CN104507159A - A method for hybrid indoor positioning based on WiFi (Wireless Fidelity) received signal strength - Google Patents

Abstract

Provided is a method for hybrid indoor positioning based on WiFi (Wireless Fidelity) received signal strength. The method comprises five steps that, in step 1, RSSI values are generated in accordance with a log-distance path loss model based on a selected indoor test area at an offline phase according to the technology of position fingerprint indoor positioning, and an offline fingerprint database is constituted; in step 2, a movement path of a target is generated at two-dimensional coordinates through dead reckoning; in step 3, a positioning result is calculated through a fuzzy k-NN; in step 4, the next position of the movement target is predicted according to a status model of a kalman filtering method; and in step 5, estimated position coordinates obtained through the technology of position fingerprint are added into a measurement matrix of the kalman filtering method and a hybrid measurement matrix is obtained, and a system state is updated through a hybrid measurement model and filtered estimated position coordinates are obtained; wherein the positioning result under the whole movement path is obtained through unceasing iteration and updating of steps 4 and 5.

Description

A kind of mixing indoor orientation method based on WiFi received signal strength

Technical field

The invention provides a kind of mixing indoor orientation method based on WiFi received signal strength, is a kind of mixing kalman filter method three limit location technologies and Indoor Position Techniques Based on Location Fingerprint merged mutually specifically.The method can solve trilateration technique and can not to conform to noise-sensitive and LF technology the problem of change, improves the positioning precision of system and robustness, belongs to WiFi indoor positioning and wireless transmission and navigation field.

Background technology

Along with modern times location and the development of airmanship, various location Based service becomes part important in Intelligent life day by day, GPS (Global Position System) (GNSS) provides high accuracy, round-the-clock positioning service for people, but because its measuring-signal can not penetrate the feature of building, effectively service cannot be positioned in highly dense groups of building district and indoor, therefore in order in the effective positioning service of indoor acquisition, indoor locating system obtains very fast development.

In recent years, due to popularizing of WiFi network, signal strength signal intensity location technology based on the WLAN (wireless local area network) (WLAN) of IEEE802.11b/g agreement comes into one's own day by day, all built-in wireless network card in the mobile device such as PDA, smart mobile phone of current main flow, from equipment for this location technology is provided convenience.In a lot of building, WiFi access point is housed now, existing infrastructure can meet location requirement substantially, can reduce financial expenditure and the assembling of extra hardware, have a extensive future.

Indoor positioning technologies based on WiFi mainly contains three limit location technologies and location fingerprint technology.Three limit location technologies utilize the range information estimating target position between target to be measured at least three known reference point.WiFi signal weakens along with the increase of propagation distance, by recording a certain signal strength signal intensity, can calculate the distance of measurement point distance access point (AP), calculates several distances and just can determine position.Localization method based on propagation model does not need the signal strength signal intensity gathering AP in advance, only need find out the propagation model of radiofrequency signal in indoor environment, basis signal propagation model and the difference in signal strength between equipment and AP carry out estimated position information, because received signal strength (RSS) is easy to by surrounding environment influence, be difficult to obtain accurate positioning result.Indoor Position Techniques Based on Location Fingerprint needs to build the mapping relations between signal strength signal intensity and position location, the database of the RSS information storing all reference points (RP) is used to carry out matching primitives, but the dependence of database to environment is larger, once larger change occurs environment, fingerprint database can be caused to lose efficacy.

Kalman filtering can exist the data of measurement noises from a series of when measuring variance and being known, estimates the state of dynamical system.A representative instance of Kalman filtering is limited from one group, to object space, comprise the coordinate position and speed that dope object in the observation sequence of noise.

Summary of the invention

The object of the invention is to: a kind of mixing indoor orientation method based on WiFi received signal strength is provided, it is a kind of mixing kalman filter method, the advantage of the WiFi trilateration technique based on RSSI and the LF technology based on Kalman filtering is merged, can not to conform to noise-sensitive and LF technology the problem of change to solve trilateration technique, improve the positioning precision of system and robustness.

Technical scheme of the present invention:

A kind of mixing indoor orientation method based on WiFi received signal strength of the present invention, it comprises following step:

Step one: according to location fingerprint indoor positioning technologies, in off-line phase, according to selected indoor test region, produces RSSI value according to log-distance path loss model model, forms off-line fingerprint database.

Wherein, due to the long distance fading characteristic obeys logarithm normal distribution of signal, common logarithm distance path loss model represents, wherein being specifically expressed as follows of log-distance path loss model model.

Wherein, P _rxrepresent receiving intensity (decibel), the receiving intensity of relative distance or the initial RSSI value (decibel) of a meter far away, be path loss index, it can change with not coexisting between 2 ~ 6 of communication environments.R _oreference distance, R is the distance between receiving equipment and transmitter.WallLoss refer to the loss caused by each face wall and.This factor is decided by architectural composition, construction material, a large amount of reflecting surface, basic common structure and mobile object.

Step 2: the motion path being produced target by dead reckoning under two-dimensional coordinate.

x _i＝x _i-1+v _iΔtcos(θ _i)

y _i＝y _i-1+v _iΔtsin(θ _i)

Wherein, x _i, y _irepresent the position of current object, x _i-1, y _i-1previous reference position in cartesian coordinate system, v _ispeed and sampling interval is represented respectively with Δ t.

Step 3: calculate positioning result by fuzzy k-NN.

Wherein, received signal strength (RSSI) vector that off-line phase gathers from each AP is on-line stage, the RSSI value collected at ad-hoc location is: wherein, N is the number of AP, and n is the number of test zone grid.Fuzzy k-NN grader adopts single order nearest neighbor algorithm, then at position γ _ivalue μ _i(γ) can be expressed as:

${\mathrm{\μ}}_i\left(\mathrm{\γ}\right)=\frac{1/{\left|\right|{\mathrm{\γ}}_i-\mathrm{\ξ}\left|\right|}^{2/(m-1)}}{\underset{j=1}{\overset{K}{\mathrm{\Σ}}}(1/{\left|\right|{\mathrm{\γ}}_i-\mathrm{\ξ}\left|\right|}^{2/(m-1)})}$

Wherein, wherein K is the quantity of arest neighbors.Vague intensity parameter m is by the weight of distance when deciding to calculate the contribution of each neighbours to functional value by much, and the span of m is (1 ,+∞).

Step 4: the next position of state model to moving target according to kalman filter method is predicted.

Wherein, the state equation of system is based on boat position presumption model.Wherein current location can be expressed as x=[p ^xp ^yv ^xv ^y] ^t, then the expression formula of the state equation of system is as follows.

$A_k=\left[\begin{array}{cccc}1& 0& T_k& 0\\ 0& 1& 0& T_k\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]$

Because log path loss model is nonlinear, so adopt following nonlinear model to predict:

x _k+1＝f _k(x _k,k)+w _k

y _k＝h _k(x _k)+v _k

Wherein, w _kfor system noise, v _kfor measurement noise.System noise and measurement noise are white Gaussian noise.The covariance matrix of system noise is E [w _kw _k ^t]=Q _k, the covariance matrix of measurement noise is E [v _kv _k ^t]=R _k, the linear first approximation of system parameters is defined as Jacobian matrix ${\mathrm{\Φ}}^{\left[1\right]}(\hat{x},k)\≈\frac{\∂f(x,k)}{\∂x}{|}_{x={\hat{x}}_k},$ $H^{\left[1\right]}(\hat{x},k)\≈\frac{\∂h(x,k)}{\∂x}{|}_{x={\hat{x}}_k}.$ Then the forecasting process of system is as follows.

${\hat{x}}_{k+1}^{-}=f_k({\hat{x}}_k,k)$

$P_{k+1}^{-}={\mathrm{\Φ}}_k{P}_{k+1}^{+}{\mathrm{\Φ}}_k^T+Q_k$

Step 5: join in the measurement matrix of kalman filter method by the estimated position coordinate obtained by location fingerprint technology, obtain hybrid measurement matrix, upgraded by hybrid measurement model to system mode, obtains filtered estimated position coordinate.Upgraded by step 4 and the continuous iteration of step 5, the positioning result under whole motion path can be obtained.

Wherein, the coordinate of the positioning result obtained by location fingerprint method in step 3 is then the expression formula of hybrid measurement matrix is as follows.

Upgraded by the measurement model of Kalman filtering, renewal process is as follows:

$K_k=P_k^{-}{H}_k^T{(H_k{P}_k^{-}{H}_k^T+R_k)}^{-1}$

$I_{k+1}=y_k+H_k{\hat{x}}_{k+1}^{-}$

${\hat{x}}_{k+1}={\hat{x}}_{k+1}^{-}+K_{k+1}{I}_{k+1}$

$P_{k+1}=P_{k+1}^{-}-K_{k+1}{H}_{k+1}{P}_{k+1}^{-}$

Wherein, K _kfor kalman gain, P _kfor error co-variance matrix, I _k+1for the error between actual value and measuring value.Upgraded by continuous iteration and can obtain final positioning result.

The invention has the advantages that:

One, the advantage of the WiFi trilateration technique based on RSSI and the LF technology based on Kalman filtering is merged, improve positioning precision;

Two, when target bearing change suddenly and ambient noise, improve the robustness of system;

Three, estimated position is continuous.

Accompanying drawing explanation

Fig. 1 indoor positioning scene graph.

Fig. 2 the method for the invention flow chart.

Fig. 3 is based on positioning result figure of the present invention.

Fig. 4 is based on position error figure of the present invention.

In figure, symbol, code name are described as follows:

AP Access Point WAP (wireless access point)

WIFI Wireless Fidelity adopting wireless fidelity technology

Embodiment

See Fig. 1, be typical indoor positioning scene, whole region is that the access point (AP) of 50m*50m, WiFi is arranged on four corners, the coordinate of four AP is respectively (0,0), (0,50), (50,50), (50,0).

See Fig. 2, it is the method for the invention flow chart.A kind of mixing indoor orientation method based on WiFi received signal strength of the present invention, its step is as follows:

Step 1: see the indoor positioning region of Fig. 1, according to location fingerprint indoor positioning technologies, in off-line phase, according to selected indoor test region, be the grid of 2m*2m by whole Region dividing, produce RSSI value according to log-distance path loss model model, form off-line fingerprint database.

Wherein, the computing formula of log path loss model is:

Wherein, the received signal strength of relative distance reference distance R _o=1m.Path loss index the loss caused by each face wall with obedience standardized normal distribution wallLoss ~ N (0, σ ²), σ ²=4.35.If the position coordinates of a kth reference point is (x _k, y _k), then the distance matrix R between reference point and four AP can be expressed as:

${\left[\begin{array}{c}{R}_1\\ R_2\\ R_3\\ R_4\end{array}\right]}_k=\left[\begin{array}{c}\sqrt{{(x_k-p_{A{P}_{s\#1}}^x)}^2+{(y_k-p_{A{P}_{s\#1}}^y)}^2}\\ \sqrt{{(x_k-p_{A{P}_{s\#2}}^x)}^2+{(y_k-p_{A{P}_{s\#2}}^y)}^2}\\ \sqrt{{(x_k-p_{A{P}_{\#3}}^x)}^2+{(y_k-p_{A{P}_{s\#3}}^y)}^2}\\ \sqrt{{(x_k-p_{A{P}_{s\#4}}^x)}^2+{(y_k-p_{A{P}_{s\#4}}^y)}^2}\end{array}\right]$

Step 2: produce quadrangle motion path by dead reckoning.

x _i＝x _i-1+v _iΔtcos(θ _i)

y _i＝y _i-1+v _iΔtsin(θ _i)

Target movement speed v=0.75m/s, sampling interval Δ t=1s.

Step 3: calculate positioning result by fuzzy k-NN.

At position γ _ivalue μ _i(γ) can expression be:

${\mathrm{\μ}}_i\left(\mathrm{\γ}\right)=\frac{1/{\left|\right|{\mathrm{\γ}}_i-\mathrm{\ξ}\left|\right|}^{2/(m-1)}}{\underset{j=1}{\overset{K}{\mathrm{\Σ}}}(1/{\left|\right|{\mathrm{\γ}}_i-\mathrm{\ξ}\left|\right|}^{2/(m-1)})}$

Wherein, the quantity K value of arest neighbors is 5.The value of vague intensity parameter m is 2.Can be in the hope of estimated position coordinate

Step 4: the next position of state model to moving target according to kalman filter method is predicted.

Wherein current location can be expressed as x=[p ^xp ^yv ^xv ^y] ^t, because log path loss model is nonlinear, so adopt following nonlinear model to predict:

x _k+1＝f _k(x _k,k)+w _k

y _k＝h _k(x _k)+v _k

Wherein, w _kfor system noise, v _kfor measurement noise.The linear first approximation of system parameters is defined as Jacobian matrix ${\mathrm{\Φ}}^{\left[1\right]}(\hat{x},k)\≈\frac{\∂f(x,k)}{\∂x}{|}_{x={\hat{x}}_k},$ $H^{\left[1\right]}(\hat{x},k)\≈\frac{\∂h(x,k)}{\∂x}{|}_{x={\hat{x}}_k}.$ Then the forecasting process of system is as follows.

${\hat{x}}_{k+1}^{-}=f_k({\hat{x}}_k,k)$

$P_{k+1}^{-}={\mathrm{\Φ}}_k{P}_{k+1}^{+}{\mathrm{\Φ}}_k^T+Q_k$

Wherein, the covariance matrix Q of system noise _k=diag{0.1,0.1,1.0,1.0} × 10 ^1.650, the covariance matrix R of measurement noise _k=I _{6 × 6}× 10 ^3.650, R (1,1)=50, R (2,2)=50

Step 5: join in the measurement matrix of kalman filter method by the estimated position coordinate obtained by location fingerprint technology, obtain hybrid measurement matrix, upgraded by hybrid measurement model to system mode, obtains filtered estimated position coordinate.Upgraded by step 4 and the continuous iteration of step 5, the positioning result under whole motion path can be obtained.

Distance wherein between current location and 4 AP can be expressed as:

${\left[\begin{array}{c}{R}_1\\ R_2\\ R_3\\ R_4\end{array}\right]}_k=\left[\begin{array}{c}\sqrt{{(x_k-p_{A{P}_{s\#1}}^x)}^2+{(y_k-p_{A{P}_{s\#1}}^y)}^2}\\ \sqrt{{(x_k-p_{A{P}_{s\#2}}^x)}^2+{(y_k-p_{A{P}_{s\#2}}^y)}^2}\\ \sqrt{{(x_k-p_{A{P}_{\#3}}^x)}^2+{(y_k-p_{A{P}_{s\#3}}^y)}^2}\\ \sqrt{{(x_k-p_{A{P}_{s\#4}}^x)}^2+{(y_k-p_{A{P}_{s\#4}}^y)}^2}\end{array}\right]$

Wherein, the coordinate of the positioning result obtained by location fingerprint method in step 3 is then the expression formula of hybrid measurement matrix is as follows.

Upgraded by the measurement model of Kalman filtering, renewal process is as follows:

$K_k=P_k^{-}{H}_k^T{(H_k{P}_k^{-}{H}_k^T+R_k)}^{-1}$

$I_{k+1}=y_k+H_k{\hat{x}}_{k+1}^{-}$

${\hat{x}}_{k+1}={\hat{x}}_{k+1}^{-}+K_{k+1}{I}_{k+1}$

$P_{k+1}=P_{k+1}^{-}-K_{k+1}{H}_{k+1}{P}_{k+1}^{-}$

Wherein, K _kfor kalman gain, P _kfor error co-variance matrix, I _k+1for the error between actual value and measuring value.Upgraded by continuous iteration and can obtain final positioning result.

As shown in Figure 3, in the scene shown in Fig. 1, the simulation result obtained is tested according to method of the present invention.X-Y axle is the coordinate of estimated position.Square graticule represents actual motion path, star graticule represents the result of location fingerprint method, triangle graticule is the result of Kalman's method, circular graticule is the result of hybrid card Germania method, result shows that the result only relying on location fingerprint method to obtain is larger with actual motion path deviation, the positioning result that hybrid card Germania method obtains is closer to the motion path of reality, motion path due to reality is square, when direction changes suddenly, the method proposed still can be good at adapting to direction change, obtains good positioning result.

As shown in Figure 4, in the scene shown in Fig. 1, the root-mean-square error between positioning result and actual value obtained is tested according to method of the present invention.The graticule of band circle represents the error of location fingerprint method, is with the graticule of rhombus to represent the error of Kalman's method, represents the error of hybrid card Germania method with foursquare graticule.From the statistics of table 1 can find out hybrid card Germania method can position error minimum, positioning precision is improved.

Error between table 1 positioning result and physical location

Method	Location fingerprint technology	Kalman's method	Hybrid card Germania method
				Average root-mean-square error (m)	8.0804	2.0658	1.6345

In sum, a kind of mixing indoor orientation method based on WiFi received signal strength provided by the present invention is the mixing kalman filter method merged mutually with Indoor Position Techniques Based on Location Fingerprint based on RSSI tri-limit location technology.Feature of the present invention is that the advantage of the WiFi trilateration technique based on RSSI and the LF technology based on Kalman filtering merges by this indoor orientation method, can solve trilateration technique to noise-sensitive and LF technology can not conform change problem, improve positioning precision and the robustness of system.

Claims (1)

1., based on a mixing indoor orientation method for WiFi received signal strength, it is characterized in that: it comprises the following steps:

Step one: according to location fingerprint indoor positioning technologies, in off-line phase, according to selected indoor test region, produces RSSI value according to log-distance path loss model model, forms off-line fingerprint database;

Wherein, due to the long distance fading characteristic obeys logarithm normal distribution of signal, common logarithm distance path loss model represents, wherein being specifically expressed as follows of log-distance path loss model model:

$P_{\mathrm{rx}}\left(\mathrm{dB}\right)=P_{d_o}\left(\mathrm{dBm}\right)-10l{\log }_{10}\left(\frac{R}{R_o}\right)-\mathrm{wallLoss}$

Wherein, P _rxrepresent receiving intensity, be the receiving intensity of relative distance or the initial RSSI value of a meter far away, l is path loss index, and it can change with not coexisting between 2 ~ 6 of communication environments; R _oreference distance, R is the distance between receiving equipment and transmitter; WallLoss refer to the loss caused by each face wall and, this factor is decided by architectural composition, construction material, a large amount of reflecting surface, basic common structure and mobile object;

Step 2: the motion path being produced target by dead reckoning under two-dimensional coordinate;

x _i＝x _i-1+v _iΔtcos(θ _i)

y _i＝y _i-1+v _iΔtsin(θ _i)

Wherein, x _i, y _irepresent the position of current object, x _i-1, y _i-1previous reference position in cartesian coordinate system, v _ispeed and sampling interval is represented respectively with Δ t;

Step 3: calculate positioning result by fuzzy k-NN;

Wherein, the received signal strength RSSI vector that off-line phase gathers from each AP is on-line stage, the RSSI value collected at ad-hoc location is: wherein, N is the number of AP, and n is the number of test zone grid; Fuzzy k-NN grader adopts single order nearest neighbor algorithm, then at position γ _ivalue μ _i(γ) be expressed as:

${\mathrm{\μ}}_i\left(\mathrm{\γ}\right)=\frac{1/{\left|\right|{\mathrm{\γ}}_i-\mathrm{\ξ}\left|\right|}^{2/(m-1)}}{\underset{j=1}{\overset{K}{\mathrm{\Σ}}}(1/{\left|\right|{\mathrm{\γ}}_i-\mathrm{\ξ}\left|\right|}^{2/(m-1)})}$

Wherein, wherein K is the quantity of arest neighbors, and vague intensity parameter m is by the weight of distance when deciding to calculate the contribution of each neighbours to functional value by much, and the span of m is (1 ,+∞);

Step 4: the next position of state model to moving target according to kalman filter method is predicted;

Wherein, the state equation of system is based on boat position presumption model, and wherein current location is expressed as x=[p ^xp ^yv ^xv ^y] ^t, then the expression formula of the state equation of system is as follows:

$A_k=\left[\begin{array}{cccc}1& 0& T_k& 0\\ 0& 1& 0& T_k\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]$

Because log path loss model is nonlinear, so adopt following nonlinear model to predict:

x _k+1＝f _k(x _k,k)+w _k

y _k＝h _k(x _k)+v _k

Wherein, w _kfor system noise, v _kfor measurement noise; System noise and measurement noise are white Gaussian noise, and the covariance matrix of system noise is the covariance matrix of measurement noise is the linear first approximation of system parameters is defined as Jacobian matrix ${\mathrm{\Φ}}^{\left[1\right]}(\hat{x},k)\≈\frac{\∂f(x,k)}{\∂x}{|}_{x={\hat{x}}_k},H^{\left[1\right]}(\hat{x},k)\≈\frac{\∂h(x,k)}{\∂x}{|}_{x={\hat{x}}_k};$ Then the forecasting process of system is as follows:

${\hat{x}}_{k+1}^{-}=f_k({\hat{x}}_k,k)$

$P_{k+1}^{-}={\mathrm{\Φ}}_k{P}_{k+1}^{+}{\mathrm{\Φ}}_k^T+Q_k$

Step 5: join in the measurement matrix of kalman filter method by the estimated position coordinate obtained by location fingerprint technology, obtain hybrid measurement matrix, upgraded by hybrid measurement model to system mode, obtains filtered estimated position coordinate; Upgraded by step 4 and the continuous iteration of step 5, obtain the positioning result under whole motion path;

Wherein, the coordinate of the positioning result obtained by location fingerprint method in step 3 is then the expression formula of hybrid measurement matrix is as follows:

$y_k=\left[\begin{array}{c}{P}_{\mathrm{LF}}^x\\ P_{\mathrm{LF}}^y\\ y{P}_{{\mathrm{rx}}_1}\\ .\\ .\\ .\\ y{P}_{{\mathrm{rx}}_1}\end{array}\right]=\left[\begin{array}{c}{P}_{\mathrm{LF}}^x\\ P_{\mathrm{LF}}^y\\ P_{R_o}-10l{\log }_{10}\left(\frac{R_1}{R_o}\right)\\ .\\ .\\ .\\ P_{R_o}-10l{\log }_{10}\left(\frac{R_N}{R_o}\right)\end{array}\right]$

Upgraded by the measurement model of Kalman filtering, renewal process is as follows:

$K_k=P_k^{-}{H}_k^T{(H_k{P}_k^{-}{H}_k^T+R_k)}^{-1}$

$I_{k+1}=y_k+H_k{\hat{x}}_{k+1}^{-}$

${\hat{x}}_{k+1}={\hat{x}}_{k+1}^{-}+K_{k+1}{I}_{k+1}$

$P_{k+1}=P_{k+1}^{-}-K_{k+1}{H}_{k+1}{P}_{k+1}^{-}$

Wherein, K _kfor kalman gain, P _kfor error co-variance matrix, I _k+1for the error between actual value and measuring value, upgraded by continuous iteration and obtain final positioning result.