CN105182288A - Indoor-positioning-system-based RSSI Kalman filtering method - Google Patents

Abstract

Disclosed in the invention is an indoor-positioning-system-based received signal strength indication (RSSI) Kalman filtering method. The method comprises the following steps: step one, constructing a bluetooth Beacon environment of an indoor scene and establishing a signal strength map (SSMap); step two, selecting an actual measurement point, obtaining RSSI data from N bluetooth Beacons and storing the data into N channels; and step three, configuring a Kalman filter to carry out filtering processing on the received signal RSSI of the N channels. With the method, a defect that RSSI fluctuation dynamic ranges of all bluetooth Beacons received by a client are large due to the indoor electromagnetic environment influence can be overcome; and with control of the dynamic ranges of RSSI of all channels, the indoor positioning precision is effectively improved.

Description

A RSSI Kalman Filter Method Based on Indoor Positioning System

technical field

The invention relates to the field of indoor precise positioning, and proposes an RSSI (Received Signal Strength Indication received signal strength indication) Kalman filtering method based on an indoor positioning system.

Background technique

Indoor positioning refers to the realization of position positioning in the indoor environment. It mainly adopts wireless communication, base station positioning, inertial navigation positioning and other technologies to form an indoor position positioning system, so as to realize the position monitoring of people and objects in the indoor space. Common indoor wireless positioning technologies include: Wi-Fi, Bluetooth, infrared, ultra-wideband, RFID (Radio Frequency Identification), ZigBee (Zigbee) and ultrasound. However, Wi-Fi signals are easily interfered by other signals, and the short transmission distance of infrared signals is easily interfered by obstacles such as walls, which limits their application in traditional mobile terminals; while ultra-bandwidth, RFID, ZigBee and ultrasonic signals are in existing It is temporarily difficult to install in mobile terminals. Therefore, based on cost performance and feasibility analysis, Beijing Lego Leroy Co., Ltd. developed the STARnet basic network system for indoor positioning and navigation based on Bluetooth signals. Just as GPS navigation needs to launch dozens of signal transmission satellites into low-earth orbit, the STARnet system also implements precise positioning and navigation services by laying comprehensive coverage of Bluetooth Beacons. After the user installs the indoor positioning software on the client terminal, he will receive the signal from the nearby Bluetooth Beacon, and analyze the RSSI of the signal to calculate the precise indoor location of the user. However, the RSSI of each Beacon received by the client is affected by the complex indoor electromagnetic environment, and there are large fluctuations, which in turn affects the client's estimation of the precise indoor position, making the entire positioning system less stable.

Contents of the invention

The purpose of the present invention is to solve the problem that the bluetooth signal received by the indoor positioning client is too large due to the interference of the indoor electromagnetic environment. The purpose of its dynamic range, thereby improving the stability of indoor positioning. The present invention first needs to be equipped with an experimental scene of indoor positioning, and divide the scene evenly into a coordinate grid; Each coordinate position of the terminal in the test scene receives signals from N Bluetooth Beacons and makes it into a SSMap (SignalStrengthMap signal strength map) database, and obtains the system state variance and the relationship between distance and signal strength according to the SSMap database information; then enter the actual measurement In the stage, the signals from N Beacons are received through the device client at any position in the room, the RSSI of these signals is divided into N channels and stored, and then the Kalman filter is designed for filtering; finally, the Kalman filtered RSSI is passed through the distance and signal strength The relationship formula is used to calculate the distance between the current device client and each Beacon, and the positioning is realized by the three-point positioning method.

A kind of RSSI Kalman filtering method based on indoor positioning system of the present invention comprises the following steps:

Step 1. Build the Bluetooth Beacon environment of the indoor scene and construct the SSMap;

Step 2. Select the actual measurement point, obtain the RSSI data from N Bluetooth Beacons, and store them in N channels;

Step 3: Design a Kalman filter to filter the RSSI of the N channels of the received signal.

The advantages of the present invention are:

The present invention overcomes the shortcoming of the large dynamic range of RSSI fluctuations received by the client from each Bluetooth Beacon caused by the influence of the indoor electromagnetic environment, and effectively improves the accuracy of indoor positioning by controlling the dynamic range of the RSSI of each channel.

Description of drawings

Fig. 1 is a system flow chart of the present invention;

Fig. 2 is the indoor scene diagram that the present invention builds;

Fig. 3 is the SSMap distribution of minor30Beacon in the whole indoor environment in the present invention;

Fig. 4 is the rendering of the first channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position before and after Kalman filter processing in the present invention;

Fig. 5 is the rendering of the second channel Bluetooth BeaconRSSI before and after Kalman filter processing at the position of grid coordinates (16,9) in the present invention;

Fig. 6 is the rendering of the third channel Bluetooth BeaconRSSI before and after Kalman filter processing at the grid coordinate (16,9) position in the present invention;

Fig. 7 is the effect diagram before and after the Kalman filter processing of the 4th channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position in the present invention;

Fig. 8 is the rendering of the fifth channel Bluetooth BeaconRSSI before and after Kalman filter processing at the grid coordinate (16,9) position in the present invention;

Fig. 9 is an effect diagram before and after the Kalman filter processing of the sixth channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position in the present invention;

Fig. 10 is the effect diagram before and after the Kalman filter processing of the 7th channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position in the present invention;

FIG. 11 is an effect diagram of the eighth channel Bluetooth BeaconRSSI at the grid coordinate (16,9) before and after Kalman filter processing in the present invention.

Detailed ways

The present invention will be further described in detail with reference to the accompanying drawings and embodiments.

The present invention is a kind of RSSI Kalman filtering method based on indoor positioning system, and its flow chart is as shown in Figure 1, comprises the following steps:

Step 1: Set up the Bluetooth Beacon environment of the indoor scene and construct the SSMap.

Specifically:

The first step is to build the indoor scene.

Assume that the scene has M floors, and each floor is evenly distributed according to the spatial distribution of W×L×H, where: W is the width of the floor, L is the length of the floor, and H is the clear height of the floor; then the horizontal plane of each floor Divide into A×B coordinate grids according to the uniform spacing of length and width; deploy a total of N Bluetooth Beacons with the same signal transmission power at equal intervals on the roof of each floor scene in a mixed manner of star network and chain network , and the broadcast signal of each Bluetooth Beacon includes the majorID recording its layer number information and the minorID recording its location information.

The second step is to build SSMap.

Equipped with a device receiving platform with a height of h ₁ , place the platform in the center of each layer of A×B coordinate grid in turn, and then divide N channels to acquire and record RSSI data for N Bluetooth Beacons respectively; at the same time, it is required to The RSSI data acquisition of the same coordinate grid requires at least t ₁ sampling time accumulation to obtain 100 data accumulations of a single channel at the coordinate position; in this way, the RSSI database of A×B×N×M×100 can be effectively Create an SSMap.

The third step is to calculate the system state variance and observation variance.

First call all the RSSI data of the SSMap under the i channel and save it as a Pow matrix, use the var function in Matlab to obtain the variance of all the RSSI data under the channel, use Q(i) to record, and then use the square of the error scale value of the observation equipment as The reference value establishes the observation variance of the system, records it with R(i), and finally traverses i from 1 to N.

Step 2: Select actual measurement points, obtain RSSI data from N Bluetooth Beacons, and store them in N channels.

Specifically:

Select the measured point at any coordinate position in the indoor scene A×B coordinate grid, and receive the RSSI data from N Bluetooth Beacons respectively through the client. It is required that the acquisition of RSSI data at the actual measurement point requires at least _t2 sampling time accumulation to obtain g RSSI data accumulations under a single channel at the actual measurement point; then calculate the maximum fluctuation of the RSSI data under each channel of the actual measurement point, The fluctuation minimum value and the fluctuation dynamic range, together with the RSSI data itself, are passed to the subsequent Kalman filter processing in the xml file format.

Step 3: Design a Kalman filter to filter the RSSI of the N channels of the received signal.

Specifically:

In the first step, the Kalman filter model is established. The design of the Kalman filter is divided into prediction and correction of the system state equation. First list the state equation of the system, starting with the general formula of the system state equation:

X(i,k+1)=AX(i,k)+W(i,k)(1)

S(i,k)=CX(i,k)+V(i,k)(2)

In the formula, X(i,k) and X(i,k+1) are system state vectors, which respectively represent the estimated RSSI values to be optimized in the i-th channel of the received signal at time k and k+1 at the measured point ; S(i,k) is the system observation vector, which represents the observed value of RSSI in the i-th channel of the received signal at the measured point at time k; and A is the system matrix, C is the system output matrix; W(i,k) and V(i,k) are the system state noise and observation noise of the i-th channel at time k, respectively, and it is approximately considered that W(i,k) and V(i,k) are mutually independent zero-mean white noise sequences, satisfying :

E[W(i,k)]=E[V(i,k)]=0(3)

E[W(i,k)W(i,k) ^T ]=Q(i)(4)

E[V(i,k)V(i,k) ^T ]=R(i)(5)

From this state equation, the prediction process of the Kalman filter equation can be listed:

P(i,k|k-1)=AP(i,k-1|k-1) ^AT +Q(i)(6)

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; A</span>}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; B</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

In the formula, X(i,k|k-1) is the result of state prediction at time k in the i-th channel using time k-1, is the optimal result of using the state at time k-1 in the i-th channel at time k, U(i,k) is the state control quantity in the i-th channel at time k, and P(i,k|k-1 )yes The corresponding covariance matrix, P(i,k-1|k-1) is The corresponding covariance matrix.

Further, the correction process of the Kalman filter equation can be listed:

K(i,k)=P(i,k|k-1)C ^T [CP(i,k|k-1)C ^T +R(i)] ^-1 (8)

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; C</span>}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$

P(i,k|k)=[I(i)-K(i,k)C]P(i,k|k-1)(10)

In the formula, K(i,k) is the Kalman gain in the i-th channel, and its function is to minimize the covariance of the posterior estimation error.

Specific to the issue of filtering and processing Bluetooth Beacon RSSI for indoor positioning, since the system design has t ₂ sampling time g RSSI data accumulation for the receiving signal processing of the same actual measurement point, the system state parameters do not occur at the same actual measurement point. changes, which means that the system matrix A and the system output matrix C are both unit matrices, so the system prediction process can be rewritten as:

P(i,k|k-1)=P(i,k-1|k-1)+Q(i)(11)

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 12</span>}<span\; class="notranslate">\; )</span>$

The system calibration process can be rewritten as:

K(i,k)=P(i,k|k-1)[P(i,k|k-1)+R(i)] ^-1 (13)

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 14</span>}<span\; class="notranslate">\; )</span>$

P(i,k|k)=[I(i)-K(i,k)]P(i,k|k-1)(15)

The second step is the program realization of the Kalman filter. The present invention uses the Karman_Filter.m program to realize the Kalman filter processing of the RSSI signal received by any given location device in the built indoor environment. First, use the xmlread function in Matlab to convert the RSSI data received by the device into the matrix data z of [N×g] for each g group of N channels received by the device; then initialize the matrix data. The meaning of the value:

The letter k represents the time node of the system state change, the letter i represents the observed channel code, and the value passed by each channel code indicates the RSSI from the Bluetooth Beacon received by the channel.

xhat is a [N×g] matrix, which represents the posterior estimated value of RSSI, that is, at time k, combined with the measured value of current RSSI and the prior estimate of RSSI at time k-1, the estimated value is updated, corresponding to the state equation X, the first element of the matrix is set to zero during initialization;

xhatminus is a [N×g] matrix, which represents the prior estimation of RSSI, that is, at time k-1, the estimation of RSSI at time k corresponds to X in the state equation, and the first element of the matrix is set to zero;

P is a [N×g] matrix, which represents the variance of the posterior estimate, and the first element of the matrix is set to zero during initialization;

Pminus is a [N×g] matrix, which represents the variance of the prior estimate, and the first element of the matrix is set to zero during initialization;

K is a [N×g] matrix, which represents the Kalman gain, and the first element of the matrix is set to zero during initialization;

Next, the present invention starts with the i-th channel, assigns xhat(i,1) in the i-th channel as the first value of the channel corresponding to the input signal, and then assigns P(i,1) in the channel to 1; So far, the matrix data initialization in the program is completed. Then in the same channel i, follow the system rewritten prediction equation and correction equation.

The system prediction equation in the code:

xhatminus(i,k)=xhat(i,k-1)(16)

Pminus(i,k)=P(i,k-1)+Q(i)(17)

The system correction equation in the code:

K(i,k)=Pminus(i,k)/(Pminus(i,k)+R(i))(18)

xhat(i,k)=xhatminus(i,k)+K(i,k)*(z(i,k)-xhatminus(i,k)(19)

P(i,k)=(1-K(i,k))*Pminus(i,k)(20)

By traversing k from 2 to g, we can get the change of the i-th channel RSSI posterior estimate with the time point, which reflects the effect of Kalman filtering; finally, by traversing i from 1 to N, we can get each channel RSSI Kalman filter effect; draw and output the effect map of each channel Bluetooth Beacon RSSI at the position of the measured point after Kalman filter processing.

The third step is to compare the change of RSSI dynamic range at the measured point before and after Kalman filtering. Call the judgeparaments.m file, where preinput configures the RSSI data of N channels before filtering (in dB), and folinput configures the RSSI data of N channels after filtering (in dB); the program outputs the maximum values of the N channels before and after Kalman filtering , the minimum value and the dynamic range of its fluctuation.

Finally, the Kalman filter-processed RSSI is solved to calculate the distance corresponding to the RSSI through the relationship between distance and signal strength. Obviously, for the Bluetooth Beacon of each channel, the upper limit and the lower limit of the distance will be calculated (corresponding to the maximum and minimum values of RSSI under the channel respectively), and then the spatial distance resection method in metrology can be used to calculate the actual measurement point in the construction scene. Mid distance position range.

Example:

The present invention is an RSSI Kalman filtering method based on an indoor positioning system, with the help of a platform carried by Beijing Lego Lehua Technology Co., Ltd., its specific implementation examples are as follows:

Step 1: Set up the Bluetooth Beacon environment of the indoor scene and construct the SSMap.

Specifically:

The first step is to build the indoor scene. As shown in Figure 2, the scene area adopts a spatial layout of 13m×7.579m×2.78m. First, the horizontal plane of the indoor scene is divided into a two-dimensional plane coordinate system of 26×16 according to the interval of 0.5m×0.5m. Set up 8 Bluetooth Beacons on the top of the scene based on , and the specific plane distribution parameters are as follows.

major minor X coordinate Y coordinate 14 29 12.60 7.30 14 30 6.50 7.30 14 31 1.00 7.30 14 32 3.35 3.70 14 33 9.67 3.70 14 34 12.70 0.30 14 35 6.50 0.30 14 36 0.30 0.30

It should be noted that these 8 Beacons have the same uuid number (namely F62D3F65-2FCB-AB76-00AB-681819202122), which is used as the hardware number for building the platform system; at the same time, major14 represents the floor of the indoor environment. Due to the test environment of this project They are all on the same level (that is, the number of floors M=1), so the major does not change; while the minor is different, which is used as a basis for the client receiver to distinguish each Beacon. In addition, each Beacon will broadcast its own unique information such as uuid, major, minor, and coordinate position while sending Bluetooth signals to the indoor space.

The second step is to build SSMap. First build a test platform with a height of 1.41m and an area of 0.5m×0.5m, place the IOS device in the center of the platform; then start from the (0,0) position of the coordinate system, and pass the frequency indication of the IOS device once per second Send 8 Bluetooth Beacons to the Bluetooth signal RSSI of the device terminal respectively, and accumulate 100s of observation time to obtain a total of 100×8 data of each Beacon at the coordinate position; traverse each coordinate in the grid according to the same method (26 ×16) RSSI data from 8 Bluetooth Beacons are made into xml format documents. Figure 3 shows the SSMap composed of the signal generated by the Bluetooth Beacon at the position of X coordinate 6.50 and Y coordinate 7.30 with uuid F62D3F65-2FCB-AB76-00AB-681819202122, major14, minor30.

The third step is to calculate the system state variance and observation variance. First call all the RSSI data of the SSMap under the i channel and save it into a Pow matrix, use the var function in Matlab to obtain the variance of all the RSSI data under the channel, and record it with Q(i); then use the square of the error scale value of the observation equipment as The reference value establishes the observation variance of the system, which is recorded by R(i); finally, i is traversed from 1 to 8.

Step 2: Select the actual measurement point, obtain the RSSI data from 8 Bluetooth Beacons, and store them in 8 channels.

Specifically:

Select the measured point with coordinates (16,9) in the indoor scene, and obtain the RSSI data from 8 Bluetooth Beacons. Also use the IOS device platform to indicate the frequency of the Bluetooth Beacon RSSI once per second. After 30 seconds of accumulation, 30 sets of RSSI data for 8 channels are obtained, and they are stored in an xml file format for post-processing. It should be noted that the design standard of the Bluetooth BeaconRSSI indicator software being developed by Lego Leroy is to indicate the Bluetooth BeaconRSSI at a frequency of 10 times per second. After 3 seconds of accumulation, 30 sets of RSSI data for 8 channels can be obtained. Currently limited by equipment, the present invention can only use the measured data under the current IOS device to simulate the design product data. However, the measured data used in the present invention to simulate the design product data only has a lag in the input time, and does not bring the data itself. change.

Step 3: Design a Kalman filter to filter the RSSI of the 8 channels of the received signal.

Specifically:

The present invention uses the Karman_Filter.m program to realize the Kalman filter processing of the RSSI signal at any given position.

The first step is data extraction before Kalman filtering. Use the xmlread function to convert the 30 groups of RSSI data received by the IOS device into [8×30] matrix data z for reading.

The second step is the initialization of matrix data. What needs to be explained is the meaning of each initial value of the program here:

The xhat matrix is the posterior estimate of RSSI, that is, at time k, the estimated value is updated by combining the measured value of current RSSI with the prior estimate of RSSI at time k-1, which corresponds to X in the state equation and is initialized to [8 ×30] zero matrix; then assign the first element of each channel to the first RSSI data of the input signal in the channel;

xhat(i,1)=z(i,1)(21)

The xhatminus matrix is the prior estimate of RSSI, that is, at time k-1, the estimate of RSSI at time k, corresponding to X in the state equation, initialized to a zero matrix of [8×30];

The P matrix is the variance of the posterior estimate, initialized to a [8×30] zero matrix; then assign the first element of each channel to 1;

P(i,1)=1(22)

The Pminus matrix is the variance of the prior estimate, initialized to a [8×30] zero matrix;

The K matrix is the Kalman gain, initialized to a [8×30] zero matrix;

The third step is to rewrite the prediction equation and correction equation. In the same channel i, rewrite the iterative equation according to the system state equation.

The system prediction equation in the code:

xhatminus(i,k)=xhat(i,k-1)(23)

Pminus(i,k)=P(i,k-1)+Q(i)(24)

The system correction equation in the code:

K(i,k)=Pminus(i,k)/(Pminus(i,k)+R(i))(25)

xhat(i,k)=xhatminus(i,k)+K(i,k)*(z(i,k)-xhatminus(i,k)(26)

P(i,k)=(1-K(i,k))*Pminus(i,k)(27)

The fourth step is to traverse k from 2 to 30. After traversal, the effect of RSSI in the same channel i after Kalman filtering is obtained.

The fifth step is to traverse i from 1 to 8. After traversal, the effect of RSSI in the 8 channels after Kalman filtering is obtained.

The sixth step is to draw and output the effect diagram of each channel Bluetooth BeaconRSSI at the position of grid coordinates (16,9) after Kalman filter processing, as shown in Figure 4-11.

Fig. 4 is the rendering of the first channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position before and after Kalman filter processing in the present invention;

Fig. 5 is the rendering of the second channel Bluetooth BeaconRSSI before and after Kalman filter processing at the position of grid coordinates (16,9) in the present invention;

Fig. 6 is the rendering of the third channel Bluetooth BeaconRSSI before and after Kalman filter processing at the grid coordinate (16,9) position in the present invention;

Fig. 7 is the effect diagram before and after the Kalman filter processing of the 4th channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position in the present invention;

Fig. 8 is the rendering of the fifth channel Bluetooth BeaconRSSI before and after Kalman filter processing at the grid coordinate (16,9) position in the present invention;

Fig. 9 is an effect diagram before and after the Kalman filter processing of the sixth channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position in the present invention;

Fig. 10 is the effect diagram before and after the Kalman filter processing of the 7th channel Bluetooth BeaconRSSI at the grid coordinate (16,9) position in the present invention;

FIG. 11 is an effect diagram of the eighth channel Bluetooth BeaconRSSI at the grid coordinate (16,9) before and after Kalman filter processing in the present invention.

The seventh step is to compare the change of RSSI dynamic range at the measured point before and after Kalman filtering. Call the judgeparaments.m file, where preinput configures 30 RSSI data (in dB) for each of the 8 channels before filtering, and folinput configures 10 RSSI data (in dB) for each of the 8 channels after filtering; the program outputs 8 before and after Kalman filtering The maximum and minimum values of each channel and the dynamic range of their fluctuations. It should be noted that in this example, it is considered that the Kalman filter tends to be stable after 20 time nodes (that is, the RSSI reading is stable after 2 seconds corresponding to the ideal device), so we make the following table to show that after the Kalman filter (that is, from the first 20 time nodes from the beginning to the end) the degree of optimization of the dynamic range of system fluctuations:

It can be seen from the above table that the dynamic range of the Bluetooth Beacon RSSI received by the device from 8 channels is significantly reduced after Kalman filtering processing, which will greatly help us achieve indoor accurate positioning through the relationship between distance and RSSI follow-up work.

Finally, the Kalman filter-processed RSSI is solved to calculate the distance corresponding to the RSSI through the relationship between distance and signal strength. Obviously, for the Bluetooth Beacon of each channel, we will calculate the upper limit and the lower limit of the distance (respectively corresponding to the maximum and minimum values of RSSI under the channel), and then use the spatial distance resection method in metrology to calculate the actual measurement point in the building. Range of distance positions in the scene.

The present invention mainly aims at the problem that the equipment terminal receives the Bluetooth Beacon RSSI fluctuation dynamic range is too large encountered in the indoor precise positioning, and designs a Kalman filter to process the RSSI, effectively controlling its dynamic range, thereby improving the positioning of the equipment terminal purpose of precision. Through example analysis, the dynamic range of the RSSI of each channel has been significantly improved after passing through the Kalman filter.

Claims (1)

1. A kind of RSSI Kalman filtering method based on indoor positioning system, comprises the following steps: Step 1. Build the Bluetooth Beacon environment of the indoor scene and construct the SSMap; Specifically: The first step is to build the indoor scene; Assume that the scene has M floors, and each floor is evenly distributed according to the spatial distribution of W×L×H, where: W is the width of the floor, L is the length of the floor, H is the clear height of the floor, and then the horizontal plane of each floor Divide into A×B coordinate grids according to the uniform spacing of length and width, and deploy a total of N Bluetooth Beacons with the same signal transmission power at equal intervals on the roof of each scene in a mixed manner of star network and chain network , and the broadcast signal of each Bluetooth Beacon contains the majorID recording its layer information and the minorID recording its location information; The second step is to build SSMap; Equipped with a device receiving platform with a height of h ₁ , place the platform in the center of each layer of A×B coordinate grid in turn, and then divide N channels to acquire and record RSSI data for N Bluetooth Beacons respectively; at the same time, it is required to The acquisition of RSSI data of the same coordinate grid requires at least t ₁ sampling time accumulation to obtain 100 data accumulations of a single channel at the coordinate position, and finally, establish SSMap through the RSSI database of A×B×N×M×100; The third step is to calculate the system state variance and observation variance; First call all the RSSI data of the SSMap under the i channel and save it into a Pow matrix, obtain the variance of all the RSSI data under the channel, record it with Q(i), and then use the square of the error scale value of the observation equipment as a reference value to establish a systematic observation Variance, recorded with R(i), and finally traverse i from 1 to N; Step 2. Select the actual measurement point, obtain the RSSI data from N Bluetooth Beacons, and store them in N channels; Specifically: Select an actual measurement point at any coordinate position in the indoor scene A×B coordinate grid, and receive RSSI data from N Bluetooth Beacons respectively through the client. The acquisition of RSSI data at the actual measurement point requires at least _t2 sampling time accumulation. Obtain g RSSI data accumulations under a single channel at the measured point; then calculate the fluctuation maximum, fluctuation minimum and fluctuation dynamic range of the RSSI data under each channel of the measured point, and pass the RSSI data to the follow-up in xml file format The processing of the Kalman filter; Step 3, designing a Kalman filter to filter the RSSI of the received signal N channels; Specifically: In the first step, the Kalman filter model is established; The state equation of the system is: X(i,k+1)=AX(i,k)+W(i,k)(1) S(i,k)=CX(i,k)+V(i,k)(2) In the formula, X(i,k) and X(i,k+1) are system state vectors, which respectively represent the estimated RSSI values to be optimized in the i-th channel of the received signal at time k and k+1 at the measured point ; S(i,k) is the system observation vector, which represents the observed value of RSSI in the i-th channel of the received signal at the measured point at time k; and A is the system matrix, C is the system output matrix; W(i,k) and V(i,k) are the system state noise and observation noise of the i-th channel at time k, respectively, and W(i,k) and V(i,k) are assumed to be mutually independent zero-mean white noise sequences, satisfying: E[W(i,k)]=E[V(i,k)]=0(3) E[W(i,k)W(i,k) ^T ]=Q(i)(4) E[V(i,k)V(i,k) ^T ]=R(i)(5) Get the prediction process of the Kalman filter equation: P(i,k|k-1)=AP(i,k-1|k-1) ^AT +Q(i)(6) $\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; A</span>}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; B</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$ In the formula, X(i,k|k-1) is the result of state prediction at time k in the i-th channel using time k-1, is the optimal result of using the state at time k-1 in the i-th channel at time k, U(i,k) is the state control quantity in the i-th channel at time k, and P(i,k|k-1 )yes The corresponding covariance matrix, P(i,k-1|k-1) is The corresponding covariance matrix; List the correction procedure for the Kalman filter equation: K(i,k)=P(i,k|k-1)C ^T [CP(i,k|k-1)C ^T +R(i)] ^-1 (8) $\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; C</span>}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$ P(i,k|k)=[I(i)-K(i,k)C]P(i,k|k-1)(10) where K(i,k) is the Kalman gain in the i-th channel; Assuming that the system matrix A and the system output matrix C are unit matrices, the system prediction process is: P(i,k|k-1)=P(i,k-1|k-1)+Q(i)(11) $\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 12</span>}<span\; class="notranslate">\; )</span>$ The system calibration process is: K(i,k)=P(i,k|k-1)[P(i,k|k-1)+R(i)] ^-1 (13) $\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 14</span>}<span\; class="notranslate">\; )</span>$ P(i,k|k)=[I(i)-K(i,k)]P(i,k|k-1)(15) The second step is the program realization of the Kalman filter; First, convert the RSSI data of each group g of N channels received by the device into [N×g] matrix data z and read it; then initialize the matrix data, set: k: represents the time node of the system state change; i: Represents the observed channel code, and the value passed by each channel code indicates the RSSI from the Bluetooth Beacon received by the channel; Xhat: It is a [N×g] matrix, which represents the posterior estimated value of RSSI, that is, at time k, combined with the measured value of current RSSI and the prior estimation of RSSI at time k-1, the estimated value is updated, corresponding to the state equation In X, the first element of the matrix is set to zero during initialization; Xhatminus: It is a [N×g] matrix, which represents the prior estimation of RSSI, that is, at time k-1, the estimation of RSSI at time k corresponds to X in the state equation, and the first element of the matrix is initialized Zero; P: It is a [N×g] matrix, which represents the variance of the posterior estimate, and the first element of the matrix is set to zero during initialization; Pminus: It is a [N×g] matrix, which represents the variance of the prior estimate, and the first element of the matrix is set to zero during initialization; K: It is a [N×g] matrix, which represents the Kalman gain, and the first element of the matrix is set to zero during initialization; For the i-th channel, assign xhat(i,1) in the i-th channel to the first value of the channel corresponding to the input signal, and then assign P(i,1) in the channel to 1, and the matrix data initialization is completed; In the same channel i, the system predicts the equation: xhatminus(i,k)=xhat(i,k-1)(16) Pminus(i,k)=P(i,k-1)+Q(i)(17) System correction equation: K(i,k)=Pminus(i,k)/(Pminus(i,k)+R(i))(18) xhat(i,k)=xhatminus(i,k)+K(i,k)*(z(i,k)-xhatminus(i,k)(19) P(i,k)=(1-K(i,k))*Pminus(i,k)(20) Traverse k from 2 to g to obtain the change of the i-th channel RSSI posterior estimate with the time point, which reflects the effect of Kalman filtering, and finally traverse i from 1 to N to obtain the RSSI Kal of each channel Mann filter effect, plotting and outputting the effect map of each channel Bluetooth BeaconRSSI at the position of the measured point after Kalman filter processing; The third step is to compare the change of RSSI dynamic range at the measured point before and after Kalman filtering; According to the RSSI data of N channels before filtering and the RSSI data of N channels after filtering, output the maximum value, minimum value and dynamic range of fluctuations of the N channels before and after Kalman filtering; Finally, the Kalman filter-processed RSSI is calculated by the relationship between distance and signal strength to calculate the distance corresponding to RSSI. For each channel of Bluetooth Beacon, the upper limit and lower limit of the distance are calculated, which correspond to the maximum value and maximum value of RSSI under the channel respectively. The minimum value is used to obtain the distance position range of the measured point in the construction scene.