CN104507159A

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

Info

Publication number: CN104507159A
Application number: CN201410682677.8A
Authority: CN
Inventors: 杨东凯; 修春娣; 杨萌; 刘源; 罗智勇
Original assignee: Beihang University
Current assignee: Beihang University
Priority date: 2014-11-24
Filing date: 2014-11-24
Publication date: 2015-04-08

Abstract

A hybrid indoor positioning method based on WiFi received signal strength, which has five major steps: Step 1: According to the location fingerprint indoor positioning technology, in the offline stage, according to the selected indoor test area, the RSSI value is generated according to the logarithmic distance path loss model , forming an offline fingerprint database; Step 2: produce the motion path of the target under the two-dimensional coordinates by dead reckoning; Step 3: calculate the positioning result by fuzzy k-NN; Step 4: analyze the motion according to the state model of the Kalman filter method Predict the next position of the target; Step 5: Add the estimated position coordinates obtained by the position fingerprint technology into the measurement matrix of the Kalman filter method to obtain a mixed measurement matrix, and update the system state through the mixed measurement model. Obtain the filtered estimated position coordinates; through the continuous iterative update of steps 4 and 5, the positioning results of the entire motion path are obtained.

Description

A Hybrid Indoor Positioning Method Based on WiFi Received Signal Strength

技术领域technical field

本发明提供一种基于WiFi接收信号强度的混合室内定位方法，具体说是一种将三边定位技术和位置指纹定位技术相融合的混合卡尔曼滤波方法。该方法能够解决三边测量技术对噪声敏感和LF技术不能适应环境变化的问题，提高系统的定位精度和鲁棒性，属于WiFi室内定位及无线传输与导航领域。The present invention provides a hybrid indoor positioning method based on WiFi received signal strength, specifically a hybrid Kalman filter method combining trilateration positioning technology and position fingerprint positioning technology. The method can solve the problems that trilateral measurement technology is sensitive to noise and LF technology cannot adapt to environmental changes, and improves the positioning accuracy and robustness of the system, belonging to the fields of WiFi indoor positioning, wireless transmission and navigation.

背景技术Background technique

随着现代定位和导航技术的发展，各种基于位置的服务日益成为智能生活中重要的组成部分，全球卫星导航系统(GNSS)为人们提供了高精度、全天候的定位服务，但是由于其测量信号不能穿透建筑物的特点，在高密集建筑群区和室内无法有效进行定位服务，因此为了在室内获得有效的定位服务，室内定位系统得到了很快的发展。With the development of modern positioning and navigation technology, various location-based services have increasingly become an important part of smart life. Global Navigation Satellite System (GNSS) provides people with high-precision, all-weather positioning services. However, due to its measurement signal Due to the characteristics of not being able to penetrate buildings, positioning services cannot be effectively provided in high-density building complex areas and indoors. Therefore, in order to obtain effective positioning services indoors, indoor positioning systems have been developed rapidly.

近年来，由于WiFi网络的普及，基于IEEE802.11b/g协议的无线局域网(WLAN)的信号强度定位技术日益受到重视，目前主流的PDA、智能手机等移动设备中都内置无线网卡，从设备上为该定位技术提供了便利。现今很多建筑中装有WiFi接入点，现有的基础设施可以基本上满足定位需求，可以减少经济开支和额外的硬件装配，应用前景广阔。In recent years, due to the popularization of WiFi networks, the signal strength positioning technology of wireless local area network (WLAN) based on IEEE802.11b/g protocol has been paid more and more attention. At present, mainstream PDAs, smart phones and other mobile devices have built-in wireless network cards. Provides convenience for this positioning technology. Nowadays, many buildings are equipped with WiFi access points. The existing infrastructure can basically meet the positioning requirements, which can reduce economic expenditure and additional hardware assembly, and has broad application prospects.

基于WiFi的室内定位技术主要有三边定位技术和位置指纹技术。三边定位技术利用待测目标到至少三个已知参考点之间的距离信息估计目标位置。WiFi信号随着传播距离的增加而减弱，通过测得某一信号强度，可以计算出测量点距离接入点(AP)的距离，计算得到若干个距离就可确定位置。基于传播模型的定位方法不需要预先采集AP的信号强度，只需找出射频信号在室内环境中的传播模型，依据信号传播模型和设备与AP之间的信号强度差来估计位置信息，因为接收信号强度(RSS)很容易受周围环境影响，很难获得精确的定位结果。位置指纹定位技术需要构建信号强度与定位位置之间的映射关系，使用存储所有参考点(RP)的RSS信息的数据库来进行匹配计算，但是数据库对环境的依赖性较大，一旦环境发生较大的改变，会导致指纹数据库失效。WiFi-based indoor positioning technology mainly includes three-sided positioning technology and location fingerprint technology. Trilateration technology uses the distance information between the target to be measured and at least three known reference points to estimate the position of the target. The WiFi signal weakens with the increase of the propagation distance. By measuring a certain signal strength, the distance between the measurement point and the access point (AP) can be calculated, and the position can be determined after calculating several distances. The positioning method based on the propagation model does not need to collect the signal strength of the AP in advance, but only needs to find out the propagation model of the RF signal in the indoor environment, and estimate the location information based on the signal propagation model and the signal strength difference between the device and the AP. The signal strength (RSS) is easily affected by the surrounding environment, and it is difficult to obtain accurate positioning results. The position fingerprint positioning technology needs to construct the mapping relationship between the signal strength and the positioning position, and use the database that stores the RSS information of all reference points (RP) to perform matching calculations, but the database is highly dependent on the environment. Changes will cause the fingerprint database to become invalid.

卡尔曼滤波在测量方差已知的情况下能够从一系列存在测量噪声的数据中，估计动态系统的状态。卡尔曼滤波的一个典型实例是从一组有限的，对物体位置的，包含噪声的观察序列中预测出物体的坐标位置及速度。Kalman filtering can estimate the state of a dynamic system from a series of data with measurement noise when the measurement variance is known. A typical example of Kalman filtering is to predict the coordinate position and velocity of an object from a limited set of observation sequences containing noise about the object's position.

发明内容Contents of the invention

本发明的目的在于：提供一种基于WiFi接收信号强度的混合室内定位方法，它是一种混合卡尔曼滤波方法，将基于RSSI的WiFi三边测量技术和基于卡尔曼滤波的LF技术的优点进行融合，以解决三边测量技术对噪声敏感和LF技术不能适应环境变化的问题，提高系统的定位精度和鲁棒性。The object of the present invention is to provide a hybrid indoor positioning method based on WiFi received signal strength, which is a hybrid Kalman filter method, which combines the advantages of RSSI-based WiFi trilateration technology and Kalman filter-based LF technology. Fusion to solve the problem that trilateration technology is sensitive to noise and LF technology cannot adapt to environmental changes, and improve the positioning accuracy and robustness of the system.

本发明的技术方案：Technical scheme of the present invention:

本发明一种基于WiFi接收信号强度的混合室内定位方法，它包括以下几个步骤：The present invention is a hybrid indoor positioning method based on WiFi received signal strength, which comprises the following steps:

步骤一：根据位置指纹室内定位技术，在离线阶段，根据选定的室内测试区域，依照对数距离路径损耗模型产生RSSI值，构成离线指纹数据库。Step 1: According to the location fingerprint indoor positioning technology, in the offline stage, according to the selected indoor test area, the RSSI value is generated according to the logarithmic distance path loss model to form an offline fingerprint database.

其中，由于信号的长距离衰落特性服从对数正态分布，常用对数距离路径损耗模型表示，其中对数距离路径损耗模型的具体表示如下。Among them, since the long-distance fading characteristics of the signal obey the logarithmic normal distribution, the logarithmic distance path loss model is often used to express it, and the specific expression of the logarithmic distance path loss model is as follows.

其中，P_rx表示接收强度(分贝)，是相对距离的接收强度或一米远的初始RSSI值(分贝)，是路径损耗指数，它能随传播环境的不同在2～6之间变化。R_o参考距离，R是接收设备与发射设备之间的距离。wallLoss是指由各面墙造成的损耗的和。这个因素决定于建筑布局，建筑材料，大量反射面，基础公共设施和移动物体。Among them, P _rx represents the receiving strength (decibel), is the received strength relative to the distance or the initial RSSI value (in decibels) one meter away, is the path loss index, which can vary between 2 and 6 depending on the propagation environment. R _o Reference distance, R is the distance between the receiving device and the transmitting device. wallLoss is the sum of the losses caused by the individual walls. This factor depends on the building layout, building materials, large number of reflective surfaces, basic public facilities and moving objects.

步骤二：通过航位推测法在二维坐标下产生目标的运动路径。Step 2: Generate the movement path of the target in the two-dimensional coordinates by the dead reckoning method.

x_i＝x_i-1+v_iΔtcos(θ_i)x _i ＝xi _-1 +v _i Δtcos(θ _i )

y_i＝y_i-1+v_iΔtsin(θ_i)y _i =y _i-1 +v _i Δtsin(θ _i )

其中，x_i，y_i表示当前物体的位置，x_i-1，y_i-1是笛卡尔坐标系统中前一个参考位置，v_i和Δt分别表示速度和取样间隔。Among them, x _i , y _i represent the current position of the object, x _i-1 , y _i-1 are the previous reference position in the Cartesian coordinate system, v _i and Δt represent the velocity and sampling interval, respectively.

步骤三：通过模糊k-NN算出定位结果。Step 3: Calculate the positioning result through fuzzy k-NN.

其中，离线阶段从各AP采集的接收信号强度(RSSI)向量为在线阶段，在特定位置采集到的RSSI值为：其中，N为AP的个数，n为测试区域栅格的个数。模糊k-NN分类器采用一阶最近邻算法，则在位置γ_i的值μ_i(γ)可以表示为：Among them, the received signal strength (RSSI) vector collected from each AP in the offline phase is In the online phase, the RSSI value collected at a specific location is: Among them, N is the number of APs, and n is the number of grids in the test area. The fuzzy k-NN classifier adopts the first-order nearest neighbor algorithm, then the value μ _i (γ) at position γ _i can be expressed as:

${μ μ}_{i i} ((γ γ)) = = \frac{11 / / {| | | | {γ γ}_{i i} - - ξ ξ | | | |}^{22 / / ((m m - - 11))}}{{Σ Σ}_{j j = = 11}^{K K} ((11 / / {| | | | {γ γ}_{i i} - - ξ ξ | | | |}^{22 / / ((m m - - 11))}))}$

其中，其中K是最近邻的数量。模糊强度参数m用来决定计算各个邻居对函数值的贡献时距离的权重由多大，m的取值范围为(1,+∞)。where K is the number of nearest neighbors. The fuzzy intensity parameter m is used to determine the weight of the distance when calculating the contribution of each neighbor to the function value. The value range of m is (1,+∞).

步骤四:根据卡尔曼滤波方法的状态模型对运动目标的下一个位置进行预测。Step 4: Predict the next position of the moving target according to the state model of the Kalman filter method.

其中，系统的状态方程是基于航位推测模型的。其中当前位置可表示为x＝[p^x p^y v^x v^y]^T，则系统的状态方程的表达式如下。Among them, the state equation of the system is based on the dead reckoning model. Where the current position can be expressed as x=[p ^x p ^y v ^x v ^y ] ^T , then the expression of the state equation of the system is as follows.

${A A}_{k k} = = [\begin{matrix} 11 & 00 & {T T}_{k k} & 00 \\ 00 & 11 & 00 & {T T}_{k k} \\ 00 & 00 & 11 & 00 \\ 00 & 00 & 00 & 11 \end{matrix}]$

由于对数路径损耗模型是非线性的，所以采用如下非线性模型进行预测：Since the logarithmic path loss model is nonlinear, the following nonlinear model is used for prediction:

x_k+1＝f_k(x_k,k)+w_k x _k+1 = f _k (x _k ,k)+w _k

y_k＝h_k(x_k)+v_k y _k =h _k (x _k )+v _k

其中，w_k为系统噪声，v_k为量测噪声。系统噪声和量测噪声均为高斯白噪声。系统噪声的协方差矩阵为E[w_k·w_k ^T]＝Q_k，量测噪声的协方差矩阵为E[v_k·v_k ^T]＝R_k，系统参数的线性一阶近似定义为雅可比矩阵 $Φ^{[1]} (\hat{x}, k) \approx \frac{&PartialD; f (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}},$ $H^{[1]} (\hat{x}, k) \approx \frac{&PartialD; h (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}} .$ 则系统的预测过程如下。Among them, w _k is the system noise, and v _k is the measurement noise. Both system noise and measurement noise are Gaussian white noise. The covariance matrix of system noise is E[w _k ·w _k ^T ]＝Q _k , the covariance matrix of measurement noise is E[v _k ·v _k ^T ]＝R _k , and the linear first-order approximation of system parameters is defined as Jacobian matrix $Φ^{[1]} (\hat{x}, k) \approx \frac{&PartialD; f (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}},$ $h^{[1]} (\hat{x}, k) \approx \frac{&PartialD; h (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}} .$ Then the prediction process of the system is as follows.

${\overset{^^}{x x}}_{k k + + 11}^{- -} = = {f f}_{k k} (({\overset{^^}{x x}}_{k k},, k k))$

${P P}_{k k + + 11}^{- -} = = {Φ Φ}_{k k} {P P}_{k k + + 11}^{+ +} {Φ Φ}_{k k}^{T T} + + {Q Q}_{k k}$

步骤五：将通过位置指纹技术得到的估计位置坐标加入到卡尔曼滤波方法的量测矩阵中，得到混合量测矩阵，通过混合量测模型对系统状态进行更新，得到滤波后的估计位置坐标。通过步骤四与步骤五不断的迭代更新，可得到整个运动路径下的定位结果。Step 5: Add the estimated position coordinates obtained by the position fingerprint technology into the measurement matrix of the Kalman filter method to obtain a mixed measurement matrix, update the system state through the mixed measurement model, and obtain the estimated position coordinates after filtering. Through the continuous iterative update of steps 4 and 5, the positioning results of the entire motion path can be obtained.

其中，步骤三中通过位置指纹方法得到的定位结果的坐标为则混合量测矩阵的表达式如下所示。Among them, the coordinates of the positioning result obtained by the location fingerprint method in step 3 are Then the expression of the mixed measurement matrix is as follows.

通过卡尔曼滤波的量测模型进行更新，更新过程如下：The measurement model is updated through the Kalman filter, and the update process is as follows:

${K K}_{k k} = = {P P}_{k k}^{- -} {H h}_{k k}^{T T} {(({H h}_{k k} {P P}_{k k}^{- -} {H h}_{k k}^{T T} + + {R R}_{k k}))}^{- - 11}$

${I I}_{k k + + 11} = = {y the y}_{k k} + + {H h}_{k k} {\overset{^^}{x x}}_{k k + + 11}^{- -}$

${\overset{^^}{x x}}_{k k + + 11} = = {\overset{^^}{x x}}_{k k + + 11}^{- -} + + {K K}_{k k + + 11} {I I}_{k k + + 11}$

${P P}_{k k + + 11} = = {P P}_{k k + + 11}^{- -} - - {K K}_{k k + + 11} {H h}_{k k + + 11} {P P}_{k k + + 11}^{- -}$

其中，K_k为卡尔曼增益，P_k为误差协方差矩阵，I_k+1为真实值与量测值之间的误差。通过不断的迭代更新可以得到最终的定位结果。Among them, K _k is the Kalman gain, P _k is the error covariance matrix, and I _k+1 is the error between the real value and the measured value. The final positioning result can be obtained through continuous iterative updating.

本发明的优点在于：The advantages of the present invention are:

一、将基于RSSI的WiFi三边测量技术和基于卡尔曼滤波的LF技术的优点进行融合，提高了定位精度；1. Integrate the advantages of RSSI-based WiFi trilateration technology and Kalman filter-based LF technology to improve positioning accuracy;

二、在目标方位突然变化和环境噪声的情况下，提高了系统的鲁棒性；2. In the case of sudden changes in target orientation and environmental noise, the robustness of the system is improved;

三、估计位置连续。Third, the estimated position is continuous.

附图说明Description of drawings

图1室内定位场景图。Figure 1 Indoor positioning scene diagram.

图2本发明所述方法流程图。Fig. 2 is a flow chart of the method of the present invention.

图3基于本发明的定位结果图。Fig. 3 is a diagram of positioning results based on the present invention.

图4基于本发明的定位误差图。Figure 4 is a positioning error map based on the present invention.

图中符号、代号说明如下：The symbols and codes in the figure are explained as follows:

AP Access Point无线接入点AP Access Point wireless access point

WIFI Wireless Fidelity无线保真技术WIFI Wireless Fidelity wireless fidelity technology

具体实施方式Detailed ways

参见图1，为典型的室内定位场景，整个区域为50m*50m，WiFi的接入点(AP)安装在四个角落，四个AP的坐标分别为(0,0)，(0,50)，(50,50)，(50,0)。See Figure 1, which is a typical indoor positioning scene. The entire area is 50m*50m. WiFi access points (APs) are installed in four corners. The coordinates of the four APs are (0,0), (0,50) , (50,50), (50,0).

参见图2，是本发明所述方法流程图。本发明一种基于WiFi接收信号强度的混合室内定位方法，其步骤如下：Referring to Fig. 2, it is a flow chart of the method of the present invention. The present invention is a hybrid indoor positioning method based on WiFi received signal strength, the steps of which are as follows:

步骤1：参见图1的室内定位区域，根据位置指纹室内定位技术，在离线阶段，根据选定的室内测试区域，将整个区域划分为2m*2m的格子，依照对数距离路径损耗模型产生RSSI值，构成离线指纹数据库。Step 1: Refer to the indoor positioning area in Figure 1, according to the location fingerprint indoor positioning technology, in the offline stage, according to the selected indoor test area, divide the entire area into 2m*2m grids, and generate RSSI according to the logarithmic distance path loss model value to form an offline fingerprint database.

其中，对数路径损耗模型的计算公式为：Among them, the calculation formula of the logarithmic path loss model is:

其中，相对距离的接收信号强度参考距离R_o＝1m。路径损耗指数由各面墙造成的损耗的和服从标准正态分布wallLoss～N(0,σ²),σ²＝4.35。设第k个参考点的位置坐标为(x_k,y_k)，则参考点与四个AP之间的距离矩阵R可以表示为:Among them, the received signal strength relative to the distance The reference distance R _o =1 m. Path Loss Index The sum of losses caused by each wall obeys the standard normal distribution wallLoss~N(0,σ ² ), σ ² =4.35. Assuming that the position coordinates of the kth reference point are (x _k , y _k ), the distance matrix R between the reference point and the four APs can be expressed as:

${[\begin{matrix} {R R}_{11} \\ {R R}_{22} \\ {R R}_{33} \\ {R R}_{44} \end{matrix}]}_{k k} = = [\begin{matrix} \sqrt{{(({x x}_{k k} - - {p p}_{A A {P P}_{s the s # # 11}}^{x x}))}^{22} + + {(({y the y}_{k k} - - {p p}_{A A {P P}_{s the s # # 11}}^{y the y}))}^{22}} \\ \sqrt{{(({x x}_{k k} - - {p p}_{A A {P P}_{s the s # # 22}}^{x x}))}^{22} + + {(({y the y}_{k k} - - {p p}_{A A {P P}_{s the s # # 22}}^{y the y}))}^{22}} \\ \sqrt{{(({x x}_{k k} - - {p p}_{A A {P P}_{# # 33}}^{x x}))}^{22} + + {(({y the y}_{k k} - - {p p}_{A A {P P}_{s the s # # 33}}^{y the y}))}^{22}} \\ \sqrt{{(({x x}_{k k} - - {p p}_{A A {P P}_{s the s # # 44}}^{x x}))}^{22} + + {(({y the y}_{k k} - - {p p}_{A A {P P}_{s the s # # 44}}^{y the y}))}^{22}} \end{matrix}]$

步骤2：通过航位推测法产生四边形运动路径。Step 2: Generate a quadrilateral motion path by dead reckoning.

x_i＝x_i-1+v_iΔtcos(θ_i)x _i ＝xi _-1 +v _i Δtcos(θ _i )

y_i＝y_i-1+v_iΔtsin(θ_i)y _i =y _i-1 +v _i Δtsin(θ _i )

目标移动速度v＝0.75m/s，采样间隔Δt＝1s。Target moving speed v=0.75m/s, sampling interval Δt=1s.

步骤3：通过模糊k-NN算出定位结果。Step 3: Calculate the positioning result through fuzzy k-NN.

在位置γ_i的值μ_i(γ)可以表示式为：The value μ _i (γ) at position γ _i can be expressed as:

其中，最近邻的数量K取值为5。模糊强度参数m的取值为2。可以求得估计位置坐标为 Wherein, the number K of the nearest neighbors is 5. The value of the blur intensity parameter m is 2. The estimated position coordinates can be obtained as

步骤4：根据卡尔曼滤波方法的状态模型对运动目标的下一个位置进行预测。Step 4: Predict the next position of the moving target according to the state model of the Kalman filter method.

其中当前位置可表示为x＝[p^x p^y v^x v^y]^T，由于对数路径损耗模型是非线性的，所以采用如下非线性模型进行预测：The current position can be expressed as x=[p ^x p ^y v ^x v ^y ] ^T . Since the logarithmic path loss model is nonlinear, the following nonlinear model is used for prediction:

x_k+1＝f_k(x_k,k)+w_k x _k+1 = f _k (x _k ,k)+w _k

y_k＝h_k(x_k)+v_k y _k =h _k (x _k )+v _k

其中，w_k为系统噪声，v_k为量测噪声。系统参数的线性一阶近似定义为雅可比矩阵 $Φ^{[1]} (\hat{x}, k) \approx \frac{&PartialD; f (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}},$ $H^{[1]} (\hat{x}, k) \approx \frac{&PartialD; h (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}} .$ 则系统的预测过程如下。Among them, w _k is the system noise, and v _k is the measurement noise. A linear first-order approximation to the system parameters is defined as the Jacobian matrix $Φ^{[1]} (\hat{x}, k) \approx \frac{&PartialD; f (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}},$ $h^{[1]} (\hat{x}, k) \approx \frac{&PartialD; h (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}} .$ Then the prediction process of the system is as follows.

其中，系统噪声的协方差矩阵Q_k＝diag{0.1,0.1,1.0,1.0}×10^1.650，量测噪声的协方差矩阵R_k＝I_6×6×10^3.650,R(1,1)＝50,R(2,2)＝50Among them, the covariance matrix of system noise Q _k =diag{0.1,0.1,1.0,1.0}×10 ^1.650 , the covariance matrix of measurement noise R _k ＝I _6×6 ×10 ^3.650 , R(1,1)＝ 50,R(2,2)=50

步骤5:将通过位置指纹技术得到的估计位置坐标加入到卡尔曼滤波方法的量测矩阵中，得到混合量测矩阵，通过混合量测模型对系统状态进行更新，得到滤波后的估计位置坐标。通过步骤四与步骤五不断的迭代更新，可得到整个运动路径下的定位结果。Step 5: Add the estimated position coordinates obtained by the position fingerprint technology into the measurement matrix of the Kalman filter method to obtain a mixed measurement matrix, update the system state through the mixed measurement model, and obtain the estimated position coordinates after filtering. Through the continuous iterative update of steps 4 and 5, the positioning results of the entire motion path can be obtained.

其中当前位置与4个AP之间的距离可表示为:The distance between the current location and the four APs can be expressed as:

如图3所示，依照本发明的方法在图1所示的场景中实验得到的仿真结果。X-Y轴为估计位置的坐标。正方形标线表示实际的运动路径，星形标线表示位置指纹方法的结果，三角形标线为卡尔曼方法的结果，圆形标线为混合卡尔曼方法的结果，结果表明只依靠位置指纹方法得到的结果与实际的运动路径偏差较大，混合卡尔曼方法得到的定位结果更接近于实际的运动路径，由于实际的运动路径为正方形，在方向突然转变的情况下，提出的方法仍然能够很好的适应方向变化，获得较好的定位结果。As shown in FIG. 3 , the simulation results obtained from the experiment in the scene shown in FIG. 1 according to the method of the present invention. The X-Y axis is the coordinate of the estimated position. The square markings represent the actual motion path, the star markings represent the results of the position fingerprint method, the triangle markings represent the results of the Kalman method, and the circular markings represent the results of the hybrid Kalman method. The results show that only relying on the position fingerprint method to obtain The result of the hybrid Kalman method deviates greatly from the actual motion path, and the positioning result obtained by the hybrid Kalman method is closer to the actual motion path. Since the actual motion path is a square, the proposed method can still perform well when the direction changes suddenly. Adapt to the change of direction and obtain better positioning results.

如图4所示，依照本发明的方法在图1所示的场景中实验得到的定位结果与实际值之间的均方根误差。带圆形的标线表示位置指纹方法的误差，带菱形的标线表示卡尔曼方法的误差，带正方形的标线表示混合卡尔曼方法的误差。从表1的统计可以看出混合卡尔曼方法可的定位误差最小，定位精度得到了提高。As shown in FIG. 4 , the root mean square error between the positioning result and the actual value is experimentally obtained in the scene shown in FIG. 1 according to the method of the present invention. The lines with circles represent the errors of the location fingerprint method, the lines with diamonds represent the errors of the Kalman method, and the lines with squares represent the errors of the hybrid Kalman method. From the statistics in Table 1, it can be seen that the positioning error of the hybrid Kalman method is the smallest, and the positioning accuracy has been improved.

表1 定位结果与实际位置之间的误差Table 1 The error between the positioning result and the actual position

方法method 位置指纹技术location fingerprinting 卡尔曼方法Kalman method 混合卡尔曼方法Hybrid Kalman Method 平均均方根误差(m)Average root mean square error (m) 8.08048.0804 2.06582.0658 1.63451.6345

综上所述，本发明所提供的一种基于WiFi接收信号强度的混合室内定位方法，是基于RSSI三边定位技术和位置指纹定位技术相融合的混合卡尔曼滤波方法。本发明的特点在于该室内定位方法将基于RSSI的WiFi三边测量技术和基于卡尔曼滤波的LF技术的优点进行融合，可以解决三边测量技术对噪声敏感和LF技术不能适应环境变化的问题，提高了系统的定位精度和鲁棒性。In summary, the hybrid indoor positioning method based on WiFi received signal strength provided by the present invention is a hybrid Kalman filtering method based on the fusion of RSSI trilateral positioning technology and location fingerprint positioning technology. The feature of the present invention is that the indoor positioning method integrates the advantages of the RSSI-based WiFi trilateration technology and the Kalman filter-based LF technology, which can solve the problems that the trilateration technology is sensitive to noise and the LF technology cannot adapt to environmental changes. The positioning accuracy and robustness of the system are improved.

Claims

1. A hybrid indoor positioning method based on WiFi received signal strength, characterized in that: it comprises the following steps:

Step 1: According to the location fingerprint indoor positioning technology, in the offline stage, according to the selected indoor test area, the RSSI value is generated according to the logarithmic distance path loss model to form an offline fingerprint database;

Among them, since the long-distance fading characteristics of the signal obey the lognormal distribution, the logarithmic distance path loss model is often used to express it, and the specific expression of the logarithmic distance path loss model is as follows:

{P P}_{rx r x} ((dB dB)) = = {P P}_{{d d}_{o o}} ((dBm dBm)) - - 1010 l l {log log}_{1010} ((\frac{R R}{{R R}_{o o}})) - - wallLoss wallLoss

Among them, P _rx represents the receiving strength, is the receiving strength relative to the distance or the initial RSSI value of one meter away, l is the path loss index, which can vary between 2 and 6 depending on the propagation environment; R _is the reference distance, and R is the distance between the receiving device and the transmitting device The distance; wallLoss refers to the sum of the loss caused by each wall, this factor depends on the building layout, building materials, a large number of reflective surfaces, basic public facilities and moving objects;

Step 2: Generate the movement path of the target in two-dimensional coordinates by dead reckoning;

x _i ＝xi _-1 +v _i Δtcos(θ _i )

y _i =y _i-1 +v _i Δtsin(θ _i )

Among them, x _i , y _i represent the position of the current object, x _i-1 , y _i-1 are the previous reference position in the Cartesian coordinate system, v _i and Δt represent the velocity and sampling interval respectively;

Step 3: Calculate the positioning result through fuzzy k-NN;

Among them, the received signal strength RSSI vector collected from each AP in the offline phase is In the online phase, the RSSI value collected at a specific location is: Among them, N is the number of APs, n is the number of grids in the test area; the fuzzy k-NN classifier adopts the first-order nearest neighbor algorithm, and the value μ _i (γ) at the position γ _i is expressed as:

{μ μ}_{i i} ((γ γ)) = = \frac{11 / / {| | | | {γ γ}_{i i} - - ξ ξ | | | |}^{22 / / ((m m - - 11))}}{{Σ Σ}_{j j = = 11}^{K K} ((11 / / {| | | | {γ γ}_{i i} - - ξ ξ | | | |}^{22 / / ((m m - - 11))}))}

Among them, where K is the number of nearest neighbors, the fuzzy intensity parameter m is used to determine the weight of the distance when calculating the contribution of each neighbor to the function value, and the value range of m is (1,+∞);

Step 4: predict the next position of moving target according to the state model of Kalman filtering method;

Among them, the state equation of the system is based on the dead reckoning model, where the current position is expressed as x=[p ^x p ^y v ^x v ^y ] ^T , then the expression of the state equation of the system is as follows:

{A A}_{k k} = = [\begin{matrix} 11 & 00 & {T T}_{k k} & 00 \\ 00 & 11 & 00 & {T T}_{k k} \\ 00 & 00 & 11 & 00 \\ 00 & 00 & 00 & 11 \end{matrix}]

Since the logarithmic path loss model is nonlinear, the following nonlinear model is used for prediction:

x _k+1 = f _k (x _k ,k)+w _k

y _k =h _k (x _k )+v _k

Among them, w _k is the system noise, v _k is the measurement noise; both the system noise and the measurement noise are Gaussian white noise, and the covariance matrix of the system noise is The covariance matrix of the measurement noise is A linear first-order approximation to the system parameters is defined as the Jacobian matrix

Φ^{[1]} (\hat{x}, k) \approx \frac{&PartialD; f (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}}, h^{[1]} (\hat{x}, k) \approx \frac{&PartialD; h (x, k)}{&PartialD; x} |_{x = {\hat{x}}_{k}};

Then the prediction process of the system is as follows:

{\overset{^^}{x x}}_{k k + + 11}^{- -} = = {f f}_{k k} (({\overset{^^}{x x}}_{k k},, k k))

{P P}_{k k + + 11}^{- -} = = {Φ Φ}_{k k} {P P}_{k k + + 11}^{+ +} {Φ Φ}_{k k}^{T T} + + {Q Q}_{k k}

Step 5: Add the estimated position coordinates obtained by the position fingerprint technology into the measurement matrix of the Kalman filter method to obtain a mixed measurement matrix, update the system state through the mixed measurement model, and obtain the estimated position coordinates after filtering; Through the continuous iterative update of steps 4 and 5, the positioning results of the entire motion path are obtained;

Among them, the coordinates of the positioning result obtained by the location fingerprint method in step 3 are Then the expression of the mixed measurement matrix is as follows:

{y the y}_{k k} = = [\begin{matrix} {P P}_{LF LF}^{x x} \\ {P P}_{LF LF}^{y the y} \\ y the y {P P}_{{rx r x}_{11}} \\ . . \\ . . \\ . . \\ y the y {P P}_{{rx r x}_{11}} \end{matrix}] = = [\begin{matrix} {P P}_{LF LF}^{x x} \\ {P P}_{LF LF}^{y the y} \\ {P P}_{{R R}_{o o}} - - 1010 l l {log log}_{1010} ((\frac{{R R}_{11}}{{R R}_{o o}})) \\ . . \\ . . \\ . . \\ {P P}_{{R R}_{o o}} - - 1010 l l {log log}_{1010} ((\frac{{R R}_{N N}}{{R R}_{o o}})) \end{matrix}]

The measurement model is updated through the Kalman filter, and the update process is as follows:

{K K}_{k k} = = {P P}_{k k}^{- -} {H h}_{k k}^{T T} {(({H h}_{k k} {P P}_{k k}^{- -} {H h}_{k k}^{T T} + + {R R}_{k k}))}^{- - 11}

{I I}_{k k + + 11} = = {y the y}_{k k} + + {H h}_{k k} {\overset{^^}{x x}}_{k k + + 11}^{- -}

{\overset{^^}{x x}}_{k k + + 11} = = {\overset{^^}{x x}}_{k k + + 11}^{- -} + + {K K}_{k k + + 11} {I I}_{k k + + 11}

{P P}_{k k + + 11} = = {P P}_{k k + + 11}^{- -} - - {K K}_{k k + + 11} {H h}_{k k + + 11} {P P}_{k k + + 11}^{- -}

Among them, K _k is the Kalman gain, P _k is the error covariance matrix, I _k+1 is the error between the real value and the measured value, and the final positioning result is obtained through continuous iterative updating.