless positioning scheme in the case of an unknown path loss model. The scheme first uses RSS to model the wireless location problem under the unknown path loss model as the minimum Pearson product moment correlation coefficient, and then gives a simple and easy solution algorithm. The specific implementation process is shown in Figure 1, and includes the following steps:

101. On the point to be measured, respectively collect RSS from N (N ≥ 3) known reference points, and the number of samples of each reference point RSS is W = l, H), and the reference point is known. The position is {(x .)} , and the first RSS from the first reference point is ^ ( = 1, 2,...^). Among them, the location of the reference point can be obtained by GPS (Global Positioning System), manual estimation, map, CAD software (computer-aided design) and other methods; RSS sample can be installed by laptop, PDA (personal digital) with wireless network card Assistant), smartphones, etc. collected. Taking the collection of Wi-Fi RSS signals as an example, after installing the wireless network monitoring software on the laptop running the Window operating system, the software can be run to collect the RSS of the surrounding Wi-Fi access points according to the definition of the Pearson product moment correlation coefficient. And the path loss model models the location estimation problem of the point to be measured to minimize the Pearson product moment correlation coefficient problem, namely:

102. Estimate the value space of the location of the point to be measured. Using the known reference point position, the RSS received on the point to be measured, the transmission distance of the wireless signal, and other information (such as the building, area, etc. of the point to be measured) about the position of the point to be measured, roughly estimate the position of the point to be measured. Value space xe [Χ,, Χ,ΙγΕ [Υ„Υ ₂ ] ₀ such as: If the point to be measured is known to be in a certain floor, the coordinate space of the floor can be used as the value space of the position to be measured; There is no reference information related to the position of the point to be measured, but according to the centroid algorithm Estimate the position of the point to be measured as, and make a square with a side length of 2D centered on the point.

Take it as the value space of the position of the point to be measured. Among them, D is the corresponding wireless transmission distance of wireless communication technology, such as 802.11g indoor transmission distance is about 38m, outdoor transmission distance is about 140m.

103. Discretize the position value space of the point to be measured, and divide the estimated value of the position of the point to be measured into a plurality of grids of equal size.

The value space of the point to be measured estimated in step 102 is divided into equal-sized square grids of area s ² m ² , with the center position of each grid (3⁄4, ,) (/ = 1, 2) ,···, ) as possible values for all positions of the point to be measured. Among them, the selection of S is determined by the expected number of iterations and the positioning accuracy requirement. The smaller the number of iterations is expected, the higher the positioning accuracy requirement is, and the smaller the value of S is.

104. Solve the corresponding Pearson product moment correlation coefficient, and determine the position space corresponding to the minimum Pearson product moment correlation coefficient: Set all possible points to be tested ^, _3⁄4) (/ = ΐ, 2, ···, ∞ ) Substitute the following formula:

Ν _

One '=1 = i

Solve the corresponding Pearson product moment correlation coefficient A (7 = 1, 2, ···, m) and find the minimum Pearson product moment correlation coefficient p _L , the network where the p _L corresponds to ( y _L ) The grid is used as the location space of the point to be measured. Steps 103-104 are repeated until the iteration termination condition is met. The position corresponding to the minimum Pearson product moment correlation coefficient ^ obtained at this time is the position estimation value (j) of the required solution. The iterative termination condition can be determined according to the actual positioning requirements according to the positioning accuracy and the operation time. If the positioning accuracy is required to be lm, the iteration can be set to terminate when 5≤1111. So far, the solution of the positioning problem based on RSS in the case of the unknown path loss model is completed. 2 is a schematic diagram of a device for wireless positioning according to an embodiment of the present invention. As shown in FIG. 2, the device in this embodiment includes a collection module 201, an estimation module 202, a division module 203, and a processing module 204, where: The module 201 is configured to: respectively collect the received signal strengths of the plurality of known reference points on the points to be tested; The estimating module 202 is configured to: estimate a value space of the location of the to-be-measured point according to the received signal strength;

The dividing module 203 is configured to: divide the estimated value space of the point to be measured into a plurality of grids of equal size;

The processing module 204 is configured to: solve the corresponding Pearson product moment correlation coefficient, and determine a minimum Pearson product moment correlation coefficient _P1 , and use the grid where the _P1 corresponding position is the position space of the point to be J.

The dividing module is configured to: divide the estimated value space of the to-be-measured point into a square grid of equal size and area m ² , with a center position of each grid ( , ) (/) = 1,2,···,∞) As the possible values of all positions of the point to be measured, the selection of s is determined by the expected number of iterations and the positioning accuracy requirements.

Wherein, the solution module is set to; the corresponding Pearson product moment correlation coefficient A (7 = 1, 2, ···, w) can be solved by the following formula;

Pi

For the number of received signal strength measurements from the first reference point, ^ is the _/· received signal strength received from the first reference point, _/· = ι,2,···Λ;

(X, ., .) is the position of the known reference point, i = \, 2 '-N , N ≥ 3.

The 3⁄4 corresponding position ( _z , y _L ) is obtained by:

x _L , y _L ) = arg mm p

A specific wireless positioning embodiment is given below. This embodiment uses the experimental data used in the Wi-Fi positioning system named COMPASS of the University of Mannheim, Germany as an example. The experimental scene is an office floor with a width of about 15m and a length of about 36m. There are 14 known Wi-Fi access points in the office area. The lab database contains RSS from the 14 access points collected at 612 known location test points, where the number of RSS samples collected at each test point is 110. For the test The positioning performance of the invention in the case of different reference points (ie different N), the following is the true position

The access point A of ( 7.125, 6.269 ) is used as the point to be tested, and N of the 612 known position test points are randomly selected as reference points for detailed positioning process description.

According to the implementation flow described above, the total number of iterations is two, the s of the first iteration is 3, and the s of the second iteration is 1. The positioning of the test point can be as follows:

201. Randomly select N from 612 known location test points as reference points, and select corresponding RSS.

The location space of the test point is determined by the known office space, xe [0,36], _ye [0,15], ie the position space length Ζ = 36, width = 15.

202. Divide the space to be measured into 60 grids of 3×3 m ² , that is, ^=60^=3. Taking the center point of each grid as the possible value of the position of the point to be measured, there are:

3⁄4 =(mod(/,f / — l)xs + W2,j), =L〃( / " xs + W2,/ = l,2,~f xJ/s ²

203. Solve _A according to the Pearson product moment correlation coefficient formula and determine a position space corresponding to the minimum Pearson product moment correlation coefficient ^.

204. Repeat steps 202-203, where s = l, = 9, with the minimum 3⁄4 corresponding position ( , ) as the position estimate of the test point.

If the mean square error (RMSE) is used as the performance index, the positioning result of the embodiment of the present invention obtained by 10,000 Monte Carlo simulations can be given by FIG. 2, and the positioning performance of the scheme increases with the number of reference points. Big and improved. Taking the reference point number N=10 as an example, the positioning RMSE of the embodiment of the present invention is 3.066m, which is about 61% and 92.6 respectively compared with the background of the traditional centroid localization algorithm of 7.852m and the technical scheme [1] of 41.027m. A significant increase in % (Background Technical Solution [1] The reason for the large error is that there are multiple iterations in 10,000 Monte Carlo simulations that do not converge to the global optimal value). If the error median is used as the performance index, when the reference point number is N=10, the positioning error of the embodiment of the present invention is 2.0125 m, which is compared with 7.0028 m of the centroid algorithm and 3.1413 m of the background technical solution [1]. There was a significant increase of nearly 71.3% and 35.6% respectively. In addition, although the algorithm of the embodiment of the present invention is more complicated than the simplest centroid algorithm, the complexity is still low. When N=10, it takes only 144s to perform 10000 positioning with the Core i5, compared with the background technical solution [1]. The 988s saved nearly 85.4% of the time. In addition, the solution of the embodiment of the present invention can also increase the reference point and increase the iteration times. The number, the size of the mesh in the iteration are reduced, and the positioning accuracy is further improved.

One of ordinary skill in the art will appreciate that all or a portion of the above steps may be accomplished by a program instructing the associated hardware, such as a read-only memory, a magnetic disk, or an optical disk. Alternatively, all or part of the steps of the above embodiments may also be implemented using one or more integrated circuits. Correspondingly, each module/unit in the above embodiment may be implemented in the form of hardware or in the form of a software function module. The invention is not limited to any specific form of combination of hardware and software.

The above is only a preferred embodiment of the present invention, and of course, the present invention may be embodied in various other embodiments without departing from the spirit and scope of the invention. Corresponding changes and modifications are intended to be included within the scope of the appended claims.

Industrial applicability

The wireless positioning method and apparatus of the above technical solution can achieve higher precision wireless positioning with lower computational complexity. Therefore, the present invention has strong industrial applicability.

Claims

claims

1. A wireless positioning method, including: collecting the received signal strengths of multiple known reference points on the point to be measured; estimating the value space of the position of the point to be measured based on the received signal strength;

Divide the estimated position value space of the point to be measured into multiple equal-sized grids; solve for the Pearson product-moment correlation coefficient corresponding to multiple equal-sized grids, and determine the minimum Pearson product-moment correlation coefficient _Pl. , use the grid where the corresponding position _{of Pl} is located as the position space of the point to be 'J.

2. The method of wireless positioning according to claim 1, wherein: the step of dividing the value space of the estimated position of the point to be measured into a plurality of equal-sized grids includes: dividing the estimated position of the point to be measured. The value space is divided into square grids of equal size and area m ^2. Taking the center position of each grid, ) (/ = 1, 2, · · · , m) as all possible positions of the point to be measured The selection of value, s is determined by the expected number of iterations and positioning accuracy requirements.

3. The method of wireless positioning according to claim 2, wherein: the solution of the Pearson product moment correlation coefficient _A (/ = 1, 2, · · ·, m) corresponding to multiple grids of equal size Steps include:

Calculate the A by the following formula: n -

Among them, ₇ = ^^ , β = ^ ^, is the number of received signal strength measurements from the th reference point, ^ is the ·th received signal strength received from the th reference point,

,

N>3.

4. The wireless positioning method according to claim 3, wherein: the corresponding position ( _z , _{_z} ) is obtained by the following formula: x _L , y _L ) = arg mm p.

5. A wireless positioning device, including a collection module, an estimation module, a dividing module and a processing module, wherein:

The collection module is configured to: separately collect the received signal strengths of multiple known reference points on the point to be measured;

The estimation module is configured to: estimate the value space of the position of the point to be measured based on the received signal strength;

The division module is configured to: divide the estimated position value space of the point to be measured into multiple grids of equal size;

The processing module is set to: solve for the Pearson product moment correlation coefficient corresponding to multiple equal-sized grids, and determine the minimum Pearson product moment correlation coefficient ¾, and use the grid where the corresponding position is located as the point to be measured. location space.

6. The wireless positioning device according to claim 5, wherein: the dividing module is configured to divide the estimated position value space of the point to be measured into a plurality of equal-sized grids in the following manner: The value space of the position of the point to be measured is divided into square grids of equal size and area m ² , and the center position of each grid, ) (/ = 1, 2, · · · , m) is used as the point to be measured All possible position values of the point, the selection of s is determined by the expected number of iterations and positioning accuracy requirements.

7. The wireless positioning device as claimed in claim 6, wherein:

The solving module is configured to solve the corresponding Pearson product-moment correlation coefficient through the following formula

A (/ = 1,2,…, :

is the number of received signal strength measurements from the th reference point, ^ is the _/th received signal strength received from the th reference point, _/· = ι,2,···Λ; β, = log ₁₀

is the position of the known reference point, ί = \,2··N, N≥3.

8. The wireless positioning device according to claim 7, wherein: the processing module is configured to obtain the corresponding position of _Pl ( _iz , y _L

xy _L ) = ^ar §? ¹¹ "P