CN110488222A - The UWB localization method that SVM is combined with barycentric coodinates under the conditions of a kind of NLOS - Google Patents

The UWB localization method that SVM is combined with barycentric coodinates under the conditions of a kind of NLOS Download PDF

Info

Publication number
CN110488222A
CN110488222A CN201910764643.6A CN201910764643A CN110488222A CN 110488222 A CN110488222 A CN 110488222A CN 201910764643 A CN201910764643 A CN 201910764643A CN 110488222 A CN110488222 A CN 110488222A
Authority
CN
China
Prior art keywords
nlos
under
conditions
svm
uwb
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201910764643.6A
Other languages
Chinese (zh)
Other versions
CN110488222B (en
Inventor
魏璇
林志赟
韩志敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Dianzi University
Hangzhou Electronic Science and Technology University
Original Assignee
Hangzhou Electronic Science and Technology University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Electronic Science and Technology University filed Critical Hangzhou Electronic Science and Technology University
Priority to CN201910764643.6A priority Critical patent/CN110488222B/en
Publication of CN110488222A publication Critical patent/CN110488222A/en
Application granted granted Critical
Publication of CN110488222B publication Critical patent/CN110488222B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
    • G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W84/00—Network topologies
    • H04W84/18—Self-organising networks, e.g. ad-hoc networks or sensor networks

Abstract

The invention discloses the UWB localization methods that SVM under the conditions of a kind of NLOS is combined with barycentric coodinates, belong to technology of wireless sensing network field.This method can be divided into following steps, step 1: identifying that is collected is the measured value under NLOS or LOS propagation conditions using SVM algorithm, and alleviate measurement error caused by NLOS；Step 2: square distance matrix is constructed using the measured value under the conditions of LOS and the measured value after NLOS alleviation, and obtains the relative position of each UWB node using MDS algorithm；Step 3: the relative position obtained according to MDS algorithm calculates the generalized barycenter coordinate about unknown node；Step 4: the actual coordinate of unknown node is calculated in conjunction with the known coordinate of generalized barycenter coordinate and each anchor node.Method provided by the invention can efficiently identify the NLOS under complex environment, and effectively improve the positioning accuracy of UWB.

Description

The UWB localization method that SVM is combined with barycentric coodinates under the conditions of a kind of NLOS
Technical field
The present invention relates to wireless sensor network technology field more particularly to a kind of NLOS (Non-Line-Of-Sight, SVM (Support Vector Machine) is combined with barycentric coodinates under the conditions of NLOS) UWB (Ultra Wide Band, UWB) localization method.
Background technique
With the fast development of wireless sensor network technology, wireless sensor location technology has become current one and grinds Study carefully hot spot, and has to business, industry with every field such as military affairs and its important meaning and boundless application prospect.Closely UWB is quickly grown over year, its high, low in energy consumption, strong interference immunity with transmission rate, strong multi-path resolved ability, strong penetration capacity etc. Advantage, these advantages make it have brilliant performance in terms of wireless sensor positioning.However, in some complex environment (such as rooms Interior, subterranean tunnel etc.) in, since barrier is intensive, cause wireless sensor that can propagate at LOS (Line-Of-Sight, LOS) It propagates with NLOS and switches at random between two ways.In NLOS environment, due to the missing of signal direct path, when causing to propagate Between postpone, generate overgauge so that the range error of UWB increases, the precision that positions of this strong influence.
NLOS is correctly identified in the environment of mixing LOS and NLOS, and carrying out alleviating to it is current research heat Point.The algorithm of current research is mostly divided into two classes, and one is recognizers, will be under the conditions of LOS and NLOS by recognizer Measured value distinguish, then only positioned using the measured value under the conditions of LOS, this method is serious in NLOS condition When, often lead to not to position, the performance of this strong influence positioning；Another kind is to alleviate algorithm, by all data All input is alleviated is handled in algorithm, and this usual computation complexity of algorithm is high, and as the increase of NLOS measurement data is fixed Position degradation.
It is an object of the invention to solve the UWB orientation problem under NLOS propagation conditions, a kind of NLOS condition is proposed The UWB localization method that lower SVM is combined with barycentric coodinates.This method utilizes the measurement under SVM algorithm identification NLOS propagation conditions Value, and can effectively alleviate error caused by NLOS, while utilizing the measured value under the conditions of LOS and the NLOS measured value after alleviation To be positioned.This mode creatively combines above-mentioned identification and alleviates two kinds of algorithms, substantially increases positioning accuracy.
Summary of the invention
Present invention aims in view of the deficiencies of the prior art, propose under the conditions of a kind of NLOS SVM mutually to tie with barycentric coodinates The UWB localization method of conjunction, to solve the problems, such as UWB be located in influenced in complex environment by NLOS caused by position it is inaccurate.
The purpose of the present invention is achieved through the following technical solutions: SVM and barycentric coodinates phase under the conditions of a kind of NLOS In conjunction with UWB localization method, this method comprises the following steps:
Step 1: using being taken multiple measurements respectively in UWB radio indoors LOS environment and NLOS environment, and adopting respectively Collect in LOS environment and the channel impulse response CPR (Channel Impulse Response, CPR) under NLOS environment, and according to These CPR extract the feature samples for representing propagation conditions.
Step 2: building SVM classifier SVC (Support Vector Classifier, SVC), and by NLOS and LOS item It is trained in the feature samples and its corresponding label constitutive characteristic Input matrix SVM extracted under part, after being trained SVC.
Step 3: building SVM returns device SVR (Support Vector Regressor, SVR), and by NLOS and LOS item The feature samples extracted under part and corresponding output valve (UWB euclidean distance between node pair), which input in SVM, to be trained, after being trained SVR。
Step 4: acquiring new channel impulse response CPR, extract the feature samples for representing propagation conditions, and be inputted SVC after training sorts out measured value using the SVC after training and belongs to LOS propagation conditions or NLOS propagation conditions, and judges Whether the number of the measured value under LOS propagation conditions, which meets location requirement, (for example, located space is two-dimensional space, then needs full Sufficient measured value is no less than 3, and located space is three-dimensional space, then needs to meet measured value and be no less than 4), if meet demand, Square distance matrix only is constructed using the measured value under LOS propagation conditions, and by square distance Input matrix MDS (Multi- Dimension Scaling, MDS) algorithm acquisition relative position；If the population of measured values under LOS propagation conditions is not able to satisfy fixed Position demand, then alleviate the measured value under the conditions of NLOS using the SVR after training, i.e., by the feature samples under the conditions of NLOS It is input to and obtains new correspondence output valve in the SVR trained, then choose the measured value under the conditions of the measured value and LOS after alleviating Square distance matrix is constructed, and relative position is obtained by MDS algorithm.
Step 5: calculating that (coordinate is unknown, section to be positioned about unknown node using the relative position that MDS algorithm obtains Point) barycentric coodinates, and calculate unknown node using the coordinate of barycentric coodinates and known anchor node (node known to coordinate) Actual position coordinate.
Further, in the step 1, include: according to the feature that CPR is extracted
Received signal strength εr, calculation formula is as follows:
Wherein r (t) is the reception signal amplitude of t moment；
Euclidean distance between node pair d, calculation formula are as follows:
D=c (ti-t0)
Wherein c is the light velocity, tiFor receiving time, t0For the respond request time；
Receive signal maximum amplitude rmax, calculation formula is as follows:
Peak value κ, calculation formula are as follows:
Wherein μ|r|For receive signal amplitude mean value,For the variance for receiving signal amplitude, T is the sampling time；
Average excess delay time TMED, calculation formula is as follows:
Wherein ψ (t)=| r (t) |2/εr
Root mean square time delay spread time TRMS, calculation formula is as follows:
Further, the step 2 specifically includes the following contents:
Nonlinear Classifier is constructed using radial base (RBF) kernel function, carries out feature using radial base (RBF) kernel function Transformation, radial direction base (RBF) kernel function K (x ', the Y ') such as following formula:
K (x ', Y ')=exp (- γ | | x '-Y ' | |2)
Wherein, γ is nuclear parameter, and x ' is input feature vector sample, and Y ' is corresponding label value.
The objective function of SVM classifier SVC such as following formula:
s.t.Yk(wTxk+ b) -1 >=0, k=1,2 ..., n
Wherein, w, b are classifier parameters, are obtained by training；xkFor k-th of feature samples, YkFor k-th of feature samples Label, n is characterized sample total number.
Above-mentioned objective function is optimized using Lagrange duality function and radial base (RBF) kernel function, optimization knot Fruit is as follows:
Wherein, αkFor Lagrange multiplier.
Further, the step 3 specifically includes the following contents:
The recurrence device SVR structure and parameter of support vector machines, the radial direction base are constructed using radial base (RBF) kernel function (RBF) kernel function K (x ', y ') such as following formula:
Wherein, γ is nuclear parameter, and σ is the width of kernel function, and x ' is input feature vector sample, and y ' is corresponding output valve, i.e. UWB Euclidean distance between node pair.
The objective function such as following formula of SVM recurrence device SVR:
s.t.|(wTxk+b)-yk|≤ε, k=1,2 ..., n
Wherein, w, b are to return device parameter, are obtained by training；N is characterized sample total number, xkFor k-th of feature samples, yk For k-th of corresponding output valve, ε is error range.
Above-mentioned objective function is optimized using Lagrange duality function and radial base (RBF) kernel function, optimization knot Fruit is as follows:
Wherein, αkFor Lagrange multiplier.
Further, in the step 4, square distance Input matrix MDS algorithm is obtained into relative position, specific as follows:
Obtain square distance matrix:
Wherein dijFor the distance between i-th of UWB node and j-th of UWB node；
Calculate center matrix J:
N be UWB node total number, I be unit matrix, 1NFor N × 1 and it is worth the matrix for being 1；
Normalized cumulant square matrices:
X progress singular value decomposition is obtained into singular value Λ and singular value vector V；
X=VAVT
By singular value Λ by sequence arrangement from big to small, and extract with located space dimension (for example, located space is two Dimension space, then number is 2；Located space is three-dimensional space, then number is the 3) maximum singular value of same number and corresponding surprise Different value vector V1, the singular value of extraction is built into diagonal matrix Λ1, by the diagonal matrix Λ of building1With singular value vector V1It is mutually multiplied To the relative coordinate matrix Q of each UWB node:
Wherein Q representation are as follows:(x′i, y 'i) be unknown node opposite position It sets, (x '1, y '1), (x '2, y '2) ..., (x 'N-1, y 'N-1) it is the 1st, 2 ..., the relative position of N-1 anchor node.
Further, in the step 5, the weight about unknown node is calculated using the relative position that MDS algorithm obtains Heart coordinate specifically: the relative position of unknown node is regarded as to the Generalized Barycentric of other nodes, and broad sense is solved by following formula Barycentric coodinates:
a1*x′1+a2*x′2+…+aN-1*x′N-1=x 'i
a1*y′1+a2*y′2+…+aN-1*y′N-1=y 'i
a1+a2+…+aN-1=1
Wherein, (a1, a2..., aN-1) it is generalized barycenter coordinate.
Further, in the step 5, actual position coordinate is calculate by the following formula to obtain:
xi=a1*x1+a2*x2+…+aN-1*xN-1
yi=a1*y1+a2*y2+…+aN-1*yN-1
Wherein, (xi, yi) be required unknown node actual position coordinate, (x1, y1), (x2, y2) ..., (xN-1, yN-1) be The true coordinate of 1st, 2 ..., N-1 anchor nodes.
Beneficial effects of the present invention: the present invention combines SVM with barycentric coodinates method, effectively will using SVM classifier Measured value under the conditions of measured value and LOS under NLOS propagation conditions is distinguished, and returns device under NLOS propagation conditions using SVM Measured value alleviated, this greatly reduced NLOS bring range error.Meanwhile utilizing the Generalized Barycentric based on MDS Coordinate method positions unknown node, solves presence of traditional geometry location algorithm because measuring noise, does not often have Intersection point or multiple intersection points lead to the problem of positioning failure.The combination of two kinds of algorithm creativeness, not only ensure that the performance of positioning, The positioning accuracy of UWB under NLOS propagation conditions is also greatly improved simultaneously.With the development of portable device, the application of UWB Range is more and more extensive, such as the tracking of personnel, unmanned plane, the positioning of unmanned vehicle etc..Of the invention has wide range of applications, to economy Development have huge effect.
Detailed description of the invention
The UWB localization method that SVM is combined with barycentric coodinates under the conditions of Fig. 1 is a kind of NLOS provided in an embodiment of the present invention Flow chart；
Fig. 2 is SVM training process figure；
Fig. 3 is the signal graph under indirect wave and direct wave propagation conditions；
Fig. 4 is MDS algorithm flow chart；
Fig. 5 is that unknown node indicates to scheme about the barycentric coodinates of anchor node.
Specific embodiment
The specific embodiment of the invention is described in further detail below in conjunction with attached drawing.
As shown in Figure 1, the UWB localization method that SVM is combined with barycentric coodinates under the conditions of a kind of NLOS provided by the invention, The following steps are included:
(1) using meet the UWB radio of FCC indoors LOS environment with carried out respectively in NLOS environment it is multiple extensive Measurement, acquiring NLOS condition respectively, (as wherein one group shown in Fig. 3 non-through with the channel impulse response CPR under the conditions of LOS Signal graph under wave and direct wave propagation conditions), and extracted according to these channel impulse responses CPR and can characterize propagation conditions Feature samples, when the feature mainly extracted includes: signal receiving strength, signal receives maximum amplitude, peak value, averagely excess postpones Between, root mean square time delay spread time etc., it is specific as follows:
Received signal strength εr, calculation formula is as follows:
Wherein r (t) is the reception signal amplitude of t moment；
Euclidean distance between node pair d, calculation formula are as follows:
D=c (ti-t0)
Wherein c is the light velocity, tiFor receiving time, t0For the respond request time；
Receive signal maximum amplitude rmax, calculation formula is as follows:
Peak value κ, calculation formula are as follows:
Wherein μ|r|For receive signal amplitude mean value,For the variance for receiving signal amplitude, T is the sampling time；
Average excess delay time TMED, calculation formula is as follows:
Wherein ψ (t)=| r (t) |2/εr
Root mean square time delay spread time TRMS, calculation formula is as follows:
(2) SVM classifier SVC (Support Vector Classifier, SVC) is constructed, and by NLOS and LOS condition It is trained in the feature samples of lower extraction and its corresponding label constitutive characteristic Input matrix SVM, after being trained SVC, as shown in Fig. 2, specific as follows:
Nonlinear Classifier is constructed using radial base (RBF) kernel function, carries out feature using radial base (RBF) kernel function Transformation, radial direction base (RBF) kernel function K (x ', the Y ') such as following formula:
K (x ', Y ')=exp (- γ | | x '-Y ' | |2)
Wherein, γ is nuclear parameter, and x ' is input feature vector sample, and Y ' is corresponding label value.
The objective function of SVM classifier SVC such as following formula:
s.t.Yk(wTxk+ b) -1 >=0, k=1,2 ..., n
Wherein, w, b are classifier parameters, are obtained by training；xkFor k-th of feature samples, YkFor k-th of feature samples Label, n is characterized sample total number.
Above-mentioned objective function is optimized using Lagrange duality function and radial base (RBF) kernel function, optimization knot Fruit is as follows:
Wherein, αkFor Lagrange multiplier.
(3) building SVM returns device SVR (Support Vector Regressor, SVR), and will be under the conditions of NLOS and LOS It is trained in the feature samples of extraction and corresponding output valve (UWB euclidean distance between node pair) input SVM, the SVR after being trained, such as It is specific as follows shown in Fig. 2:
The recurrence device SVR structure and parameter of support vector machines, the radial direction base are constructed using radial base (RBF) kernel function (RBF) kernel function K (x ', y ') such as following formula:
Wherein, γ is nuclear parameter, and σ is the width of kernel function, and x ' is input feature vector sample, and y ' is corresponding output valve, i.e. UWB Euclidean distance between node pair.
The objective function such as following formula of SVM recurrence device SVR:
s.t.|(wTxk+b)-yk|≤ε, k=1,2 ..., n
Wherein, w, b are to return device parameter, are obtained by training；N is characterized sample total number, xkFor k-th of feature samples, yk For k-th of corresponding output valve, ε is error range.
Above-mentioned objective function is optimized using Lagrange duality function and radial base (RBF) kernel function, optimization knot Fruit is as follows:
Wherein, αkFor Lagrange multiplier.
(4) new channel impulse response CPR is acquired, extracts the feature samples that can characterize propagation conditions, and be inputted Classifier SVC after training, the measured value for sorting out acquisition belongs to LOS propagation conditions or NLOS propagation conditions, and judges LOS Whether the number of the measured value under propagation conditions, which meets location requirement, (for example, located space is two-dimensional space, then needs to meet survey Magnitude is no less than 3, and located space is three-dimensional space, then needs to meet measured value and be no less than 4), if meet demand, utilize Measured value under all LOS propagation conditions constructs square distance matrix, and by square distance Input matrix MDS (Multi- Dimension Scaling, MDS) relative position of corresponding node is obtained in algorithm；If the measured value under LOS propagation conditions Number is not able to satisfy location requirement, then is alleviated using the SVR after training to the measured value under the conditions of NLOS, i.e., by NLOS condition Under feature samples be input to and obtain new correspondence output valve in the SVR trained, then choose measured value and LOS item after alleviating Measured value under part constructs square distance matrix, and the relative position of corresponding node is obtained by MDS algorithm.
Be illustrated in figure 4 the process of MDS algorithm, present invention utilizes based on the generalized barycenter coordinate method of MDS come to UWB into Row positioning, wherein the specific embodiment of MDS algorithm is as follows:
Step 1, square distance matrix D will be obtained by the processed measurement distance progress square of SVM.
Wherein dijFor the distance between i-th of UWB node and j-th of UWB node；
Step 2, center matrix J is calculated:
Wherein N be UWB node total number, I be unit matrix, 1NFor N × 1 and it is worth the matrix for being 1；
Step 3, normalized cumulant square matrices X:
Step 4, X progress singular value decomposition is obtained into singular value Λ and singular value vector V:
X=VAVT
Step 5, singular value Λ is sorted by sequence from big to small, and extracted with located space dimension (for example, positioning is empty Between be two-dimensional space, then number be 2；Located space is three-dimensional space, then number be 3) maximum singular value of same number with it is right The singular value vector V answered1, the singular value of extraction is built into diagonal matrix Λ1, by the diagonal matrix Λ of building1With singular value vector V1 Multiplication obtains the relative coordinate matrix Q of each UWB node:
Wherein Q representation are as follows:(x′i, y 'i) be unknown node opposite position It sets, (x '1, y '1), (x '2, y '2) ..., (x 'N-1, y 'N-1) it is the 1st, 2 ..., the relative position of N-1 anchor node.
(5) using the relative position of each node obtained by MDS algorithm as, unknown node is regarded to the Generalized Barycentric of anchor node (any 3 anchor nodes need to meet not conllinear condition), the Generalized Barycentric for then calculating unknown node relative to anchor node are sat Mark.The true coordinate of unknown node is calculated to the true coordinate of related anchor node using generalized barycenter coordinate.
Fig. 5 show barycentric coodinates expression figure of the unknown node about anchor node, wherein each node is to pass through in specific example The expression of the resulting relative position MDS in a coordinate system is crossed, node l is unknown node, and node i, j, k are anchor node.MDS algorithm The distance between any two points relationship is not changed, therefore the Generalized Barycentric calculated using the resulting relative coordinate of MDS algorithm is sat Generalized barycenter coordinate between mark and actual coordinate is to maintain consistent.Wherein, the calculating of generalized barycenter coordinate is as follows:
a1*x′1+a2*x′2+…+aN-1*x′N-1=x 'i
a1*y′1+a2*y′2+…+aN-1*y′N-1=y 'i
a1+a2+…+aN-1=1
Wherein, (a1, a2..., aN-1) it is generalized barycenter coordinate.
Actual position coordinate is calculate by the following formula to obtain:
xi=a1*x1+a2*x2+…+aN-1*xN-1
yi=a1*y1+a2*y2+…+aN-1*yN-1
Wherein, (xi, yi) be required unknown node actual position coordinate, (x1, y1), (x2, y2) ..., (xN-1, yN-1) be The true coordinate of 1st, 2 ..., N-1 anchor nodes.
Finally illustrate, the above are the preferable specific embodiments of the present invention, but protection scope of the present invention is not limited to This, anyone skilled in the art in the technical scope disclosed by the present invention, can be easily to technology of the invention Scheme is modified or replaced equivalently, and is all intended to be within the scope of the claims of the invention.

Claims (8)

1. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of a kind of NLOS, which is characterized in that the method includes Following steps:
Step 1: using being taken multiple measurements respectively in UWB radio indoors LOS environment and NLOS environment, and acquiring respectively Channel impulse response CPR under LOS environment and NLOS environment, and the feature sample for representing propagation conditions is extracted according to these CPR This.
Step 2: building SVM classifier SVC, and the feature samples and its corresponding mark that will be extracted under the conditions of NLOS and LOS It is trained in label constitutive characteristic Input matrix SVM, the SVC after being trained.
Step 3: building SVM returns device SVR, and the feature samples extracted under the conditions of NLOS and LOS are inputted with corresponding output valve It is trained in SVM, the SVR after being trained, the output valve is UWB euclidean distance between node pair.
Step 4: acquiring new channel impulse response CPR, extract the feature samples for representing propagation conditions, and be inputted training SVC afterwards sorts out measured value using the SVC after training and belongs to LOS propagation conditions or NLOS propagation conditions, and judges LOS Whether the number of the measured value under propagation conditions meets location requirement, if meet demand, only utilizes the survey under LOS propagation conditions Magnitude constructs square distance matrix, and square distance Input matrix MDS algorithm is obtained relative position；If under LOS propagation conditions Population of measured values be not able to satisfy location requirement, then the measured value under the conditions of NLOS is alleviated using the SVR after training, i.e., Feature samples under the conditions of NLOS are input to and obtain new correspondence output valve in the SVR trained, then choose the survey after alleviating Magnitude constructs square distance matrix with the measured value under the conditions of LOS, and obtains relative position by MDS algorithm.
Step 5: calculating the barycentric coodinates about unknown node using the relative position that MDS algorithm obtains, and sat using center of gravity The coordinate of mark and known anchor node calculates the actual position coordinate of unknown node.
2. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of NLOS according to claim 1, feature exist In in the step 1, the feature according to CPR extraction includes:
Received signal strength εr, calculation formula is as follows:
Wherein r (t) is the reception signal amplitude of t moment.
Euclidean distance between node pair d, calculation formula are as follows:
D=c (ti-t0)
Wherein c is the light velocity, tiFor receiving time, t0For the respond request time.
Receive signal maximum amplitude rmax, calculation formula is as follows:
Peak value κ, calculation formula are as follows:
Wherein μ|r|For receive signal amplitude mean value,For the variance for receiving signal amplitude, T is the sampling time.
Average excess delay time TMED, calculation formula is as follows:
Wherein ψ (t)=| r (t) |2/εr。
Root mean square time delay spread time TRMS, calculation formula is as follows:
3. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of NLOS according to claim 1, feature exist In the step 2 specifically includes the following contents:
Nonlinear Classifier is constructed using radial base (RBF) kernel function, carries out feature change using radial base (RBF) kernel function It changes, radial direction base (RBF) kernel function K (x ', the Y ') such as following formula:
K (x ', Y ')=exp (- γ | | x '-Y ' | |2)
Wherein, γ is nuclear parameter, and x ' is input feature vector sample, and Y ' is corresponding label value.
The objective function of SVM classifier SVC such as following formula:
s.t.Yk(wTxk+ b) -1 >=0, k=1,2 ..., n
Wherein, w, b are classifier parameters, are obtained by training；xkFor k-th of feature samples, YkFor the mark of k-th of feature samples Label, n are characterized sample total number.
Above-mentioned objective function is optimized using Lagrange duality function and radial base (RBF) kernel function, optimum results are such as Under:
Wherein, αkFor Lagrange multiplier.
4. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of NLOS according to claim 1, feature exist In the step 3 specifically includes the following contents:
The recurrence device SVR structure and parameter of support vector machines, the radial direction base are constructed using radial base (RBF) kernel function (RBF) kernel function K (x ', y ') such as following formula:
Wherein, γ is nuclear parameter, and σ is the width of kernel function, and x ' is input feature vector sample, and y ' is corresponding output valve.
The objective function such as following formula of SVM recurrence device SVR:
s.t.|(wTxk+b)-yk|≤ε, k=1,2 ..., n
Wherein, w, b are to return device parameter, are obtained by training；N is characterized sample total number, xkFor k-th of feature samples, ykIt is K corresponding output valve, ε is error range.
Above-mentioned objective function is optimized using Lagrange duality function and radial base (RBF) kernel function, optimum results are such as Under:
Wherein, αkFor Lagrange multiplier.
5. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of NLOS according to claim 1, feature exist In, in the step 4, the location requirement specifically: located space is two-dimensional space, then needs to meet measured value and be no less than 3 A, located space is three-dimensional space, then needs to meet measured value and be no less than 4.
6. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of NLOS according to claim 1, feature exist In, in the step 4, square distance Input matrix MDS algorithm is obtained into relative position, specific as follows:
Obtain square distance matrix:
Wherein dijFor the distance between i-th of UWB node and j-th of UWB node；
Calculate center matrix J:
N be UWB node total number, I be unit matrix, 1NFor N × 1 and it is worth the matrix for being 1；
Normalized cumulant square matrices:
X progress singular value decomposition is obtained into singular value Λ and singular value vector V；
X=VAVT
By singular value Λ by sequence arrangement from big to small, and extract with the maximum singular value of located space dimension same number with Corresponding singular value vector V1, the singular value of extraction is built into diagonal matrix Λ1, by the diagonal matrix Λ of building1With singular value vector V1Multiplication obtains the relative coordinate matrix Q of each UWB node
Wherein Q representation are as follows:(x′i, y 'i) be unknown node relative position, (x′1, y '1), (x '2, y '2) ..., (x 'N-1, y 'N-1) it is the 1st, 2 ..., the relative position of N-1 anchor node.
7. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of NLOS according to claim 6, feature exist In, in the step 5, using MDS algorithm obtain relative position calculate the barycentric coodinates about unknown node specifically: will The relative position of unknown node regards the Generalized Barycentric of other nodes as, and solves generalized barycenter coordinate by following formula:
a1*x′1+a2*x′2+…+aN-1*x′N-1=x 'i
a1*y′1+a2*y′2+…+aN-1*y′N-1=y 'i
a1+a2+…+aN-1=1
Wherein, (a1, a2..., aN-1) it is generalized barycenter coordinate.
8. the UWB localization method that SVM is combined with barycentric coodinates under the conditions of NLOS according to claim 7, feature exist In in the step 5, actual position coordinate is calculate by the following formula to obtain:
xi=a1*x1+a2*x2+…+aN-1*xN-1
yi=a1*y1+a2*y2+…+aN-1*yN-1
Wherein, (xi, yi) be required unknown node actual position coordinate, (x1, y1), (x2, y2) ..., (xN-1, yN-1) it is the 1st, The true coordinate of 2 ..., N-1 anchor nodes.
CN201910764643.6A 2019-08-19 2019-08-19 UWB positioning method combining SVM (support vector machine) and barycentric coordinate under NLOS (non line of sight) condition Active CN110488222B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910764643.6A CN110488222B (en) 2019-08-19 2019-08-19 UWB positioning method combining SVM (support vector machine) and barycentric coordinate under NLOS (non line of sight) condition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910764643.6A CN110488222B (en) 2019-08-19 2019-08-19 UWB positioning method combining SVM (support vector machine) and barycentric coordinate under NLOS (non line of sight) condition

Publications (2)

Publication Number Publication Date
CN110488222A true CN110488222A (en) 2019-11-22
CN110488222B CN110488222B (en) 2021-06-22

Family

ID=68551506

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910764643.6A Active CN110488222B (en) 2019-08-19 2019-08-19 UWB positioning method combining SVM (support vector machine) and barycentric coordinate under NLOS (non line of sight) condition

Country Status (1)

Country Link
CN (1) CN110488222B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111294731A (en) * 2020-01-16 2020-06-16 珠海格力电器股份有限公司 UWB indoor positioning method, device, electronic equipment and storage medium
CN111432364A (en) * 2020-03-25 2020-07-17 哈尔滨工程大学 Radial basis function neural network-based non-line-of-sight error suppression method
CN111541988A (en) * 2020-04-17 2020-08-14 北京理工大学重庆创新中心 Three-dimensional indoor positioning method based on barycentric coordinate and Taylor expansion
CN112929826A (en) * 2021-03-23 2021-06-08 浙江工业大学 Indoor hybrid positioning method based on UWB triangulation positioning and fingerprint information
WO2021228209A1 (en) * 2020-05-15 2021-11-18 大唐移动通信设备有限公司 Measurement reporting method, measurement reporting device, and positioning server
CN114928811A (en) * 2022-05-05 2022-08-19 南京理工大学 Ultra-wideband mine underground personnel accurate positioning method based on neural network

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101561495A (en) * 2009-06-01 2009-10-21 长讯通信服务有限公司 Method for three-dimensionally positioning network node of wireless sensor

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017100706A1 (en) * 2015-12-09 2017-06-15 Origin Wireless, Inc. Method, apparatus, and systems for wireless event detection and monitoring
US20180061037A1 (en) * 2016-08-24 2018-03-01 The Boeing Company Dynamic, persistent tracking of multiple field elements
CN106772231B (en) * 2016-12-23 2019-10-11 深圳市金溢科技股份有限公司 Electronic tag, base station, positioning system and method based on UWB
CN106950537A (en) * 2017-04-28 2017-07-14 浙江大学 A kind of Distributed localization method based on UWB
CN107563437B (en) * 2017-08-31 2020-12-04 广州中海达创新科技集团有限公司 Ultra-wideband non-line-of-sight identification method based on random forest

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101561495A (en) * 2009-06-01 2009-10-21 长讯通信服务有限公司 Method for three-dimensionally positioning network node of wireless sensor

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111294731A (en) * 2020-01-16 2020-06-16 珠海格力电器股份有限公司 UWB indoor positioning method, device, electronic equipment and storage medium
CN111432364A (en) * 2020-03-25 2020-07-17 哈尔滨工程大学 Radial basis function neural network-based non-line-of-sight error suppression method
CN111432364B (en) * 2020-03-25 2021-05-11 哈尔滨工程大学 Radial basis function neural network-based non-line-of-sight error suppression method
CN111541988A (en) * 2020-04-17 2020-08-14 北京理工大学重庆创新中心 Three-dimensional indoor positioning method based on barycentric coordinate and Taylor expansion
CN111541988B (en) * 2020-04-17 2021-11-23 北京理工大学重庆创新中心 Three-dimensional indoor positioning method based on barycentric coordinate and Taylor expansion
WO2021228209A1 (en) * 2020-05-15 2021-11-18 大唐移动通信设备有限公司 Measurement reporting method, measurement reporting device, and positioning server
CN112929826A (en) * 2021-03-23 2021-06-08 浙江工业大学 Indoor hybrid positioning method based on UWB triangulation positioning and fingerprint information
CN112929826B (en) * 2021-03-23 2022-07-15 浙江工业大学 Indoor hybrid positioning method based on UWB triangulation positioning and fingerprint information
CN114928811A (en) * 2022-05-05 2022-08-19 南京理工大学 Ultra-wideband mine underground personnel accurate positioning method based on neural network
CN114928811B (en) * 2022-05-05 2024-05-07 南京理工大学 Ultra-wideband mine underground personnel accurate positioning method based on neural network

Also Published As

Publication number Publication date
CN110488222B (en) 2021-06-22

Similar Documents

Publication Publication Date Title
CN110488222A (en) The UWB localization method that SVM is combined with barycentric coodinates under the conditions of a kind of NLOS
CN106886023B (en) A kind of Radar Echo Extrapolation method based on dynamic convolutional neural networks
CN108696932B (en) Outdoor fingerprint positioning method using CSI multipath and machine learning
CN105929378B (en) Combine the direct tracking of time delay and Doppler frequency based on external sort algorithm
CN103605110B (en) Based on the indoor passive object localization method of received signal strength
CN103501538B (en) Based on the indoor orientation method of multipath energy fingerprint
CN111308471B (en) Rain, snow and hail classification monitoring method based on semi-supervised domain adaptation
CN104519571B (en) A kind of indoor orientation method based on RSS
CN108717175A (en) Indoor fingerprint positioning method based on region division and sparse support vector regression
CN106793082A (en) A kind of positioning of mobile equipment method in WLAN/ bluetooth heterogeneous network environments
CN104807460A (en) Indoor positioning method and system for unmanned aerial vehicle
CN107087256A (en) A kind of fingerprint cluster method and device based on WiFi indoor positionings
CN106842118A (en) For the time difference positioning method of Distributed Multi positioning monitoring system
CN103886207A (en) Nest multiple-input and multiple-output radar DOA estimating method based on compressed sensing
CN106941718A (en) A kind of mixing indoor orientation method based on signal subspace fingerprint base
CN107703480A (en) Mixed kernel function indoor orientation method based on machine learning
CN103885030A (en) Locating method of mobile node in wireless sensor network
CN105259550A (en) Compressed sensing-based multi-input and multi-output radar two-dimensional angle estimation method
CN104616059B (en) A kind of Wave arrival direction estimating method based on quantum particle swarm
CN107436427A (en) Space Target Motion Trajectory and radiation signal correlating method
CN109922427A (en) Utilize the intelligent radio positioning system and method for large scale array antenna
CN110049549A (en) More fusion indoor orientation methods and its system based on WiFi fingerprint
CN103616661A (en) Robust far-field narrowband signal source number estimation method
CN107360552A (en) A kind of indoor orientation method of multi-categorizer overall situation dynamic fusion
CN106353720A (en) Multi-station continuous positioning model based on TDOA/GROA (time different of arrival/gain ratio of arrival)

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant