CN105163385A - Localization algorithm based on sector overlapping area of clustering analysis - Google Patents

Abstract

The invention discloses a localization algorithm based on sector overlapping area of clustering analysis. The method comprises the following steps: S1, fitting a Gaussian ranging curve of received signal strength indicators (RSSIs); S2, performing partitioning based on a sector overlapping area; S3, performing clustering analysis; S4, preforming distance correction; and S5, performing dynamic weighted centroid localization. In the method, Gaussian sieving is added according to errors caused during RSSI measurement in order to eliminate the RSSIs having relatively great influences; and then, the sector overlapping area based on the clustering analysis is established according to the difference in localization distribution of the RSSIs, and correction factors are added in different partitioned areas through geometrical locations among reference nodes to perform distance correction. Meanwhile, localization is performed through a dynamic weighted centroid localization algorithm to get the location of an unknown node. The localization algorithm has the advantages of high localization accuracy, low computation complexity, convenience in hardware implementation and the like.

Description

A kind of location algorithm based on the cluster analysis of fan-shaped overlapping region

Technical field

The present invention relates to wireless sensor network technology field, particularly a kind of location algorithm based on the cluster analysis of fan-shaped overlapping region.

Background technology

The locating information of wireless sensor network (WSN, Wirelesssensornetwork) mobile node (MN, MobileNode) has great importance to aspects such as personnel positioning, coal mine environment, military surveillances.WSN location technology can be divided into based on range finding and without the need to ranging technology, but higher based on the location technology positioning precision of range finding.Ranging technology based on ranging localization algorithm have based on RSSI (Receivedsignalstrengthindicator, received signal strength indicator value), based on the time of advent, based on the time of advent difference with based on angle of arrival.Wherein, based on RSSI ranging localization technology because of have without the need to extra hardware benefits widely studied scholar study.Based on RSSI location technology by known base station (BS, BaseStation) with mobile node (MN, MobileNode) set up the propagation loss model between propagation loss and distance, the location algorithms such as associating maximum-likelihood method, three limit positioning modes, weighted mass center determine the position of MN.

Due to RSSI in WSN transmitting procedure owing to being subject to the obstruction of the barriers such as fixed obstacle (floor, wall) and moving obstacle (animal, human body), thus produce rapid fading and slow fading affects its accuracy.Therefore, the ranging localization technology based on RSSI improves the accuracy of MN position by range finding and localization method.In the impact that minimizing range error in indoor is brought from weakening non line of sight range finding (barrier), in outdoor reduction range error then from the range finding impact that weakening multipath, scattering, diffraction cause.Localization method weakens position error and positions from the distinct regions that Region dividing is equal.

According to above-mentioned weakening range finding and position error method, the region partitioning algorithm proposed in recent years mainly contains: IndooradaptiveRSSIlocalizationalgorithmbasedongraphtheor yandfuzzyclusteringinwirelesssensornetworks (IAL-GT-FC), WSNLocalizationMethodUsingIntervalDataClustering (RSSI-D), based on the constraint KNN Indoor Locating Model of geometry cluster fingerprint base, a kind of Localization Algorithm for Wireless Sensor Networks etc. utilizing K mean cluster; Have in patent documentation " Localization Algorithm for Wireless Sensor Networks selecting anchor node based on distance cluster " (application number CN201510070653.1), " indoor orientation method based on signal receiving strength instruction correlation " (application number CN201410546502.4), " a kind of node positioning method of many precision based on regional determination " (application number CN201410400684.4).The computational complexity of above-mentioned cluster analysis location algorithm is larger.Therefore, calculate simple in the urgent need to one, the cluster analysis location algorithm that performance is better.

Summary of the invention

The object of the invention is to overcome the shortcoming of prior art and deficiency, a kind of location algorithm based on the cluster analysis of fan-shaped overlapping region is provided.

Object of the present invention is achieved through the following technical solutions: a kind of location algorithm based on the cluster analysis of fan-shaped overlapping region (Localizationalgorithmbasedonsectoroverlappingareaofclust eringanalysis, be abbreviated as LA-SOACA), specifically comprise the following steps:

The Gauss of S1, matching received signal strength indicator value RSSI finds range curve: calculate the signal strength loss in Gauss Distribution Fitting during Signal transmissions 1m in current environment and RSSI Gauss described in signal intensity attenuation index n thus matching finds range mathematic(al) representation;

S2, to divide based on fan-shaped overlapping region: the feature according to RSSI distribution with regional differentiation, using square area as application scenarios, a reference node is respectively placed at square four summits A, B, C, D, be respectively BS1, BS2, BS3, BS4, region in square, using described foursquare length of side R as fan-shaped covering radius, is divided into blind area, non-concentration zones and concentration zones according to by the number of times of fan-shaped covering by described reference node;

S3, cluster analysis: the barycenter to choose in blind area, non-concentration zones and concentration zones bunch, the RSSI data set unknown mobile node MN screened through Gauss according to Euclidean distance criterion calculates with the barycenter RSSI data set in three regions bunch respectively, calculates minimum value person and then can judge that unknown mobile node MN is at this area coverage;

S4, distance correction: before selection reference node location, select reference node far away of adjusting the distance with the reference node of unknown node MN close together and carry out measuring distance correction;

S5, dynamic weighting center coordination: dynamic weighting centroid localization algorithm has regional center characteristic in conjunction with blind area, non-concentration zones, concentration zones RSSI data set distributional difference characteristic and weighted mass center location WCL, add dynamic factor and improve weighted mass center location WCL positioning precision, thus unknown node MN position is solved.

Described step S1 specifically comprises the following steps:

S11, Maximum-likelihood estimation is carried out to the k group data of each range points d of current environment sample, estimate Gaussian Profile P respectively _raverage with variance be shown below:

$\widehat{P_r}\left(d\right)=\frac{1}{k}{\mathrm{\Σ}}_{j=1}^k{P}_r\[d\left(j\right)\]=\stackrel{\‾}{P_r\left(d\right)},$

$\widehat{\mathrm{\σ}}{\left(d\right)}^2=\frac{1}{k}{\mathrm{\Σ}}_{j=1}^k{\[P_r\[d\left(j\right)\]-\stackrel{\‾}{P_r\left(d\right)}\]}^2;$

S12, utilize normpdf to solve each range points d probability to be respectively α ₁p _r1(d) and α ₂p _r2(d), solution procedure is as follows:

$F\[P_r\left(d\right)\]=\frac{1}{\widehat{\mathrm{\σ}}\left(d\right)\sqrt{2\mathrm{\π}}}{\∫}_{-\mathrm{\∞}}^{P_R}{e}^{-\frac{{(x-\stackrel{\‾}{P_r\left(d\right)})}^2}{2\widehat{\mathrm{\σ}}{\left(d\right)}^2}}dx,$

P[P _r(d)≤P _r1(d)]＝F[P _r1(d)]＝α ₁，

P[P _r(d)≥P _r2(d)]＝1-F[P _r2(d)]＝1-α ₂；

S13, to filter out each range points probability be [α ₁, α ₂] interval P _rvalue, obtains at P _r1(d) and P _r2p between (d) _rvalue, by the P of screening _rvalue is defined as high-probability event, screening the P obtained through Gaussian Profile _rbe stored in Beacon_val_gauss [m], each range points RSSI average is tried to achieve by following formula

${\stackrel{\‾}{P_r\left(d\right)}}^{\′}=\frac{1}{m}{\mathrm{\Σ}}_{j=1}^{m}Beacon\_val\_gauss\[j\],$

Wherein m is through Gaussian Profile screening P _rnumber;

S14, by what try to achieve substituting into following formula utilizes least square method to carry out curve fitting, and draws fit mathematics expression formula

$P_r\left(d\right)={\stackrel{\‾}{P_r\left(d\right)}}^{\′}-10{\mathrm{nlog}}_{10}\left(d\right).$

Described step S3 specifically comprises the following steps:

S31, choose the barycenter that barycenter in blind area, non-concentration zones and region, 3, concentration zones does each region bunch;

S32, to calculate in described 3 regions barycenter to the distance of four reference nodes BS1, BS2, BS3, BS4, the RSSI Gauss substituted in described step S14 models fitting curve of finding range can obtain the RSSI data set of the barycenter respective base station in each region bunch, the barycenter RSSI data set in each region bunch is sorted from big to small, obtains the barycenter RSSI data set RSSI in each region bunch _x;

S33, unknown node MN receive the RSSI data set of four reference nodes BS1, BS2, BS3, BS4, also sorted from big to small, obtain the RSSI data set RSSI of unknown node MN by Gaussian data process _mN;

S34, calculate the barycenter distinctiveness ratio in unknown node MN and each region bunch, selection similarity function Euclidean distance function is as the judgement of similarity degree, and Euclidean distance formula is shown below, wherein RSSI _xfor the barycenter RSSI data set in each region bunch, the number of p contained by RSSI data centralization

$D({\mathrm{RSSI}}_{MN},{\mathrm{RSSI}}_x)=\left|\right|{\mathrm{RSSI}}_{MN}-{\mathrm{RSSI}}_x|{|}_2={\left({\mathrm{\Σ}}_{j=1}^p{|{\mathrm{RSSI}}_{MN}\left(j\right)-{\mathrm{RSSI}}_x\left(j\right)|}^2\right)}^{\frac{1}{2}};$

S35, compare the data set RSSI of unknown node MN _mNwith the barycenter RSSI data set D (RSSI in each region bunch _mN, RSSI _x) difference, difference reckling, can judge that unknown node MN belongs to RSSI data set RSSI _xpoint in this region.

Described step S4 specifically comprises the following steps:

S41, record AB respectively, the RSSI value RSSI between BC, CD, DA _aS, RSSI _bC, RSSI _cD, RSSI _dA;

S42, data set RSSI at unknown node MN _mNin, try to achieve matching RSSI Gauss curve of finding range according to step S14 and obtain unknown node MN and four the distance d that reference node BS1, BS2, BS3, BS4 are nearest _min;

S43, selection RSSI range finding calibration model, the impact that erasure signal strength retrogression index n brings: try to achieve unknown node MN data set RSSI according to following formula _mNclosest approach is to the reference distance d of reference node _{refer_x},

$d_{refer\_x}={10}^{\frac{P_r\left(d_1\right)-{\mathrm{RSSI}}_{MN\_x}}{P_r\left(d_1\right)-{\mathrm{RSSI}}_{xy}}},$

Wherein, RSSI _xyfor recording AB, the RSSI reference measurement values between BC, CD, DA, RSSI _{mN_x}with d _{refer_x}in x be the distance correction number of times carried out according to territory, sector coverage area, each region number, being 1 at blind area distance correction number of times, is 2 at distance correction number of times, does not need to carry out distance correction in concentration zones;

S44, ask minimum distance d _minwith reference distance d _{refer_x}between Error Absolute Value d _{error_x};

S45, foundation data set RSSI _mNclosest approach to distance correction coefficient gamma corresponding to reference node BS, according to formula

d _{max_x}＝γ[P _r(d ₁)-10nlog ₁₀(d)]

Ask data set RSSI _mNcomparatively far point distance d _{max_x}, wherein γ is distance correction coefficient, and the computing formula of described distance correction coefficient is

γ＝(1-β) ^τ，

τ is distance correction index, and β is distance correction percentage, and described distance correction percentage refers to Error Absolute Value d _{error_x}account for minimum distance d _minpercentage, computing formula is

$\mathrm{\β}=\frac{d_{error\_x}}{d_{\min }}.$

In described step S2, described concentration zones take radius as the region of the fan-shaped covering four times of R, non-concentration zones take radius as the fan-shaped covering three times of R and removes the region of concentration zones, and blind area take radius as the fan-shaped covering twice of R and removes the region of concentration zones and non-concentration zones.

In described step S5, dynamic weighting centroid localization algorithm defines blind area, non-concentration zones and concentration zones dynamic factor and use weighted mass center to locate displace analysis that WCL carries out unknown node MN respectively, is shown below.

$MN(x,y)=\frac{{\mathrm{\Σ}}_{j=1}^z\left(\frac{1}{d^{{\mathrm{\ω}}_j}}{\mathrm{BS}}_j\right(x,y\left)\right)}{{\mathrm{\Σ}}_{j=1}^z{w}_j}$

Wherein, ω _jfor the dynamic factor of a corresponding jth reference node, MN (x, y) is unknown node MN coordinate, BS _j(x, y) is the coordinate of a jth reference node, and z is reference node number.

Blind area dynamic factor is chosen:

RSSI _Mn(1),RSSI _MN(2)→ω ₁＝ω ₂＝1

RSSI _MN(3),RSSI _MN(4)→ω ₃＝ω ₄＝2

Non-concentration zones dynamic factor is chosen:

RSSI _MN(1),RSSI _MN(2)→ω ₁＝ω ₂＝1

RSSI _MN(3)→ω ₃＝1

RSSI _MN(4)→ω ₄＝2

Concentration zones dynamic factor is chosen:

RSSI _MN(1),RSSI _MN(2)→ω ₁＝ω ₂＝1

RSSI _MN(3),RSSI _MN(4)→ω ₃＝ω ₄＝1

Compared with prior art, tool has the following advantages and beneficial effect in the present invention:

1, the present invention is according to actual environment, adds Gauss and screens RSSI data, namely reject low probability event, makes matching RSSI range finding curve more meet the requirement of current environment, thus improves the accuracy of range finding.

2, the present invention is based on fan-shaped overlapping region and be divided into three regions, be i.e. blind area, non-concentration zones and blind area, zoning is less, and reference node quantity only needs 4, and positioning precision is higher, calculates simple.

3, the present invention adds distance correction coefficient according to varying environment, makes range accuracy higher by the actual distance correction coefficient choosing the best.

4, the present invention is in conjunction with the feature of blind area, non-concentration zones and concentration zones, and the dynamic factor in dynamic weighting centroid localization algorithm chooses the location requirement adapting to each region, improves positioning precision.

Accompanying drawing explanation

Fig. 1 is the flow chart of a kind of location algorithm based on the cluster analysis of fan-shaped overlapping region of the present invention.

The fan-shaped overlapping region that Fig. 2 is location algorithm described in Fig. 1 divides schematic diagram.

Fig. 3 is that in the embodiment of the present invention, received signal strength indicator value RSSI Gauss finds range matched curve schematic diagram.

Fig. 4 is based on location algorithm (LA-SOACA) the position error absolute average of fan-shaped overlapping region cluster analysis and τ value relation curve schematic diagram in the embodiment of the present invention.

Fig. 5 locates (WCL) position error based on the location algorithm (LA-SOACA) of fan-shaped overlapping region cluster analysis and weighted mass center in the embodiment of the present invention to contrast schematic diagram.

Fig. 6 contrasts schematic diagram based on the location algorithm (LA-SOACA) of fan-shaped overlapping region cluster analysis, weighted mass center location (WCL) with maximum-likelihood method (MaximumLikelihoodEstimate, MLE) position error in the embodiment of the present invention.

Embodiment

Below in conjunction with embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are not limited thereto.

A kind of location algorithm based on the cluster analysis of fan-shaped overlapping region (Localizationalgorithmbasedonsectoroverlappingareaofclust eringanalysis, LA-SOACA), as shown in Figure 1, specifically comprises the following steps:

The Gauss of S1, matching received signal strength indicator value RSSI finds range curve: calculate the signal strength loss in Gauss Distribution Fitting during Signal transmissions 1m in current environment and RSSI Gauss described in signal intensity attenuation index n thus matching finds range mathematic(al) representation;

S2, to divide based on fan-shaped overlapping region: the feature according to RSSI distribution with regional differentiation, using square area as application scenarios, a reference node is respectively placed at square four summits A, B, C, D, be respectively BS1, BS2, BS3, BS4, region in square, using described foursquare length of side R as fan-shaped covering radius, is divided into blind area, non-concentration zones and concentration zones according to by the number of times of fan-shaped covering by described reference node;

S3, cluster analysis: the barycenter to choose in blind area, non-concentration zones and concentration zones bunch, be a region in three regions unknown mobile node MN through the RSSI data set classification that Gauss screens according to Euclidean distance criterion, make the barycenter similitude in RSSI data set and each region bunch large as much as possible;

S4, distance correction: before selection reference node location, select reference node far away of adjusting the distance with the reference node of unknown node MN close together and carry out measuring distance correction;

S5, dynamic weighting center coordination: dynamic weighting centroid localization algorithm has regional center characteristic in conjunction with blind area, non-concentration zones, concentration zones RSSI data set distributional difference characteristic and weighted mass center location (WCL), add dynamic factor and improve WCL positioning precision, thus unknown node MN position is solved.

Further, step S1 specifically comprises the following steps:

Step S11, carries out Maximum-likelihood estimation to the k group data of each range points d of current environment sample, estimates Gaussian Profile P respectively _raverage with variance be shown below:

$\widehat{P_r}\left(d\right)=\frac{1}{k}\underset{j=1}{\overset{k}{\Σ}}{P}_r\[d\left(j\right)\]=\stackrel{\‾}{P_r\left(d\right)}$

$\widehat{\mathrm{\σ}}{\left(d\right)}^2=\frac{1}{k}\underset{j=1}{\overset{k}{\Σ}}{\[P_r\[d\left(j\right)\]-\stackrel{\‾}{P_r\left(d\right)}\]}^2$

Step S12, utilizes normpdf to ask each range points d probability to be respectively α ₁p _r1(d) and α ₂p _r2(d), its process is as follows:

$F\[P_r\left(d\right)\]=\frac{1}{\widehat{\mathrm{\σ}}\left(d\right)\sqrt{2\mathrm{\π}}}{{\∫}_{-\mathrm{\∞}}^{P_R}e}^{-\frac{{(x-\stackrel{\‾}{P_r\left(d\right)})}^2}{2\widehat{\mathrm{\σ}}{\left(d\right)}^2}}dx$

P[P _r(d)≤P _r1(d)]＝F[P _r1(d)]＝α ₁

P[P _r(d)≥P _r2(d)]＝1-F[P _r2(d)]＝1-α ₂

Step S13, can filter out each range points probability is [α ₁, α ₂] interval P _rvalue, obtains at P _r1(d) and P _r2p between (d) _rvalue, by the P of screening _rvalue is defined as high-probability event, screening the P obtained through Gaussian Profile _rbe stored in Beacon_val_gauss [m], each range points RSSI average is tried to achieve by following formula, and wherein m is through Gaussian Profile screening P _rnumber.

${\stackrel{\‾}{P_r\left(d\right)}}^{\′}=\frac{1}{m}\underset{j=1}{\overset{m}{\Σ}}Beacon\_val\_gauss\[j\]$

Step S14, by what try to achieve substituting into following formula utilizes least square method to carry out curve fitting, and can draw fit mathematics expression formula.

$P_r\left(d\right)={\stackrel{\‾}{P_r\left(d\right)}}^{\′}-10{\mathrm{nlog}}_{10}\left(d\right)$

Further, in step S2, A, B, C, D do fan-shaped in tetra-summits with radius R, in square, hand over E respectively, F, G, H tetra-intersection points.Each region is respectively with fan-shaped covering number definition blind area, non-concentration zones and region, three, concentration zones.Described concentration zones take radius as the region of the fan-shaped covering four times of R, and non-concentration zones take radius as the fan-shaped covering three times of R and removes the region of concentration zones, and blind area take radius as the fan-shaped covering twice of R and removes the region of concentration zones and non-concentration zones.Blind area is defined as reference node with radius R only territory, two sector coverage area, and namely specify A-E-B, B-H-C, C-G-D, D-F-A region is blind area; Non-concentration zones is defined as reference node with radius R only territory, three sector coverage area, and namely specify A-F-E, B-E-H, C-H-G, D-G-F region is non-concentration zones; Concentration zones is defined as reference node with the four fan-shaped overlay area of radius R, i.e. E-F-G-H, as shown in Figure 2.

Further, step S3 specifically comprises the following steps:

Step S31, choose the barycenter that barycenter in blind area, non-concentration zones and region, 3, concentration zones does each region bunch, namely the barycenter in blind area bunch is Blind_Centroid (x, y), barycenter in non-concentration zones bunch is Decentralized_Centroid (x, y), the barycenter in concentration zones bunch is Concentration_Centroid (x, y).

Step S32, barycenter Blind_Centroid (the x to calculate in blind area bunch, y) the barycenter Decentralized_Centroid (x, in non-concentration zones bunch, y) with concentration zones in bunch barycenter Concentration_Centroid (x, y) to the distance of four reference nodes BS1, BS2, BS3, BS4, substitute into the RSSI Gauss that tries to achieve of step S13 models fitting curve of finding range and can obtain the RSSI data set of the barycenter respective base station in each region bunch, the barycenter RSSI data set in each region bunch is sorted from big to small, obtains the barycenter RSSI data set RSSI of blind area bunch _blind, the barycenter RSSI data set RSSI of non-concentration zones bunch _{decentralized}, the barycenter RSSI data set RSSI of concentration zones bunch _{concentration}.

Step S33, unknown node MN receive the RSSI data set of four reference nodes BS1, BS2, BS3, BS4, also sorted from big to small, obtain the RSSI data set RSSI of unknown node MN by Gaussian data process _mN.

Step S34, in order to judge which bunch the data set RSSI of unknown node MN belongs to, need to calculate the barycenter distinctiveness ratio in unknown node MN and each region bunch, selection similarity function Euclidean distance function is as the judgement of similarity degree, Euclidean distance formula is shown below, wherein RSSI _xfor the barycenter RSSI data set in each region bunch, the number of p contained by RSSI data centralization.

$D({\mathrm{RSSI}}_{MN},{\mathrm{RSSI}}_x)=\left|\right|{\mathrm{RSSI}}_{MN}-{\mathrm{RSSI}}_x|{|}_2={\left({\mathrm{\Σ}}_{j=1}^p{|{\mathrm{RSSI}}_{MN}\left(j\right)-{\mathrm{RSSI}}_x\left(j\right)|}^2\right)}^{\frac{1}{2}}$

Step S35, compares the data set RSSI of unknown node MN _mNwith the barycenter RSSI data set D (RSSI in each region bunch _mN, RSSI _x) difference, difference reckling, can judge that unknown node MN belongs to RSSI data set RSSI _xpoint in this region.

Further, step S4 specifically comprises the following steps:

Step S41, records AB respectively, the RSSI value RSSI between BC, CD, DA _aB, RSSI _bC, RSSI _cD, RSSI _dA.

Step S42, at the data set RSSI of unknown node MN _mNin, try to achieve matching RSSI Gauss curve of finding range according to step S14 and obtain unknown node MN and four the distance d that reference node BS1, BS2, BS3, BS4 are nearest _min.

Step S43, select RSSI range finding calibration model, the impact that erasure signal strength retrogression index n brings, tries to achieve unknown node MN data set RSSI according to following formula _mNclosest approach is to the reference distance d of reference node _{refer_x}.Wherein, RSSI _xyfor recording AB, the RSSI reference measurement values between BC, CD, DA, RSSI _{mN_x}with d _{refer_x}middle x, for carry out distance correction according to territory, sector coverage area, each region number, is 1 at blind area distance correction number of times, is 2 at distance correction number of times, does not need to carry out distance correction in concentration zones.

$d_{refer\_x}={10}^{\frac{P_r\left(d_1\right)-{\mathrm{RSSI}}_{MN\_x}}{P_r\left(d_1\right)-{\mathrm{RSSI}}_{xy}}}$

Step S44, asks minimum distance d _minwith reference distance d _{refer_x}between Error Absolute Value d _{error_x}.

Definition 1: β is distance correction percentage, i.e. Error Absolute Value d _{error_x}account for minimum distance d _minpercentage, be shown below, wherein, x and x in step S43 express look like consistent.

$\mathrm{\β}=\frac{d_{error\_x}}{d_{min}}$

Definition 2: γ is distance correction coefficient, and be shown below, wherein, τ is distance correction index.

γ＝(1-β) ^τ

Step S45, according to data set RSSI _mNclosest approach, to distance correction coefficient gamma corresponding to reference node BS, asks data set RSSI _mNcomparatively far point distance d _{max_x}, be shown below, wherein, x and x in step S43 express look like consistent.

d _{max_x}＝γ[P _r(d ₁)-10nlog ₁₀(d)]

Further, step S5 detailed process is as follows:

Dynamic weighting centroid localization algorithm has regional center characteristic in conjunction with blind area, non-concentration zones, concentration zones RSSI data set distributional difference characteristic and weighted mass center location (WCL), adds dynamic factor and improves WCL positioning precision.According to above-mentioned thinking, dynamic weighting centroid localization algorithm defines blind area, non-concentration zones and concentration zones dynamic factor respectively and uses WCL to carry out the displace analysis of unknown node MN, is shown below.

$MN(x,y)=\frac{{\mathrm{\Σ}}_{j=1}^z\left(\frac{1}{d^{{\mathrm{\ω}}_j}}{\mathrm{BS}}_j\right(x,y\left)\right)}{{\mathrm{\Σ}}_{j=1}^z{w}_j}$

Wherein, ω _jfor the dynamic factor of a corresponding jth reference node, MN (x, y) is unknown node MN coordinate, BS _j(x, y) is the coordinate of a jth reference node, and z is reference node number.

Blind area dynamic factor is chosen:

RSSI _MN(1),RSSI _MN(2)→ω ₁＝ω ₂＝1

RSSI _MN(3),RSSI _MN(4)→ω ₃＝ω ₄＝2

Non-concentration zones dynamic factor is chosen:

RSSI _MN(1),RSSI _MN(2)→ω ₁＝ω ₂＝1

RSSI _MN(3)→ω ₃＝1

RSSI _MN(4)→ω ₄＝2

Concentration zones dynamic factor is chosen:

RSSI _MN(1),RSSI _MN(2)→ω ₁＝ω ₂＝1

RSSI _MN(3),RSSI _MN(4)→ω ₃＝ω ₄＝1

Two, about the determination of each relevant parameter of the inventive method:

Experiment is one: RSSI Gauss find range matched curve

Step one: choose the Bluetooth Low Energy (BluetoothLowEnergy that Texas Instrument produces, BLE) chip CC2541 is as node, Range finding experiments environment is a slice depletion region, Range finding experiments node is all set as uniform height (the height 1m on distance ground), and node transmitting power is set as 0dbm.

Distance measurement ranges between step 2: BS and MN is 0 ~ 20m, and measure k=100 group RSSI data at interval of 0.5m, data collection cycle is 1 time per second.

Step 3: choose α ₁=0.3, α ₂=0.7 screens concentrated probability as Gauss, the RSSI data of experiment sampling is carried out finding range matched curve as shown in Figure 3 as step S11, S12, S13, S14 process had both obtained RSSI Gauss, =-61dbm, n=-1.9.

Experiment two: distance correction index τ affects LA-SOACA

Step one: choose the Bluetooth Low Energy (BluetoothLowEnergy that Texas Instrument produces, BLE) chip CC2541 is as node, positioning experiment environment area is 8m × 8m area of space (square length of side R is 8m), positioning experiment node is all set as uniform height (the height 1m on distance ground), and node transmitting power is set as 0dbm.

Step 2: choose R=8m, at A, B, C, D places 4 reference node BS, and overlapping region as fan-shaped in step S2 is divided into 3 regions, and wherein there are 4 zonules blind area and non-concentration zones, concentration zones only has 1 zonule, puts unknown node MN respectively at random and position experiment test in 3 regions.

Step 3: in experimentation, measuring distance modified index τ on the impact of blind area and non-concentration zones algorithm positioning performance, using following formula as evaluation index, in formula, MN (x, y) _realfor testing the actual coordinate of unknown node, MN (x, y) is LA-SOACA location estimation coordinate, and num is experiment coordinate number.Error _averfor the error absolute average of LA-SOACA multiple bearing, error _averreflection LA-SOACA location algorithm locating effect.Error _averless, positioning precision is higher, and position error absolute average and τ value relation curve are as shown in Figure 4.

${\mathrm{error}}_{aver}=\frac{1}{num}{\mathrm{\Σ}}_{j=1}^{num}|MN(x,y)-MN{(x,y)}_{real}|$

Step 4: as shown in Figure 4, there is optimum value in current environment measuring distance modified index τ, blind area and non-concentration zones positioning performance the best when τ=0.5.

Step 5: this algorithm and WCL algorithm, maximum likelihood algorithm (MLE) are compared, according to the analysis of the modified index τ that adjusts the distance, setpoint distance modified index τ be 0.5, unknown node MN in the experiment position error of the algorithms of different of area of space each position as shown in Fig. 5,6 and table 1.Table 1 is experimental verification table of the present invention: LA-SOACA, WCL and MLE position error contrast signal table.

Table 1

Above-described embodiment is the present invention's preferably execution mode; but embodiments of the present invention are not restricted to the described embodiments; change, the modification done under other any does not deviate from Spirit Essence of the present invention and principle, substitute, combine, simplify; all should be the substitute mode of equivalence, be included within protection scope of the present invention.

Claims (6)

1. based on a location algorithm for fan-shaped overlapping region cluster analysis, it is characterized in that, specifically comprise the following steps:

The Gauss of S1, matching received signal strength indicator value RSSI finds range curve: calculate the signal strength loss in Gauss Distribution Fitting during Signal transmissions 1m in current environment and RSSI Gauss described in signal intensity attenuation index n thus matching finds range mathematic(al) representation;

S2, to divide based on fan-shaped overlapping region: the feature according to received signal strength indicator value RSSI distribution with regional differentiation, using square area as application scenarios, a reference node is respectively placed at square four summits A, B, C, D, be respectively BS1, BS2, BS3, BS4, region in square, using described foursquare length of side R as fan-shaped covering radius, is divided into blind area, non-concentration zones and concentration zones according to by the number of times of fan-shaped covering by described reference node;

S3, cluster analysis: the barycenter to choose in blind area, non-concentration zones and concentration zones bunch, the RSSI data set unknown mobile node MN screened through Gauss according to Euclidean distance criterion calculates with the barycenter RSSI data set in three regions bunch respectively, calculates minimum value person and then can judge that unknown mobile node MN is at this area coverage;

S4, distance correction: before selection reference node location, select reference node far away of adjusting the distance with the reference node of unknown node MN close together and carry out measuring distance correction;

S5, dynamic weighting center coordination: dynamic weighting centroid localization algorithm has regional center characteristic in conjunction with blind area, non-concentration zones, concentration zones RSSI data set distributional difference characteristic and weighted mass center location WCL, add dynamic factor and improve weighted mass center location WCL positioning precision, thus unknown node MN position is solved.

2. the location algorithm based on the cluster analysis of fan-shaped overlapping region according to claim 1, is characterized in that: described step S1 specifically comprises the following steps:

S11, Maximum-likelihood estimation is carried out to the k group data of each range points d of current environment sample, estimate Gaussian Profile P respectively _raverage with variance be shown below:

S12, utilize normpdf to solve each range points d probability to be respectively α ₁p _r1(d) and α ₂p _r2(d), solution procedure is as follows:

P[P _r(d)≤P _r1(d)]＝F[P _r1(d)]＝α ₁，

P[P _r(d)≥P _r2(d)]＝1-F[P _r2(d)]＝1-α ₂；

S13, to filter out each range points probability be [α ₁, α ₂] interval P _rvalue, obtains at P _r1(d) and P _r2p between (d) _rvalue, by the P of screening _rvalue is defined as high-probability event, screening the P obtained through Gaussian Profile _rbe stored in Beacon_val_gauss [m], each range points RSSI average is tried to achieve by following formula

Wherein m is through Gaussian Profile screening P _rnumber;

S14, by what try to achieve substituting into following formula utilizes least square method to carry out curve fitting, and draws fit mathematics expression formula

3. the location algorithm based on the cluster analysis of fan-shaped overlapping region according to claim 2, is characterized in that: described step S3 specifically comprises the following steps:

S31, choose the barycenter that barycenter in blind area, non-concentration zones and region, 3, concentration zones does each region bunch;

S32, to calculate in described 3 regions barycenter to the distance of four reference nodes BS1, BS2, BS3, BS4, the RSSI Gauss substituted in described step S14 models fitting curve of finding range can obtain the RSSI data set of the barycenter respective base station in each region bunch, the barycenter RSSI data set in each region bunch is sorted from big to small, obtains the barycenter RSSI data set RSSI in each region bunch _x;

S33, unknown node MN receive the RSSI data set of four reference nodes BS1, BS2, BS3, BS4, also sorted from big to small, obtain the RSSI data set RSSI of unknown node MN by Gaussian data process _mN;

S34, calculate the barycenter distinctiveness ratio in unknown node MN and each region bunch, selection similarity function Euclidean distance function is as the judgement of similarity degree, and Euclidean distance formula is shown below, wherein RSSI _xfor the barycenter RSSI data set in each region bunch, the number of p contained by RSSI data centralization

$D({\mathrm{RSSI}}_{MN},{\mathrm{RSSI}}_x)=\left|\right|{\mathrm{RSSI}}_{MN}-{\mathrm{RSSI}}_x|{|}_2={\left({\mathrm{\Σ}}_{j=1}^p{|{\mathrm{RSSI}}_{MN}\left(j\right)-{\mathrm{RSSI}}_x\left(j\right)|}^2\right)}^{\frac{1}{2}};$

S35, compare the data set RSSI of unknown node MN _mNwith the barycenter RSSI data set D (RSSI in each region bunch _mN, RSSI _x) difference, difference reckling, can judge that unknown node MN belongs to RSSI data set RSSI _xpoint in this region.

4. the location algorithm based on the cluster analysis of fan-shaped overlapping region according to claim 2, is characterized in that: described step S4 specifically comprises the following steps:

S41, record AB respectively, the RSSI value RSSI between BC, CD, DA _aB, RSSI _bC, RSSI _cD, RSSI _dA;

S42, data set RSSI at unknown node MN _mNin, try to achieve matching RSSI Gauss curve of finding range according to step S14 and obtain unknown node MN and four the distance d that reference node BS1, BS2, BS3, BS4 are nearest _min;

S43, selection RSSI range finding calibration model, the impact that erasure signal strength retrogression index n brings: try to achieve unknown node MN data set RSSI according to following formula _mNclosest approach is to the reference distance d of reference node _{refer_x},

$d_{refer\_x}={10}^{\frac{P_r\left(d_1\right)-{\mathrm{RSSI}}_{MN\_x}}{P_r\left(d_1\right)-{\mathrm{RSSI}}_{xy}}},$

Wherein, RSSI _xyfor recording AB, the RSSI reference measurement values between BC, CD, DA, RSSI _{mN_x}with d _{refer_x}in x be the distance correction number of times carried out according to territory, sector coverage area, each region number, being 1 at blind area distance correction number of times, is 2 at distance correction number of times, does not need to carry out distance correction in concentration zones;

S44, ask minimum distance d _minwith reference distance d _{refer_x}between Error Absolute Value d _{error_x};

S45, foundation data set RSSI _mNclosest approach to distance correction coefficient gamma corresponding to reference node, according to formula

d _{max_x}＝γ[P _r(d ₁)-10nlog ₁₀(d)]

Ask data set RSSI _mNcomparatively far point distance d _{max_x}, wherein γ is distance correction coefficient, and the computing formula of described distance correction coefficient is

γ＝(1-β) ^τ，

τ is distance correction index, and β is distance correction percentage, and described distance correction percentage refers to Error Absolute Value d _{error_x}account for minimum distance d _minpercentage, computing formula is

$\mathrm{\β}=\frac{d_{error\_x}}{d_{\min }}.$

5. the location algorithm based on the cluster analysis of fan-shaped overlapping region according to claim 1, it is characterized in that: in described step S2, described concentration zones take radius as the region of the fan-shaped covering four times of R, non-concentration zones take radius as the fan-shaped covering three times of R and removes the region of concentration zones, and blind area take radius as the fan-shaped covering twice of R and removes the region of concentration zones and non-concentration zones.

6. the location algorithm based on the cluster analysis of fan-shaped overlapping region according to claim 1, it is characterized in that: in described step S5, dynamic weighting centroid localization algorithm defines blind area, non-concentration zones and concentration zones dynamic factor and use weighted mass center to locate displace analysis that WCL carries out unknown node MN respectively, and formula used is

$MN(x,y)=\frac{{\mathrm{\Σ}}_{j=1}^z\left(\frac{1}{d^{{\mathrm{\ω}}_j}}{\mathrm{BS}}_j\right(x,y\left)\right)}{{\mathrm{\Σ}}_{j=1}^z{w}_j},$

Wherein, ω _jfor the dynamic factor of a corresponding jth reference node, MN (x, y) is unknown node MN coordinate, BS _j(x, y) is the coordinate of a jth reference node, and z is reference node number; The dynamic factor of described reference node is divided into three kinds of situations:

Blind area dynamic factor is chosen:

RSSI _MN(1)，RSSI _MN(2)→ω ₁＝ω ₂＝1，

RSSI _MN(3)，RSSI _MN(4)→ω ₃＝ω ₄＝2；

Non-concentration zones dynamic factor is chosen:

RSSI _MN(1)，RSSI _MN(2)→ω ₁＝ω ₂＝1，

RSSI _MN(3)→ω ₃＝1，

RSSI _MN(4)→ω ₄＝2；

Concentration zones dynamic factor is chosen:

RSSI _MN(1)，RSSI _MN(2)→ω ₁＝ω ₂＝1，

RSSI _MN(3)，RSSI _MN(4)→ω ₃＝ω ₄＝1。