CN102395197A - TDOA cellular positioning method based on residual weighting - Google Patents

Abstract

The invention discloses a TDOA (Time Difference of Arrival) cellular positioning method based on residual weighting, which belongs to the field of radio positioning technology. The positioning method comprises the following steps: firstly, a cellular base station data is divided into several groups; and secondly, TDOA data corresponding to each suffix set Sk is individually estimated to obtain a medium estimated value, the medium estimated value is utilized to calculate a squared residual error or a universal set residual error with symbol to obtain a corresponding weighting value, and the weighting value is used for weighting the medium estimated value in a normalizing way to obtain the final position value of a mobile station. In the second step of the method, a universal set squared residual error function or a universal set residual error function with symbol is used for selecting root. The positioning method realizes the fusion of the simplicity and high efficiency of Chan and the good robustness and the ability of NLOS resistance of Rwgh, improves the estimate precision without increasing overmuch calculation load, and verifies that the algorithm of the invention can realize the function of fast and efficiently positioning the mobile station in the honeybomb communication system and has the relative good ability of NLOS resistance.

Description

A kind of TDOA honeycomb locating method based on the residual error weighting

Technical field

The present invention relates to a kind of method that in cellular communications networks, realizes the reliable location of mobile radio station, belong to the radio position finding radio directional bearing technical field.

Background technology

TDOA (Time Difference Of Arrival; Differential delay) history of radio position finding radio directional bearing technology is of long duration; The detection and the sonar signal that just are widely used in underwater sound signal as far back as the sixties are handled; The upsurge of research time difference estimation once appearred in the initial stage eighties, afterwards because the complexity constraints of hardware speed and algorithm gets into cold and still relatively period.Got into since the nineties, the time difference estimates to obtain again people's attention, and its range of application has expanded to general signal of communication and radar signal from underwater sound signal, and research means greatly enriches.In addition, military's demand is like radiation source direction finding, location (also being named passive direction finding, location respectively), all the time all in the sustainable development that is promoting the radio position finding radio directional bearing technology.Recently, domestic honeycomb communication (3G/LTE/4G) flourish expedited the emergence of in a large number based on the top service of travelling carriage high precision position information, and its vast market prospect has excited people to radio position finding radio directional bearing (especially cellular localization) Study on Technology enthusiasm.

1994; Chan and Ho have proposed a kind of method based on the TDOA hyperbolic fix (Chan method) (Y.T.Chan and K.C.Ho; " A simple and efficient estimator for hyperboliclocation, " IEEE Trans.Signal Processing, vol.42; Pp.1905-1915, August 1994.).According to the Chan method, if the TDOA noise of measuring is less and do not contain the NLOS deviation, the Chan method can utilize Measurement Variance information simply and efficiently to try to achieve location estimation; And the performance of this location estimation value is near CR circle; Belong to the narrower optimum of a kind of range of application location,, have big NLOS (Non-Line-Of-Sight owing to measure in the big or measured value of noise when localizing environment complicated (like city) at building dense; Non line of sight) deviation, the performance of Chan method can descend rapidly.

For suppressing the influence of NLOS deviation; Chen had provided a kind of method (Rwgh method) (P.-C.Chen based on squared residual weighting location in 1999; " A non-line-of-sight error mitigation algorithm inlocation estimation, " in Proc.IEEE Wireless Communications NetworkingConference, vol.1; Pp.316-320,1999.).In general, under least square (LS) meaning, generally can there be relatively large squared residual in the base station that NLOS takes place.The Rwgh method is measured as NLOS with squared residual; Through finding the solution different base stations combination non-linear LS location estimation down, calculate the interior squared residual of its combination, and with the inverse of this residual error this block position is separated and to be carried out weighting; To suppress the NLOS effect, finally separated.Under various measured values, the Rwgh method always can provide squared residual tolerance estimated value down, but adopts non-linear LS when finding the solution owing to various combination, and amount of calculation is relatively large.

As two kinds of radio position finding radio directional bearing methods that obtain more application at present, Chan method and Rwgh method all can be applicable to the location of cellular system, but the two is applied independently in present cellular positioning system separately; The limitation that all has himself; Therefore, be necessary to further investigate the model basis and the application scenario of two kinds of methods, merge Chan method and Rwgh method advantage separately; Revise the NLOS tolerance of squared residual under its LS; Give the accurate NLOS tolerance (proposing tape symbol residual error SR) more that makes new advances,, realize the raising of estimated accuracy simultaneously exceeding under the prerequisite that increases amount of calculation like the present invention.

Summary of the invention

The objective of the invention is to propose a kind of new NLOS tolerance, and under this tolerance, Chan is combined with Rwgh, provide one in cellular system, to realize the new method that high efficient and reliable is located based on the TDOA mode to existing technical problem.

A kind of TDOA honeycomb locating method based on the residual error weighting of the present invention comprises the steps:

Step 1, the cellular basestation data are divided into groups

To M the location and cellular basestation location aware of participation mobile radio station, wherein serving BS is BS ₁, with BS ₁As location reference point, and BS is set ₁Position coordinates is x ₁=(x ₁, y ₁)=(0,0); Other each cellular basestation is with BS _mExpression, wherein 2≤m≤M; The position coordinates of remembering said m base station is known vector x _m=(x _m, y _m);

Utilize serving BS BS ₁Receive the arrival time delay τ of mobile station signal ₁, all the other M-1 base station also receives the arrival time delay τ of this signal of mobile radio station respectively simultaneously _m, then with each τ _mAll with τ ₁Differ from, then can obtain all the other M-1 cellular basestation with respect to serving BS BS ₁M-1 TDOA data

That is:

${\hat{r}}_{31}=c({\mathrm{\&tau;}}_3-{\mathrm{\&tau;}}_1),...,{\hat{r}}_{\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">m</figure-callout>}1}=c({\mathrm{\&tau;}}_m-{\mathrm{\&tau;}}_1),...,{\hat{r}}_{\mathrm{<figure-callout\; id="1"\; label="M"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">M</figure-callout>}1}=c({\mathrm{\&tau;}}_M-{\mathrm{\&tau;}}_1),$ Wherein c is the electric wave signal propagation velocity, τ _mIt is the arrival time delay that receives of m base station from the signal of mobile radio station;

In the M-1 that is obtained TDOA data, to appoint and to get wherein that i TDOA data make up and 2≤i≤M-1 as one, the indexed set of remembering TDOA data in this combination is a S set _k, total N indexed set S that satisfies this condition _k, i.e. 1≤k≤N;

Step 2, grouped data is carried out the centre estimate

N the pairing TDOA data of indexed set are carried out single respectively estimate, obtain N middle estimated value, said to each indexed set S _kPairing TDOA data carry out single estimate to obtain in the middle of the operation of estimated value following:

Judge S _kGesture | S _k| be S _kMiddle element number, and coefficient matrix G _aColumn rank, wherein said coefficient matrix G _aBe the matrix of (M-1) * 3 dimension, first row walk to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe abscissa value opposite number, i.e. [x ₂,-x ₃... ,-x _M] ^T, secondary series walks to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe ordinate value opposite number, i.e. [y ₂,-y ₃... ,-y _M] ^T, the 3rd row walk to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe opposite number of TDOA data, promptly

Divide following situation to obtain middle estimated value then:

1. if | S _k|=2 and G _aThe row full rank, (x, equation group y) are found the solution this equation group, and with x, y uses variable about mobile station location coordinate x=then to utilize the non-linear least square method structure

Show out, the expression mobile station location is to BS ₁Euclidean distance; Finding the solution r ₁Quadratic expression the time, if r ₁Two positive root p are arranged ₁And p ₂, then try to achieve two groups of location estimation value p ₁=(x _P1, y _P1), p ₂=(x _P2, y _P2), then select the root rule therefrom to select one group to separate, as middle estimated value according to predetermined

2. if | S _k|＞2 and G _aThe row full rank then adopts two step weighted least require methods to find the solution, and when decorrelation is handled, from four groups of roots that obtained, selects the root rule to select one group to separate according to what be scheduled to, as middle estimated value

Said four groups of method for root that when decorrelation is handled, obtain are:

Decorrelation process result Za3 is 2 * 1 column vector; The matrix R1 of structure 2 * 2; Two elements that two elements of said matrix R1 first row are respectively among the Za3 ask positive square root to get the result of real part more respectively separately, and second classifies two elements negate square root result that gets real part more respectively separately among the Za3 as; Construct the R2 matrix then, it also is 2 * 2 matrixes, first row of the first behavioural matrix R1 of matrix R2, the opposite number of second row of the second behavioural matrix R1 of matrix R2; Two column vectors of matrix R1 are regarded as two groups of roots, and two column vectors of matrix R2 also are regarded as two groups of roots

3. if G _aBe listed as not full rank, when utilizing the localization method of linear array to find the solution, earlier weight matrix Ф be changed to unit matrix I, obtain initial direct least square solution, utilize this direct least square solution to obtain diagonal matrix B=diag (r ₂, r ₃..., r _M) approximation, r wherein ₂, r ₃..., r _MRepresent that respectively mobile station location is to base station BS ₂, BS ₃..., BS _MEuclidean distance; Combine TDOA covariance matrix Q can try to achieve new weight matrix Ф=BQB again according to B; Last according to the new weight matrix and the localization method of this linear array; Obtain the least square solution of weighting, if relate to that many groups are separated then select the root rule to select one group to separate, as middle estimated value according to predetermined

Above-mentioned N the middle estimated value that storage is obtained, said process can its corresponding S set _kBe labeled as

(1≤k≤N);

Estimated value

is calculated squared residual in the middle of step 3, the utilization, obtains corresponding weight value

To each indexed set S _kMiddle estimated value

Ask squared residual in the set, squared residual is defined as mobile station location coordinate x to be asked and indexed set S in the set _kFunction:

$R_{\mathrm{es}}(x,S_k)=\underset{l\&Element;S_k}{\mathrm{\&Sigma;}}\left\{{[{\hat{r}}_{l1}-\left(\right||x-x_l||-||x-x_1|\left|\right)]}^2\right\}\&ForAll;k.$

L ∈ S wherein _k, TDOA data value in the set

Position coordinates x with the base station _lBe known quantity, symbol || || for asking Euclidean distance; Squared residual function R in the pair set then _Es(x, S _k) carry out normalization and handle:

$R_{\mathrm{es}}^{\&prime;}(x,S)=\frac{R_{\mathrm{es}}(x,S_k)}{\left|S_k\right|}.$

| S _k| be described indexed set S _kGesture; With each Squared residual function R in value is gathered as the above-mentioned normalization of mobile station location coordinate x substitution ' _Es(x, S), i.e. altogether substitution N time, then by the N that is obtained normalization gather the squared residual function R ' _Es(x, inverse S) obtains the weight w of corresponding indexed set _k(1≤k≤N), that is,, weights are arranged to each k:

$w_k=\frac{1}{R_{\mathrm{es}}^{\&prime;}({\hat{x}}_k,S_k)}$

Step 4, middle estimated value

is done the normalization weighting, obtain the final position estimated value of this mobile radio station

Middle estimated value to the step 2 acquisition

Corresponding weight w with the step 3 setting _kWeighting, and do normalization and handle, the final position estimated value of mobile radio station obtained

$\hat{x}=\frac{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{\hat{x}}_k{w}_k}{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{w}_k}.$

So far, described TDOA honeycomb locating method disposes.

Further, in step 2, described being scheduled to selects the root rule to be: in the said three kinds of situation of step 2, relate at a plurality of limited p _nIn when electing, choose the p that makes that following J (x) functional value is minimum _nAs correct Solution

$J\left(x\right)=\underset{m=2}{\overset{M}{\mathrm{\&Sigma;}}}\left\{{[{\hat{r}}_{m1}-\left(\right||x-x_m||-||x-x_1|\left|\right)]}^2\right\}$

Wherein, || || for asking Euclidean distance; Function J (x) is called complete or collected works' squared residual function at x place, the quadratic sum of translational movement when its numerical value equals all auditory localization cues through translation and through location point x.

Further, as preferred scheme, in the step 2, described select the root rule can also for: in the said three kinds of situation of step 2, relate at a plurality of limited p _nIn when electing, can also adopt as follows the method for root that selects based on the tape symbol residual error:

At first, definition based on put on the x-y plane x=(x, the complete or collected works' tape symbol residual error function J ' that y) locates (x):

$J^{\&prime;}\left(x\right)=\underset{m=2}{\overset{M}{\mathrm{\&Sigma;}}}[{\hat{r}}_{m1}-\left(\right||x-x_m||-||x-x_1|\left|\right)]$

Wherein, || || for asking Euclidean distance;

The judgement set | J ' (p ₁) |, | J ' (p ₂) | ..., | J ' (p _n) | ... the minimum number of middle functional value, || for scalar is asked signed magnitude arithmetic(al);

1. if minimum value is unique, then get pairing of this minimum value as correct Solution

2. if minimum value is not unique; Promptly there is 1 above functional value to equate and the while minimum; And the actual value of these minimum values also equates, then from these minimum values, appoints and gets one group as correct Solution

3. if minimum value is not unique; Promptly there is 1 above functional value to equate and the while minimum; But the actual value of these minimum values is unequal; Then get that and make that its complete or collected works' tape symbol residual error function value is positive person, as correct Solution

Further; As preferred scheme, said step 3 can also adopt following method to utilize middle estimated value

to calculate complete or collected works' tape symbol residual error function and obtain corresponding weight value:

To each S set _kMiddle estimated value Ask its complete or collected works' tape symbol residual error function

That is the position coordinates x of serving BS, ₁, the TDOA data value

Position x with each base station _m(1≤m≤M-1) is a known quantity, with each Value as the said complete or collected works' tape symbol of mobile station location coordinate x substitution residual error function J ' (x) is asked N time altogether;

Then by the N that is obtained complete or collected works' tape symbol residual error function value

Calculate the weight w of corresponding indexed set _kAnd 1≤k≤N, method is following:

judges according to its actual value to complete or collected works' tape symbol residual error function value; Complete or collected works' tape symbol residual error function value

as is when being negative value; Its absolute value is big more; Then representative corresponding in the middle of estimated value

to contain the possibility of NLOS error big more; So the reliability of this centre estimated value

is more little, give the less weights of other weights relatively in view of the above this centre estimated value

.

As preferred scheme; Said give relatively for middle estimated value

the less weights of other weights can adopt the linear weighted function rule; Promptly to each k; Each

value as the said complete or collected works' tape symbol of mobile station location coordinate x substitution residual error function J ' (x) is asked N time altogether;

Then according to following formula by the N that is obtained complete or collected works' tape symbol residual error function value Calculate the weight w of corresponding indexed set _kAnd 1≤k≤N,

$w_k=\frac{J^{\&prime;}\left({\hat{x}}_k\right)-\min \left(J^{\&prime;}\left({\hat{x}}_k\right)\right)+1}{\max \left(J^{\&prime;}\left({\hat{x}}_k\right)\right)-\min \left(J^{\&prime;}\left({\hat{x}}_k\right)\right)+1}$

Where denote the N the minimum and maximum values .

The contrast prior art, beneficial effect of the present invention is:

Handle through the method; Can realize the fusion of advantages such as the good and anti-NLOS ability of robustness of the simple efficient and Rwgh method of Chan method; The present invention is based on the NLOS tolerance of squared residual under the LS that revises the Chan method; (select the root rule in new accurate more NLOS tolerance like squared residual associating tape symbol residual error; Or independent tape symbol residual error weighting etc.) under, can realize exceeding the raising that realizes estimated accuracy when increasing amount of calculation, emulation shows among the present invention that algorithm can be implemented in the cellular communication system mobile radio station positioning function and anti-relatively preferably NLOS ability fast and efficiently.

Description of drawings

Fig. 1 is that sketch map is implemented in 4 base stations of tape symbol residual error tolerance NLOS of the present invention;

Fig. 2 is that sketch map is implemented in 7 base stations of tape symbol residual error tolerance NLOS of the present invention;

Fig. 3 is LE, the UR performance comparison diagram of 7 base station embodiment of the present invention;

The 4th, the LE of 7 base station embodiment of the present invention, SR performance comparison diagram;

Fig. 5 is J (x) distribution maps of 7 base station embodiment of the present invention when the NLOS deviation is 0m;

Fig. 6 is J (x) distribution maps of 7 base station embodiment of the present invention when the NLOS deviation is 500m;

Fig. 7 is (x) distribution maps of 7 Js ' of base station embodiment when the NLOS deviation is 0m of the present invention;

Fig. 8 is (x) distribution maps of 7 Js ' of base station embodiment when the NLOS deviation is 500m of the present invention;

Fig. 9 implementing procedure figure that to be Chan method of the present invention directly combine with the Rwgh method;

Figure 10 is the implementing procedure figure of tape symbol residual error weighting positioning mode of the present invention.

Embodiment

To combine accompanying drawing and embodiment that the present invention is specified below; Technical problem and beneficial effect that technical scheme of the present invention solves have also been narrated simultaneously; It is pointed out that described embodiment only is intended to be convenient to understanding of the present invention, and it is not played any qualification effect.

Step 1, the cellular basestation data are divided into groups

To M the location and cellular basestation location aware of participation mobile radio station, wherein serving BS is BS ₁, with BS ₁As location reference point, be succinct form, consider two-dimensional case, BS ₁Position coordinates is x ₁=(x ₁, y ₁)=(0,0); Other each cellular basestation is with BS _m(2≤m≤M) expression, the position coordinates of remembering said m base station is known vector x _m=(x _m, y _m) (2≤m≤M).

Utilize serving BS BS ₁Receive the arrival time delay τ of mobile station signal ₁, all the other M-1 base station also receives the arrival time delay τ of mobile station signal respectively simultaneously _m, then with each τ _mAll with τ ₁Differ from, then can obtain all the other M-1 cellular basestation with respect to serving BS BS ₁M-1 TDOA data

That is: ${\hat{r}}_{31}=c({\mathrm{\&tau;}}_3-{\mathrm{\&tau;}}_1),...,{\hat{r}}_{\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">m</figure-callout>}1}=c({\mathrm{\&tau;}}_m-{\mathrm{\&tau;}}_1),...,{\hat{r}}_{\mathrm{<figure-callout\; id="1"\; label="M"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">M</figure-callout>}1}=c({\mathrm{\&tau;}}_M-{\mathrm{\&tau;}}_1),$ Wherein c is the electric wave signal propagation velocity, τ _mBe the arrival time delay from the signal that receives mobile radio station that receives of m base station, this arrival time delay can utilize the correlator of signal to obtain, and does not discuss at this.

In the M-1 that is obtained TDOA data, get wherein i TDOA data as a combination and 2≤i≤M-1, the indexed set of remembering TDOA data in this combination is a S set _k, such as the situation (M=4) of four base stations, if the 1st indexed set S ₁Only comprise base station BS ₂, BS ₃With respect to serving BS BS ₁The TDOA data

With Indexed set S then ₁=21, and 31}, the rest may be inferred also has indexed set S ₂=21,41}, S ₃=31,41}, S ₄={ 21,31,41} writes down all such indexed set S _k, the indexed set S that all satisfy condition _kAdd up to N, i.e. 1≤k≤N; As N=4 then being arranged, 1≤k≤4 in the above-mentioned four base station examples;

Step 2, grouped data is carried out the centre estimate

Utilize the Chan method that N the pairing TDOA data of indexed set are carried out single respectively and estimate, obtain N middle estimated value, distinguishing slightly place is summarized as follows among its estimation routine and the Chan; The detailed process of not launching to describe is please referring to document (Y.T.Chan and K.C.Ho; " A simple and efficient estimatorfor hyperbolic location, " IEEE Trans.Signal Processing, vol.42; Pp.1905-1915, August 1994.).To each indexed set S _kPairing TDOA data are carried out single and are estimated that the middle estimated value of acquisition comprises the steps:

Judge S _kGesture | S _k| (be S _kIn element number) and coefficient matrix G _a(be the G in the literary composition _a) column rank, wherein said coefficient matrix G _aBe the matrix of (M-1) * 3 dimension, first row walk to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe abscissa value opposite number, i.e. [x ₂,-x ₃.. ,-x _M] ^T, secondary series walks to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe ordinate value opposite number, i.e. [y ₂,-y ₃.. ,-y _M] ^T, the 3rd row walk to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe opposite number of TDOA data, promptly

1. if | S _k|=2 and G _aThe row full rank, (x, equation group y) are found the solution this equation group, and with x, y uses variable about mobile station location coordinate x=then to utilize the direct structure of non-linear least square technology

Show out, the expression mobile station location is to BS ₁Euclidean distance.Finding the solution r ₁Quadratic expression the time, if r ₁Two positive roots are arranged, can try to achieve two groups of location estimation p ₁=(x _P1, y _P1), p ₂=(x _P2, y _P2), then separate, as middle estimated value according to selecting the root rule to select one group

2. if | S _k|＞2 and G _aThe row full rank then adopts two step weighted least require methods to find the solution, and when decorrelation is handled, from four roots that obtained, selects, and concrete m program does;

R1=real ([sqrt (Za3)-sqrt (Za3)]); Four selections of %

R2＝[R1(1，:)；-R1(2，:)]；

RSR＝[R1 R2]′；

Variable Za3 in the program (result after decorrelation is handled) is the Z ' in the decorrelation processing in the Chan method literary composition _a(seeing formula 24 in this article); Za3 is 2 * 1 column vector; R1 is 2 * 2 matrixes; R1 first classifies two elements of matrix Za3 as and asks positive square root to get the column vector of real part more respectively respectively, second classify as two elements of Za3 respectively the negate square root get the vector of real part respectively, two column vectors among the R1 are two groups of roots among the Chan; The R2 matrix also is 2 * 2 matrixes, and first row of the first behavior R1 of R2 is gone two element intermodulation of second row position of R1 matrix as second of R2, and two column vectors of R2 also are regarded as two groups of roots; Thereby, further R1 and R2 being obtained 4 * 2 matrix RSR by contigency merging transposition again in the program, every row of RSR matrix corresponds to one group and separates, and is Z ' _aPositive and negative square root is got four kinds of combinations behind the real part; Select the root rule from these four groups are separated, to select one group to separate according to predetermined, as middle estimated value

3. if G _aBe listed as not full rank, when the localization method of linear array is found the solution in utilizing Chan method literary composition, earlier weight matrix Ф be changed to unit matrix I, obtain initial direct least square solution, utilize this direct least square solution to obtain diagonal matrix B=diag (r ₂, r ₃.., r _M) approximation, with r ₁Mark is consistent, wherein r ₂, r ₃.., r _M, represent that respectively mobile station location is to base station BS ₂, BS ₃..., BS _MEuclidean distance.Combine TDOA covariance matrix Q can try to achieve new weight matrix Ф=BQB again according to B; Last according to the new weight matrix and the localization method of this linear array; Obtain the least square solution of weighting; If group more than relating to is separated then is separated according to selecting the root rule to select one group, as middle estimated value

Wherein, the described root rule of selecting is: in above-mentioned three kinds of situation, relate to limited a plurality of four situation for example: { p ₁, p ₂, p ₃, p ₄In when electing, we choose and make minimum the separating as correct Solution of following J (x) functional value

Promptly

${\hat{x}}_k=\mathrm{<figure-callout\; id="1"\; label="arg\; min\; p"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">arg</figure-callout>}\underset{p}{\min }\underset{}{1,\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="HDA0000104816150000081.png,HDA0000104816150000091.png"\; state="\{\{state\}\}">p</figure-callout>}2,p3,p4}\{J\left(p_1\right),J\left(p_2\right),J\left(p_3\right),J\left(p_4\right)\}$

$J\left(x\right)=\underset{m=2}{\overset{M}{\mathrm{\&Sigma;}}}\left\{{[{\hat{r}}_{m1}-\left(\right||x-x_m||-||x-{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">x</figure-callout>}}_1|\left|\right)]}^2\right\}$

Wherein, || || for asking Euclidean distance, the selection of the root that relates among the present invention is at most four selections, but this rule is also abideed by in two selections, only needs J (p ₁), J (p ₂) make comparisons and get its little person and get final product.Function J (x) is called complete or collected works' squared residual function at x place; Its numerical value equals the quadratic sum of all auditory localization cues (being hyperbola during TDOA) through translation translational movement when passing through location point x; Can prove; In the localizing environment of no NLOS deviation, when the measurement noise of all base stations was consistent, this separates was that getting well under the maximum likelihood meaning separated.

Above-mentioned N the middle estimated value that storage is obtained, said process can its corresponding S set _kBe labeled as

(1≤k≤N);

Estimated value is calculated squared residual in the middle of step 3, the utilization, obtains corresponding weight value

To each indexed set S _kMiddle estimated value

Ask squared residual in the set, because TDOA data value in the set

Position coordinates x with the base station _l(l ∈ S _k) be known quantity, then squared residual is defined as unknown mobile station location coordinate x and S set in the set _kFunction:

$R_{\mathrm{es}}(x,S_k)=\underset{l\&Element;S_k}{\mathrm{\&Sigma;}}\left\{{[{\hat{r}}_{l1}-\left(\right||x-x_l||-||x-{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">x</figure-callout>}}_1|\left|\right)]}^2\right\}\&ForAll;k.$

|| || for asking Euclidean distance, it should be noted that R _Es(x, S _k) also inequality with the J (x) of preceding text, J (x) is merely the function of mobile station location coordinate.To function R _Es(x, S _k) carry out normalization and handle:

$R_{\mathrm{es}}^{\&prime;}(x,S)=\frac{R_{\mathrm{es}}(x,S_k)}{\left|S_k\right|}.$

| S _k| be S set mentioned above _kGesture (i.e. set element number).With each

Squared residual function R in value is gathered as the above-mentioned normalization of mobile station location coordinate x substitution ' _Es(x, S), i.e. altogether substitution N time, then by N squared residual function R that each substitution obtained ' _Es(x, inverse S) obtains the weight w of corresponding indexed set _k(1≤k≤N), that is, and to each k:

$w_k=\frac{1}{R_{\mathrm{es}}^{\&prime;}({\hat{x}}_k,S_k)}$

Step 4, middle estimated value

is done the normalization weighting, obtain the final position estimated value of this mobile radio station

Middle estimated value to the step 2 acquisition

Corresponding weight w with the step 3 setting _kWeighting, and do normalization and handle, the final position estimated value of mobile radio station obtained

$\hat{x}=\frac{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{\hat{x}}_k{w}_k}{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{w}_k}$

So far, dispose based on Chan method and the direct TDOA honeycomb locating method that combines of Rwgh method.Fig. 9 implementing procedure figure that to be Chan method of the present invention directly combine with the Rwgh method; As stated, the objective of the invention is that combination through Chan algorithm and Rwgh algorithm realizes.Know at present with regard to us; The location algorithm that Chan method and Rwgh method directly combine does not provide in current document; So above-mentioned steps has provided the algorithm steps that thought that Chan method and Rwgh method are contained and characteristics combine; And the present invention's some concrete vary in details when using Chan method and Rwgh method have been introduced in detail, following some other preferred improvement project that provides.Figure 10 is the implementing procedure figure of tape symbol residual error weighting positioning mode of the present invention.

It is pointed out that " separating " related among the present invention, be a estimated value, when the centre is estimated and finally estimated when not distinguishing, all to can be described as one " separating " perhaps " location solution ", below do not remake specified otherwise mobile station location.

Further, the root rule of selecting described in the step 2 can also adopt following method:

In the described three kinds of situation of step 2, relate to when in a plurality of, electing ({ p for example ₁, p ₂, p ₃, p ₄), adopt the method for root that selects based on the tape symbol residual error, at first, define one based on x-y planar point x=(x, complete or collected works' tape symbol residual error function y):

$J^{\&prime;}\left(x\right)=\underset{m=2}{\overset{M}{\mathrm{\&Sigma;}}}[{\hat{r}}_{m1}-\left(\right||x-x_m||-||x-{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">x</figure-callout>}}_1|\left|\right)]$

Wherein, || || for asking Euclidean distance, J ' (x) is called complete or collected works' tape symbol residual error function at x place; Select root rule branch following steps to carry out:

The judgement set | J ' (p ₁) |, | J ' (p ₂) | ..., | J ' (p _n) | ... in the minimum number of functional value: for example when four roots, judge set | J ' (p ₁) |, | J ' (p ₂) |, | J ' (p ₃) |, | J ' (p ₄) | the minimum number of middle functional value;

If minimum value is unique, then get the corresponding root of this minimum value as correct Solution

|| here for scalar is asked signed magnitude arithmetic(al);

If minimum value is not unique, promptly have 1 above functional value to equate and the while minimum, and the actual value of these minimum values is also equal, then therefrom appoints and gets one group as correct Solution

Such as | J ' (p ₁) |, | J ' (p ₂) | equate and the while minimum, and the actual value between these minimum values is also equal, i.e. J ' (p ₁)=J ' (p ₂), then from p ₁, p ₂In appoint and to get one group as correct Solution

If minimum value is not unique, promptly have 1 above functional value to equate and the while minimum, but the actual value of these minimum values is unequal, then gets that and makes that its complete or collected works' tape symbol residual error function value is positive person, as correct Solution

Such as | J ' (p ₁) |, | J ' (p ₂) | equate and the while minimum, but the actual value between these minimum values is unequal, i.e. J ' (p ₁J ' (the p of)=- ₂), then get its complete or collected works' tape symbol residual error function J ' and be worth and be positive person, even J ' (p ₁J ' (the p of)=- ₂P is then got in)＞0 ₁Be correct Solution

This kind selects the geometric meaning of root rule judgment to be; When square metric equality of displacement causes this tolerance to lose efficacy; We launch the directivity tolerance that translation adds; Discussion by embodiment part can know, when inside and outside translation absolute measure equated, this was separated and is getting well under the NLOS environment of doing translation laterally and separates.

Further; As preferred scheme; In the TDOA honeycomb locating method based on the weighting of tape symbol residual error of the present invention, step 3 can also adopt following method to utilize middle estimated value to calculate the tape symbol residual error and obtain corresponding weight value:

To each S set _kMiddle estimated value Ask its complete or collected works' tape symbol residual error

Because TDOA data value

Position x with the base station _m(1≤m≤M-1) is a known quantity, with each

Value as the above-mentioned tape symbol residual error function of mobile station location coordinate x substitution J ' (x) is asked N time altogether, then by the N that is obtained a tape symbol residual error

Calculate the weight w of corresponding indexed set _k(1≤k≤N), method is following:

To the tape symbol residual error according to its actual value (promptly no matter positive and negative all compare) according to actual value; General; When the tape symbol residual error of

is the negative value residual error; If its negative value is big more; Then representative corresponding in the middle of estimated value

contain the possibility of NLOS error bigger (in the accompanying drawing 1 four base station scenario can simple declaration); So the reliability of this centre estimated value is more little; Thereby, need to give this centre estimated value to give more little weights.Concrete weighting scheme has a variety of, like linear weighted function, square law weighting or the like, has provided simple the most a kind of linear weighted function rule based on siding-to-siding block length in the accompanying drawing 10, promptly to each k,

$w_k=\frac{J^{\&prime;}\left({\hat{x}}_k\right)-\min \left(J^{\&prime;}\left({\hat{x}}_k\right)\right)+1}{\max \left(J^{\&prime;}\left({\hat{x}}_k\right)\right)-\min \left(J^{\&prime;}\left({\hat{x}}_k\right)\right)+1}$

Where

denote the N the minimum and maximum values .

Below in conjunction with two embodiment execution mode of the present invention is further specified.

Embodiment 1

Typical regular hexagon honeycomb shown in accompanying drawing 1, a main serving BS BS ₁, 3 assistant base station BS that participate in the location ₂, BS ₃, BS ₄, obtain 3 groups of TDOA data values thus

Suppose only BS ₄NLOS deviation (for the purpose of discussing intuitively, this official holiday location survey amount noise is zero) takes place, its TDOA data value

Have one on the occasion of deviation, the 3rd hyperbola that then should meet at MS mark among the figure originally is (by BS ₁And BS ₄Decision) will be toward main serving BS BS ₁Translation, and this curve is can be along with the amount of deflection of NLOS bigger and continue to main serving BS BS ₁Draw close, the numerical value of setting the NLOS deviation be 0m to 900m, and base station radius R=1000m.

The positional value that the Chan method estimates; Provide with cross mark in the drawings; Because the mathematical characteristic (LS characteristic) of Chan method has determined its estimated value only can suppress the absolute value (claim absolute residuals UR, be the positive square root value of squared residual J (x)) of the error between the TDOA data value under true TDOA data value and the estimated value calculating; Thereby this estimated position only possibly appear near some hyp edge; And along with the pollution of NLOS deviation greatly, intersecting the curvilinear triangle area that forms between curve can be increasing, shown in black region among Fig. 1.Thereby the precision that is in the location estimation value of borderline Chan method will be worse and worse; At this moment; Will there be unreliability in the rule that only adopts UR or J (x) to measure NLOS; Introduce the tape symbol residual error and then further reflected the performance of separating under this NLOS environment, and the tape symbol residual error can convert UR or J (x) easily on the numerical value.If certain location solution to the tape symbol residual error SR of three base stations all for (being and the UR contrast that when not specifying the related base station of tape symbol residual error, the tape symbol residual error is also general term SR later on just; Complete or collected works' tape symbol residual error J ' of front definition (x) then only refers in particular to the UR summation of all being participated in the base station of location), explain that then this is separated every curve in three hyperbolas all is a side of the non-main serving BS of deflection, is referred to as the outside; Corresponding if symbol residual error is for negative; Then this separates the inboard that belongs to said curve, obviously, is in inboard separating under this analysis; Bigger NLOS possibility is arranged, thereby need suppress its influence with little weights.Thereby, when inside and outside translation absolute measure equates, when promptly UR or J (x) equate, under the NLOS environment, make the possibility that separating of translation has bigger true mobile station location laterally.General; When the relative NLOS deviation of TDOA measurement noise can be ignored; To the actual value of SR value, we have such superiority-inferiority ordering of separating: little positive number＞little negative＞big positive number＞big negative, and " little " here and the judgement of " greatly " can be measured 3 times of values of noise criteria difference with reference to TDOA; The tape symbol residual error SR value of measuring 3 times of values of noise criteria difference less than this TDOA is " little ", otherwise is " greatly ".

On the other hand; SR tolerance NLOS also has disadvantageous one side, and promptly when doing counteracting between the positive and negative residual error, SR can not well reflect the precision of separating; So; The present invention on the selective rule of second kind of selective root, in conjunction with squared residual and tape symbol residual error judge, and with the master that is judged as of squared residual.

But, to middle estimated value

Ask weight w _kThe time, point out based on the experiment of numerical value:

1, the selective rule of two roots of the present invention's proposition; Both selected unanimities under most situations; (concrete simulation example no longer provides) generally seldom carried out in a judgement that is promptly increased in the selective rule based on the root of tape symbol residual error, but in case squared residual tolerance because of the identical situation that lost efficacy under, in special base station geometric layout; Or on some special mobile station locations; When making single when organizing specific INSTANTANEOUS OBSERVATION again or to certain and estimating,, bring the very big lifting of positioning performance possibly if utilize the selective rule of the root after changing to revise; Because of only increasing tape symbol residual computations, the comparison that a judgement and small probability are carried out, computation complexity too much increases, and but makes algorithm itself obtain bigger accuracy guarantee, reaches the fault-tolerance to particular values information (like base station location value, measured value etc.).

2, can utilize the tape symbol residual error to construct weights separately; Like the tape symbol residual error weighting positioning mode that provides among the present invention; There are not complicated square operation and computing reciprocal in its weighting procedure, and the inverse weight basically identical of its performance and squared residual, thereby resource saved.

More than be the reasonability basis as NLOS tolerance with SR.

Further; Can be used as a kind of accurate more common metric of weighing the NLOS deviation to the SR of single base station design, and, since to the J ' of complete or collected works' design (x) or SR to the insensitivity of NLOS deviation and algorithm; Can the Chan method that find the solution at this place be changed and make any one other estimator efficiently; Thereby this tolerance can combine with squared residual, is applied as the design principle and the test stone of other location approach (like TOA etc.) widely.

Embodiment 2

Shown in accompanying drawing 2, embodiment 2 also is typical regular hexagon honeycomb: a main serving BS BS ₁, 7 assistant base station BS that participate in the location ₂, BS ₃, BS ₄, BS ₅, BS ₆, BS ₇Obtain 6 groups of TDOA data values thus

Still suppose only BS ₄The NLOS deviation takes place, i.e. its TDOA data value

Exist one on the occasion of deviation, the numerical value of setting the NLOS deviation be 0m to 900m, radius of society R=1000m.The estimated result of Chan method still provides with cross mark in the drawings, and can further confirm the characteristic of the Chan method of being proved among the embodiment 1: the estimated position only appears near the hyperbola.

Be the performance of simulation algorithm, can in the sub-district of main serving BS, get fixed mobile station location at random, the actual position that this instance is got mobile radio station is x=(476.91,829.92), selection of base stations radius R=1000m, and TDOA measurement standard difference is 30m.The positioning performance that has provided Chan method, Chan method and the direct combined techniques of Rwgh (claiming the Chan+Rwgh method) and tape symbol residual error weighting method (claiming the SRwgh method) among the figure compares; Its positioning performance is measured with position error (LE) and absolute residuals value (UR) and tape symbol residual values (SR); Being defined as here:

wherein; The unit of LE, UR and SR all is a rice (m);

is final estimated value; The functional value that

locates in final estimation for aforementioned functions J (x); The functional value that

(x) locates in final estimation

for aforementioned functions J ', || || for asking Euclidean distance.

The NLOS of transverse axis deviation change to from 50m 900m in a big way in, can see that from the LE curve of accompanying drawing 3 LE of Chan+Rwgh method and SRwgh method is approaching basically, and be superior to independent Chan method (positioning accuracy exceeds 200m approximately) more significantly.But in the scope of NLOS deviation less than 50m; Independent Chan method positioning accuracy is high slightly, and (Chan method precision is best when the NLOS deviation is 0m; Exceed 30m approximately); This with a lot of documents in the conclusion that provides consistent, that is: when the NLOS deviation is not obvious, have the algorithm (like the Chan method) that algorithm (like Chan+Rwgh method and SRwgh method) performance that the NLOS deviation suppresses is not so good as no NLOS deviation inhibition.Though it is pointed out that the fine tolerance locating accuracy of LE value ability, because of in the practical position system, the LE value is calculated need use actual position value x, and this actual position value x is unknown, and the application of all LE indexs only limits to numerical simulation and theory analysis.

In addition, can see in the UR curve from accompanying drawing 3, in three kinds of algorithms; Relational expression is satisfied in its UR value ordering basically: Chan method＜Chan+Rwgh method＜SRwgh method; Thereby the tolerance of UR value as positioning accuracy is described, when the NLOS deviation is big, is had reliability scarcely.Can see in the SR curve from accompanying drawing 4; In three kinds of algorithms; Relational expression is also satisfied in the ordering of its SR value basically: Chan method＜Chan+Rwgh method＜SRwgh method, and what SR was different with UR is, SR has reflected that the estimated value of Chan method compares with actual value and have inclined to one side characteristic in certain; Be SR negative value always, the estimated value that then means the Chan method in the actual location is always to BS ₁Direction produces deviation, and comparatively speaking, the SR value of Chan+Rwgh method is the minimum positive number among the three, thereby positioning accuracy is for the highest, has confirmed foregoing superiority-inferiority ordering of separating: little positive number＞little negative＞big positive number＞big negative.The precision that the SR value of SRwgh method is conciliate also meets described superiority-inferiority ordering of separating substantially, but described superiority-inferiority ordering of separating is always suitable, relative embodiment 2, and the applicability that this described superiority-inferiority ordering of separating of SR sorts by size than UR is good.

Further specify (x) distribution character of J (x) and J ' below in conjunction with accompanying drawing 5 to accompanying drawing 8, to help the reasonability of understanding data in the foregoing description and Algorithm design thinking.

Accompanying drawing 5 and accompanying drawing 6 are respectively NLOS deviations when being 0m and 500m, the contour distribution map of J (x), and successively decrease to the center in the outer that isocontour value is served as reasons among the figure, and its background is 6 and locatees hyperbolas, and the black blockage shown in the figure is the actual position value.From accompanying drawing 5 and accompanying drawing 6, can find out: when the NLOS deviation was 0m, more near actual value, its J (x) value was then more little on the x-y plane for anchor point, as can be seen from the figure, J this moment (x) with the actual position value as minimal point; When the NLOS deviation is 500m; J (x) still can measure locating accuracy preferably, when anchor point on the x-y plane during more near actual value, its J (x) value becomes little; But the existing deviation of the minimum convergence point of squared residual function J this moment (x), the actual position value is on the edge of central area.Generally, J (x) has measured locating accuracy preferably, but the reliability of this tolerance reduces with the increase of NLOS deviation.

Accompanying drawing 7 and accompanying drawing 8 are respectively NLOS deviations when being 0m and 500m, J ' contour distribution map (x), and successively decrease to the center in the outer that isocontour value is served as reasons among the figure, and its background also is 6 and locatees hyperbolas that the black blockage shown in the figure is the actual position value.From accompanying drawing 7 and accompanying drawing 8, can find out: when the NLOS deviation changes; J ' contour distribution map (x) has consistency (only being that isocontour numerical value is different); The position of not reflecting actual value among the figure; And in this embodiment, J ' contour distribution map (x) has the center symmetry characteristic based on initial point.Below we explain separately J ' (x) this center symmetry characteristic and to the insensitive characteristic of NLOS deviation.

Investigate J ' definition (x):

$J^{\&prime;}\left(x\right)=\underset{m=2}{\overset{M}{\mathrm{\&Sigma;}}}[{\hat{r}}_{m1}-\left(\right||x-x_m||-||x-{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="HDA0000104816150000031.png,HDA0000104816150000081.png"\; state="\{\{state\}\}">x</figure-callout>}}_1|\left|\right)]$

After given one group of TDOA data, we just can draw (x) curved surface of a specific J ', and in essence, J ' center symmetry (x) is from the geometric distributions of 7 base stations, and promptly each root contour is all from such curve: non-main serving BS (BS sues for peace ₂, BS ₃, BS ₄, BS ₅, BS ₆, BS ₇) distance with to main serving BS BS ₁The difference of distance, should and value be the track of the point of constant.Among this embodiment, the J ' at center origin place (x) is worth for minimum, explain initial point near BS ₁Point, also promptly inboard point, consistent with the analysis in the previous embodiment 1.Again in the J ' definition (x); Summation to all TDOA data values is arranged, wherein must comprise the influence of (claiming the NLOS total deviation) of NLOS deviation in all TDOA data, thereby; When the NLOS total deviation increases progressively; J ' only is to be in quantitatively at identical x to increase progressively as equivalent (x), because two summations of back then are that J ' (x) makes translation transformation at the z axle to being identical value to identical x from the graphics reflection; So the contour distribution map under its different N LOS deviation all duplicates; But we also possibly be used this specific character in the navigation system of measuring history based on TDOA, or utilize to the relation between non-complete or collected works' SR curved surface and come algorithm for design, specifically no longer set off a discussion here

The above; Be merely embodiment of the present invention, but protection scope of the present invention is not limited thereto, anyly is familiar with this technological people in the technical scope that the present invention disclosed; Can understand conversion and the replacement expected; All should be encompassed in of the present invention comprising within the scope, therefore, protection scope of the present invention should be as the criterion with the protection range of claims.

Claims (4)

1. the TDOA honeycomb locating method based on the residual error weighting is characterized in that, comprises following steps:

Step 1, the cellular basestation data are divided into groups

To M the location and cellular basestation location aware of participation mobile radio station, wherein serving BS is BS ₁, with BS ₁As location reference point, and BS is set ₁Position coordinates is x ₁=(x ₁, y ₁)=(0,0); Other each cellular basestation is with BS _mExpression, wherein 2≤m≤M; The position coordinates of remembering said m base station is known vector x _m=(x _m, y _m);

Utilize serving BS BS ₁Receive the arrival time delay τ of mobile station signal ₁, all the other M-1 base station also receives the arrival time delay τ of this signal of mobile radio station respectively simultaneously _m, then with each τ _mAll with τ ₁Differ from, then can obtain all the other M-1 cellular basestation with respect to serving BS BS ₁M-1 TDOA data

That is:

Wherein c is the electric wave signal propagation velocity, τ _mIt is the arrival time delay that receives of m base station from the signal of mobile radio station;

In the M-1 that is obtained TDOA data, to appoint and to get wherein that i TDOA data make up and 2≤i≤M-1 as one, the indexed set of remembering TDOA data in this combination is a S set _k, total N indexed set S that satisfies this condition _k, i.e. 1≤k≤N;

Step 2, grouped data is carried out the centre estimate

N the pairing TDOA data of indexed set are carried out single respectively estimate, obtain N middle estimated value, said to each indexed set S _kPairing TDOA data carry out single estimate to obtain in the middle of the operation of estimated value following:

Judge S _kGesture | S _k| be S _kMiddle element number, and coefficient matrix G _aColumn rank, wherein said coefficient matrix G _aBe the matrix of (M-1) * 3 dimension, first row walk to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe abscissa value opposite number, i.e. [x ₂,-x ₃... ,-x _M] ^T, secondary series walks to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe ordinate value opposite number, i.e. [y ₂,-y ₃... ,-y _M] ^T, the 3rd row walk to the capable base station BS that is followed successively by of M-1 from first ₂, BS ₃..., BS _MThe opposite number of TDOA data, promptly

Divide following situation to obtain middle estimated value then:

1. if | S _k|=2 and G _aThe row full rank, (x, equation group y) are found the solution this equation group, and with x, y uses variable about mobile station location coordinate x=then to utilize the non-linear least square method structure

Show out, the expression mobile station location is to BS ₁Euclidean distance; Finding the solution r ₁Quadratic expression the time, if r ₁Two positive root p are arranged ₁And p ₂, then try to achieve two groups of location estimation value p ₁=(x _P1, y _P1), p ₂=(x _P2, y _P2), then select the root rule therefrom to select one group to separate, as middle estimated value according to predetermined

2. if | S _k|＞2 and G _aThe row full rank then adopts two step weighted least require methods to find the solution, and when decorrelation is handled, from four groups of roots that obtained, selects the root rule to select one group to separate according to what be scheduled to, as middle estimated value

Said four groups of method for root that when decorrelation is handled, obtain are:

Decorrelation process result Za3 is 2 * 1 column vector; The matrix R1 of structure 2 * 2; Two elements that two elements of said matrix R1 first row are respectively among the Za3 ask positive square root to get the result of real part more respectively separately, and second classifies two elements negate square root result that gets real part more respectively separately among the Za3 as; Construct the R2 matrix then, it also is 2 * 2 matrixes, first row of the first behavioural matrix R1 of matrix R2, the opposite number of second row of the second behavioural matrix R1 of matrix R2; Two column vectors of matrix R1 are regarded as two groups of roots, and two column vectors of matrix R2 also are regarded as two groups of roots

3. if G _aBe listed as not full rank, when utilizing the localization method of linear array to find the solution, earlier weight matrix Ф be changed to unit matrix I, obtain initial direct least square solution, utilize this direct least square solution to obtain diagonal matrix B=diag (r ₂, r ₃..., r _M) approximation, r wherein ₂, r ₃..., r _MRepresent that respectively mobile station location is to base station BS ₂, BS ₃..., BS _MEuclidean distance; Combine TDOA covariance matrix Q can try to achieve new weight matrix Ф=BQB again according to B; Last according to the new weight matrix and the localization method of this linear array; Obtain the least square solution of weighting, if relate to that many groups are separated then select the root rule to select one group to separate, as middle estimated value according to predetermined

Above-mentioned N the middle estimated value that storage is obtained, said process can its corresponding S set _kBe labeled as 1≤k≤N wherein;

One of step 3, the following two kinds of methods of employing, estimated value in the middle of utilizing

Obtain corresponding weight value w _k:

(method 1): to each indexed set S _kMiddle estimated value

Ask squared residual in the set, squared residual is defined as mobile station location coordinate x to be asked and indexed set S in the set _kFunction:

L ∈ S wherein _k, TDOA data value in the set

Position coordinates x with the base station _lBe known quantity, then || || for asking Euclidean distance; Squared residual function R in the pair set then _Es(x, S _k) carry out normalization and handle:

| S _k| be described indexed set S _kGesture; With each

Squared residual function R in value is gathered as the above-mentioned normalization of mobile station location coordinate x substitution ' _Es(x, S), i.e. altogether substitution N time, then by the N that is obtained normalization gather the squared residual function R ' _Es(x, inverse S) obtains the weight w of corresponding indexed set _kAnd 1≤k≤N promptly, to each k, has weights:

(method 2): to each S set _kMiddle estimated value

Ask its complete or collected works' tape symbol residual error function

That is the position coordinates x of serving BS, ₁, the TDOA data value

Position x with each base station _mAnd 1≤m≤M-1 is a known quantity, with each

Value as the said complete or collected works' tape symbol of mobile station location coordinate x substitution residual error function J ' (x) is asked N time altogether;

Then by the N that is obtained complete or collected works' tape symbol residual error function value

Calculate the weight w of corresponding indexed set _kAnd 1≤k≤N, method is following:

To complete or collected works' tape symbol residual error function value

Judge according to its actual value, when

Complete or collected works' tape symbol residual error function value

During for negative value, its absolute value is more greatly then given this centre estimated value

Give the less weight w of other weights relatively _k

Step 4, to middle estimated value

Do the normalization weighting, obtain the middle estimated value of the final position estimated value of this mobile radio station the step 2 acquisition

Corresponding weight w with the step 3 setting _kWeighting, and do normalization and handle, the final position estimated value of mobile radio station obtained

2. according to the said a kind of TDOA honeycomb locating method of claim 1 based on the residual error weighting; It is characterized in that; In the method 2 of step 3; The said middle estimated value of giving is given the less weights employing linear weighted function rule of other weights relatively; Promptly to each k; Each

value as the said complete or collected works' tape symbol of mobile station location coordinate x substitution residual error function J ' (x) is asked N time altogether;

Then according to following formula by the N that is obtained complete or collected works' tape symbol residual error function value

Calculate the weight w of corresponding indexed set _kAnd 1≤k≤N,

Where

denote the N

the minimum and maximum values .

3. according to claim 1 or 2 said a kind of TDOA honeycomb locating methods, it is characterized in that in the step 2, the described root rule of selecting is based on the residual error weighting:

In the said three kinds of situation of step 2, relate at a plurality of limited p _nIn when electing, choose the p that makes that following J (x) functional value is minimum _nAs correct Solution

Wherein, || || for asking Euclidean distance; Function J (x) is called complete or collected works' squared residual function at x place, the quadratic sum of translational movement when its numerical value equals all auditory localization cues through translation and through location point x.

4. according to claim 1 or 2 said a kind of TDOA honeycomb locating methods, it is characterized in that in the step 2, the described root rule of selecting is based on the residual error weighting:

In the said three kinds of situation of step 2, relate at a plurality of limited p _nIn when electing, adopt as follows the method for root that selects based on the tape symbol residual error:

At first, definition based on put on the x-y plane x=(x, the complete or collected works' tape symbol residual error function J ' that y) locates (x):

Wherein, || || for asking Euclidean distance;

The judgement set | J ' (p ₁) |, | J ' (p ₂) | ..., | J ' (p _n) | ... the minimum number of middle functional value, || for scalar is asked signed magnitude arithmetic(al);

1. if minimum value is unique, then get pairing of this minimum value as correct Solution

2. if minimum value is not unique; Promptly there is 1 above functional value to equate and the while minimum; And the actual value of these minimum values also equates, then from these minimum values, appoints and gets one group as correct Solution

3. if minimum value is not unique; Promptly there is 1 above functional value to equate and the while minimum; But the actual value of these minimum values is unequal; Then get that and make that its complete or collected works' tape symbol residual error function value is positive person, as correct Solution

Images