RSS fingerprint database self-generated and self-updated indoor visible light anti-tilt positioning method

Abstract

This invention discloses an indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database. The steps are as follows: Assuming there is an LED light source indoors, initialize the PL index and obtain the estimated vertical RSS value of the user's first position; initialize the Kalman filter parameters, predict the three-dimensional coordinates of the i-th position and calculate the predicted distances between the i-th and j-th positions and the i-th LED light source; use Kalman filtering to obtain the estimated RSS value of the current position; estimate and update the PL index of each LED light source and the fingerprint database under the current receiving state based on the estimated vertical RSS values ​​of the current position and the previous position; if the user moves to the i-th position, continue estimation; if the user leaves the indoor positioning scene, end all estimations.

Description

Technical Field

[0001] This invention belongs to the field of indoor visible light positioning technology, specifically involving an indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database. Background Technology

[0002] The Global Positioning System (GPS) is widely used for navigation, positioning, tracking, and timing of targets in outdoor environments. However, due to the inability of GPS signals to penetrate walls and the presence of multipath effects, providing accurate positioning services for people or objects in indoor environments is quite challenging.

[0003] Visible Light Positioning (VLP) uses light-emitting diodes (LEDs) as transmitters and photodetectors as receivers. It uses visible light with a wavelength range of 380nm to 780nm as the positioning signal to estimate the position of indoor targets at the centimeter level. It can be widely used in advanced application scenarios such as position-assisted channel estimation, drone swarms and fast-moving robots, and applications requiring centimeter-level positioning accuracy and hundreds of hertz positioning frequencies.

[0004] In indoor scenes, diffraction and reflection of visible light signals are generally negligible, and the line-of-sight (LOS) component is generally considered to dominate. Trilateration methods based on received-signal-strength (RSS) estimate the distance between LEDs and PDs based on measured received light intensity. This method is low-cost and easy to implement. However, most existing RSS-based VLPs assume the receiver's normal vector is vertically upward. Under the ideal assumption of vertical reception, the distance model between LEDs and PDs can be expressed as a function of RSS, but it cannot accurately characterize the analytical relationship between the distance between LEDs and PDs and RSS when the receiver is tilted. The distances between LEDs and PDs estimated based on the vertical reception distance model and the RSS under tilted reception will inevitably deviate from the actual distances, leading to decreased positioning accuracy. Fingerprint positioning combined with artificial neural networks can complete the complex mapping from the RSS of vertical or tilted reception to the two-dimensional or three-dimensional coordinates of the receiving terminal, solving the problem of rapid deterioration in positioning accuracy under tilted reception. However, the positioning accuracy of existing RSS-based fingerprint positioning depends on the size of the fingerprint database established offline; the smaller the grid of the fingerprint database, the higher the positioning accuracy. Current fingerprint databases are all multi-parameter sets measured under the condition that the receiver's normal vector is vertically upward. To reduce the time complexity of fingerprint database construction, researchers proposed a method for manually compiling fingerprint databases based on the Path Loss (PL) exponent. This method only requires measuring a small number of RSS samples to compile a dense fingerprint database with high positioning accuracy. However, the PL exponent model used for fingerprint database compilation is only applicable to vertical reception and not to positioning under tilted receiver scenarios. To address this issue, researchers proposed a method based on the PL exponent... This paper presents a method for constructing a tilted fingerprint database. This method allows each user to correct the vertical fingerprint database based on the real-time receiving direction of the receiver. Using WKNN, centimeter-level positioning can be achieved, avoiding the time and labor costs of offline fingerprint database construction. It is a highly efficient and reliable positioning method. This method is used to generate the PL index of the fingerprint database. It is the PL index measured at multiple receiving locations. The average value, and it is assumed that this value is constant during the online phase. The PL index of a real LED light source. It is related to the half-power angle, and the relationship between the two is as follows: The half-power angle of the light source is a statistical mean, so the actual... It is a random variable, and its standard deviation changes with the light decay of the LED. The RSS measured in the online stage is the value under the current PL index. Unavoidable noise and errors during the measurement process cause the measured RSS value to differ from the offline stage value, which is based on a constant value. The generated fingerprint database will inevitably contain biases, and these biases will increase over time, leading to a continuous decrease in positioning accuracy and limiting its application in real-world indoor positioning scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide an indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database, which solves the problem of sharp deterioration in positioning accuracy when the receiver is tilted.

[0006] The technical solution adopted in this invention is an indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database, the steps of which are as follows: Step 1, assuming there is indoor... One LED light source, initialize the PL index and obtain the estimated vertical RSS value of the user's first position; Step 2, initialize Kalman filter parameters, predict the first... The three-dimensional coordinates of the position are calculated and the first position is calculated. The position, the The positions are respectively related to the first... Predicted distance between individual LED light sources; Step 3: Use Kalman filtering to obtain the estimated RSS value of the current position; Step 4: Estimate and update the PL index of each LED light source based on the estimated vertical RSS values ​​of the current position and the previous position. Step 5, if the user moves to the... For each location, repeat steps 3-4 to continue estimating. If the user leaves the indoor positioning scene, end all estimations.

[0007] The invention is further characterized by: The specific process of step 1 is as follows: Step 1.1: Initialize the power index (PL) of the light source based on its half-power angle. , , ; Step 1.2, calculate the vertical received power at the receiving position directly below each LED light source 1. ; The expression is: , (1) In equation (1), The transmission power of the LED light source, The effective detection area of ​​the photodetector. The vertical height between the LED and the receiver. For the first The initial value of the Lambertian index of each LED light source. For the first The distance between each LED light source and the receiving position directly below the LED light source; Step 1.3, divide the receiving plane into The grid is a set of grids, with the center of each grid representing the location of a fingerprint point in the fingerprint database. The 1st grid is generated using formula (2). The fingerprint point received the first Vertical receiving power of each LED light source ; (2) In equation (2), For the first The distance between each LED light source and the receiving position directly below the LED light source For the first The LED light source and the first The distance between fingerprint points , .

[0008] Step 1.4, if the polar angle and azimuth angle of the receiver at the user's initial position are respectively and According to equation (3), the tilt reception power of each fingerprint point at the receiver under the current receiving direction is obtained. This forms the initial vertical RSS fingerprint database; , , (3) In equation (3), This represents the gain factor introduced by the polar angle and azimuth angle of the receiver at each fingerprint point for different LED light sources at the current position. (4) In equation (4), For the first The three-dimensional coordinates of an LED light source For the first The three-dimensional coordinates of the fingerprint point; Step 1.5: Estimate the two-dimensional coordinates of the user's initial position using the WKNN matching algorithm. ; Step 1.6, based on the two-dimensional coordinates of the user's initial position. Initialize the first expression (8) Vertical from the initial position of each LED light source , As the estimated vertical RSS value for the first location; , (8) In equation (8), For the first The three-dimensional coordinates of an LED light source This represents the photoelectric conversion efficiency of the photodetector.

[0009] The specific process of step 1.5 is as follows: Step 1.5.1, if the user is at the initial position, the receiver measures the signal from the first... The receiving power of each LED light source is Calculate using equation (5) With the fingerprint database Euclidean distance between fingerprint points ; The expression is: , (5) Step 1.5.2, Euclidean distance Arrange the items from smallest to largest, and record the distance after the arrangement as . ,Record The horizontal coordinate of the corresponding fingerprint point and ordinate ; Step 1.5.3, calculate the weighting coefficients using equation (6). ; The expression is: , (6) Step 1.5.4: Estimate the two-dimensional coordinates of the user's initial position using equation (7). ; The expression is: (7) In equation (7), To find the optimal number of nearest neighbors, we typically take... ; Weighting coefficients The former indivual, .

[0010] The specific process of step 2 is as follows: Step 2.1: Initialize the posterior error covariance matrix based on experience. Process noise covariance and measurement noise covariance ; Step 2.2, assuming every If the PL index is estimated once per second, the number of estimations required is: Users with speed Continue moving to the 1st The position, according to the first Azimuth angle of the receiver measured at each location and the Estimated two-dimensional coordinates of each location Predicting the first The three-dimensional coordinates of each position ; Step 2.3, calculate the first... The position and the first Predicted distance between LED light sources Calculate the first The position and the first Predicted distance between LED light sources ; The expression is:

[0011] .

[0012] In step 2.2, predict the first... The three-dimensional coordinates of each position The expression is: , (9).

[0013] The specific process of step 3 is as follows: Step 3.1, the first The vertical RSS value at each location is denoted as . Then for The dynamic linear model for prediction is represented as follows: , (10) In equation (10), for The state transition matrix; Indicates the first The estimated vertical RSS values ​​at each location are of size [value missing]. ; for The control input matrix; Indicates size is Control input; Let be the process noise, with a mean of zero and a process noise covariance of . Gaussian distribution; in, (11) (12) In equation (12), Indicates the first The position comes from the first Control input factor of an LED light source ; For the first The position The Lambert index of an LED light source; (13) Step 3.2, using the prior error covariance matrix Indicates the first The uncertainty in estimating the vertical RSS value at each location, namely: , (14) In equation (14), Let Q be the posterior error covariance matrix, representing the uncertainty of the estimated vertical RSS value at the previous position; k Represents the process noise covariance; Step 3.3, if the first The polar angle and azimuth angle measured by the receiver at each location are: and The calculation of the first LED light source is performed using equation (4). The polar angle of the receiver at each position and azimuth Gain factors introduced at each fingerprint point ;according to Formula (3) is used to recalculate the tilt reception power of each fingerprint point in the fingerprint database. This will obtain an updated RSS fingerprint database; if the user is in the first... The measurement from the terminal at the location is obtained from the first location. The vertical RSS value of each LED light source is ,in accordance with Get the first Received power at each location Repeat steps 1.5 through 1.8 to estimate the user's [number]. Two-dimensional coordinates of each position ; Step 3.4, the The observation equation for the vertical RSS at each location is: , (15) In equation (15), for The observation matrix Indicates the first The vertical RSS values ​​measured at each location, with a magnitude of [value missing]. ; The noise to be measured follows a mean of zero, and the measured noise covariance is... Gaussian distribution; in, , (16) In equation (16), Calculated using formula (4), it represents the value of different LED light sources in the first... The gain factor introduced by the polar angle and azimuth angle of the receiver at each fingerprint point at each location; Step 3.5, Calculate the Kalman gain : , (17) Step 3.6, based on Kalman gain Calculate the first The optimal estimate of the vertical RSS value at each location, i.e.: , (18) Step 3.7, using the posterior error covariance matrix This represents the uncertainty in estimating the vertical RSS value at the current location, i.e.: , (19) In equation (19), I represents N × N The identity matrix.

[0014] The specific process of step 4 is as follows: Step 4.1, calculate the first... The position and the first The distance between LED light sources , No. The position and the first Estimated distance between LED light sources ; The expression is:

[0015]

[0016] Step 4.2, calculate the first... The position is the first PL index of an LED light source The estimate is expressed as: , (20) Step 4.3, according to Update the Lambert index .

[0017] The specific process of step 5 is as follows: Step 5.1, the user moves to the... For the position, repeat step 3, for the position... Estimate the vertical RSS at the nth position, and estimate the vertical RSS at the nth position. Two-dimensional coordinates of each position Make an estimate; Step 5.2, repeat step 4, for the first... PL index at each position Make an estimate; Step 5.3: If the user leaves the indoor scene, end the estimation of the vertical RSS value, the estimation of the two-dimensional coordinates of the location, and the estimation of the PL index for the new location; Step 5.4: If the user does not leave the indoor scene, repeat steps 5.1 to 5.4.

[0018] The beneficial effects of this invention are as follows: This invention provides an indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database. It utilizes Kalman filtering to estimate and update the PL index of the LED light source in the online stage. Based on the estimated and updated link loss index, the RSS fingerprint database is updated online during user movement. The WKNN matching algorithm is used to achieve positioning, avoiding the problem that the PL index of the LED light source needs to be measured and updated offline. It also avoids the decrease in positioning accuracy caused by random changes in the PL index due to LED light decay. It is suitable for real-time positioning with both vertical and tilt reception, and is a universal positioning method with low complexity and low cost applicable to practical scenarios. Attached Figure Description

[0019] Figure 1 This is a hardware structure diagram of the indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database of the present invention. Figure 2 This is a flowchart of the indoor visible light anti-tilt positioning method for self-generated and self-updated RSS fingerprint database of the present invention; Figure 3 The PL index estimated for the first LED light source at 16 positions during tilting movement in Example 8. and measured PL index Comparison chart; Figure 4 The PL index estimated for the second LED light source at 16 positions during tilting movement in Example 8. and measured PL index Comparison chart; Figure 5 The PL index estimated for the third LED light source at 16 positions during tilting movement in Example 8. and measured PL index Comparison chart; Figure 6 The PL index estimated for the fourth LED light source at 16 positions during tilting movement in Example 8. and measured PL index Comparison chart; Figure 7 The estimated PL index in Example 8 The generated fingerprint features are compared with those based on measured fingerprint features and measured PL index. A comparison chart of the localization error probabilities of fingerprint features generated from the mean.

[0020] In the diagram, 1. LED light source, 2. gyroscope, 3. photodetector, 4. receiving terminal. Detailed Implementation

[0021] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0022] Example 1 The hardware device used in the self-generating and self-updating indoor visible light anti-tilt positioning method of the RSS fingerprint database in this invention, namely the indoor visible light fingerprint positioning system, has the following structure: Figure 1 As shown, the device includes several LED light sources 1 installed on the roof, and a receiving terminal 4 installed below the LED light sources 1. The receiving terminal 4 is equipped with a gyroscope 2 and a photodetector 3. The LED light sources 1 are white LEDs with a half-power angle of 45 degrees. The photodetector 3 is used to convert light signals into current signals with a sensitivity of 0.4 A / W. The gyroscope 2 is used to measure the polar angle and azimuth angle of the receiving terminal 4. The receiving terminal 4 includes, but is not limited to, smartphones, tablets, etc., equipped with the photodetector 3.

[0023] Example 2 Based on Example 1, the present invention provides an indoor visible light tilt-resistant positioning method with self-generated and self-updated RSS fingerprint database, such as... Figure 2 As shown, the steps are as follows: Step 1, assuming there is indoor... One LED light source, initialize PL index. And obtain the estimated vertical RSS value of the user's first location; Step 2, initialize Kalman filter parameters, predict the first... The three-dimensional coordinates of the nth position are calculated, and the nth position is calculated. The position, the The positions are respectively related to the first... Predicted distance between individual LED light sources; Step 3: Use Kalman filtering to obtain the estimated RSS value of the current position; Step 4: Estimate and update the PL index of each LED light source based on the estimated vertical RSS values ​​of the current and previous positions. ; Step 5, if the user moves to the... Repeat steps 3-4 at the first position, continuing for the second position. Estimate the vertical RSS value at the nth position, and estimate the vertical RSS value at the nth position. Estimate the two-dimensional coordinates of the i-th position, and estimate the i-th position's coordinates. The PL index for the first location is estimated; if the user leaves the indoor positioning scenario, the positioning process ends. Estimate the vertical RSS value, two-dimensional coordinates, and PL index at each location.

[0024] Example 3 Based on Example 2, the specific process of step 1 is as follows: Step 1.1: Initialize the power index (PL) of the light source based on its half-power angle. , , ; Step 1.2, calculate the vertical received power at the receiving position directly below each LED light source 1. ; The expression is: , (1) In equation (1), The transmission power of the LED light source, The effective detection area of ​​the photodetector. The vertical height between the LED and the receiver. For the first The initial value of the Lambertian index of each LED light source. For the first The distance between each LED light source and the receiving position directly below the LED light source; Step 1.3, divide the receiving plane into The grid is a set of grids, with the center of each grid representing the location of a fingerprint point in the fingerprint database. The 1st grid is generated using formula (2). The fingerprint point received the first Vertical receiving power of each LED light source ; (2) In equation (2), For the first The distance between each LED light source and the receiving position directly below the LED light source For the first The LED light source and the first The distance between fingerprint points , .

[0025] Step 1.4, if the polar angle and azimuth angle of the receiver at the user's initial position are respectively and According to equation (3), the tilt reception power of each fingerprint point at the receiver under the current receiving direction is obtained. This forms the initial vertical RSS fingerprint database; , , (3) In equation (3), This represents the gain factor introduced by the polar angle and azimuth angle of the receiver at each fingerprint point for different LED light sources at the current position. (4) In equation (4), For the first The three-dimensional coordinates of an LED light source For the first The three-dimensional coordinates of the fingerprint point; Step 1.5: Estimate the two-dimensional coordinates of the user's initial position using the WKNN matching algorithm. ; The specific process is as follows: Step 1.5.1, if the user is at the initial position, the receiver measures the signal from the first... The receiving power of each LED light source is Calculate using equation (5) With the fingerprint database Euclidean distance between fingerprint points ; The expression is: , (5) Step 1.5.2, Euclidean distance Arrange the items from smallest to largest, and record the distance after the arrangement as . ,Record The horizontal coordinate of the corresponding fingerprint point and ordinate ; Step 1.5.3, calculate the weighting coefficients using equation (6). ; The expression is: , (6) Step 1.5.4: Estimate the two-dimensional coordinates of the user's initial position using equation (7). ; The expression is: (7) In equation (7), To find the optimal number of nearest neighbors, we typically take... ; Weighting coefficients The former indivual, ; Step 1.6, based on the two-dimensional coordinates of the user's initial position. Initialize the first expression (8) Vertical from the initial position of each LED light source , As the estimated vertical RSS value for the first location; , (8) In equation (8), For the first The three-dimensional coordinates of an LED light source This represents the photoelectric conversion efficiency of the photodetector.

[0026] Example 4 Based on Example 3, the specific process of step 2 is as follows: Step 2.1: Initialize the posterior error covariance matrix based on experience. Process noise covariance and measurement noise covariance ; Step 2.2, assuming every If the PL index is estimated once per second, the number of estimations required is: Users with speed Continue moving to the 1st The position, according to the first Azimuth angle of the receiver measured at each location and the Estimated two-dimensional coordinates of each location Predicting the first The three-dimensional coordinates of each position : , (9) Step 2.3, calculate the first... The position and the first Predicted distance between LED light sources Calculate the first The position and the first Predicted distance between LED light sources ; The expression is:

[0027] .

[0028] Example 5 Based on Example 4, the specific process of step 3 is as follows: Step 3.1, the first The vertical RSS value at each location is denoted as . Then for The dynamic linear model for prediction is represented as follows: , (10) In equation (10), for The state transition matrix; Indicates the first The estimated vertical RSS values ​​at each location are of size [value missing]. ; for The control input matrix; Indicates size is Control input; Let be the process noise, with a mean of zero and a process noise covariance of . Gaussian distribution; in, (11) (12) In equation (12), Indicates the first The position comes from the first Control input factor of an LED light source ; For the first The position The Lambert index of an LED light source; (13) Step 3.2, using the prior error covariance matrix Indicates the first The uncertainty in estimating the vertical RSS value at each location, namely: , (14) In equation (14), Let Q be the posterior error covariance matrix, representing the uncertainty of the estimated vertical RSS value at the previous position; k Represents the process noise covariance; Step 3.3, if the first The polar angle and azimuth angle measured by the receiver at each location are: and The calculation of the first LED light source is performed using equation (4). The polar angle of the receiver at each position and azimuth Gain factors introduced at each fingerprint point ;according to Formula (3) is used to recalculate the tilt reception power of each fingerprint point in the fingerprint database. This will obtain an updated RSS fingerprint database; if the user is in the first... The measurement from the terminal at the location is obtained from the first location. The vertical RSS value of each LED light source is ,in accordance with Get the first Received power at each location Repeat steps 1.5 through 1.8 to estimate the user's [number]. Two-dimensional coordinates of each position ; Step 3.4, the The observation equation for the vertical RSS at each location is: , (15) In equation (15), for The observation matrix Indicates the first The vertical RSS values ​​measured at each location, with a magnitude of [value missing]. ; The noise to be measured follows a mean of zero, and the measured noise covariance is... Gaussian distribution; in, , (16) In equation (16), Calculated using formula (4), it represents the value of different LED light sources in the first... The gain factor introduced by the polar angle and azimuth angle of the receiver at each fingerprint point at each location; Step 3.5, Calculate the Kalman gain : , (17) Step 3.6, based on Kalman gain Calculate the first The optimal estimate of the vertical RSS value at each location, i.e.: , (18) Step 3.7, using the posterior error covariance matrix This represents the uncertainty in estimating the vertical RSS value at the current location, i.e.: , (19) In equation (19), I represents N × N The identity matrix.

[0029] The vertical RSS value at each location is estimated using Kalman filtering, which is a cyclical process that begins when the user enters the room and the positioning starts at the first location. k =1, initialize the vertical RSS value with the estimated vertical RSS1 of the first position, then start estimating the second position, estimate the vertical RSS value of the second position, estimate the PL index at the second position, the user continues to move to the third position, continue to estimate the position, estimate the vertical RSS value of the third position, estimate the PL index at the third position, and so on until the user leaves the positioning scene, thus avoiding the problem that the PL index of the LED light source needs to be measured and updated in the offline stage, and improving the positioning accuracy.

[0030] Example 6 Based on Example 5, the specific process of step 4 is as follows: Step 4.1, calculate the first... The position and the first The distance between LED light sources , No. The position and the first Estimated distance between LED light sources ; The expression is:

[0031]

[0032] Step 4.2, calculate the first... The position is the first PL index of an LED light source The estimate is expressed as: , (20) Step 4.3, according to Update the Lambert index .

[0033] Example 7 Based on Example 6, the specific process of step 5 is as follows: Step 5.1, the user moves to the... For the position, repeat step 3, for the position... Estimate the vertical RSS at the nth position, and estimate the vertical RSS at the nth position. Two-dimensional coordinates of each position Make an estimate; Step 5.2, repeat step 4, for the first... PL index at each position Make an estimate; Step 5.3: If the user leaves the indoor scene, end the estimation of the vertical RSS value, the estimation of the two-dimensional coordinates of the location, and the estimation of the PL index for the new location; Step 5.4: If the user does not leave the indoor scene, repeat steps 5.1 to 5.4.

[0034] Example 8 The indoor visible light fingerprint positioning system used in this embodiment includes 4 LED light sources arranged in a matrix. The vertical distance between the LED light sources and the receiving plane is 1.2m. The receiving plane is a square with a side length of 1.2m. A receiving terminal is set on the receiving plane, and the receiving terminal is equipped with a gyroscope and a photodetector.

[0035] like Figure 3 As shown, the PL index of the first LED light source was measured offline. The variance is 0.0211, and the PL index of the first LED light source is estimated online based on Kalman filtering. The variance was 0.0393, which is consistent with the offline measured PL index. The variances are not significantly different, indicating that the method of the present invention can adjust the PL index of the first LED light source when the user moves. Make a reliable estimate.

[0036] like Figure 4 As shown, the PL index of the second LED light source was measured offline. The variance is 0.0271, and the PL index of the second LED light source is estimated online based on Kalman filtering. The variance was 0.0405, which is consistent with the offline actual measurement of the PL index. The variances are not significantly different, indicating that the method of the present invention can adjust the PL index of the second LED light source when the user moves. Make a reliable estimate.

[0037] like Figure 5 As shown, the PL index of the third LED light source was measured offline. The variance is 0.0304, and the PL index of the third LED light source is estimated online based on Kalman filtering. The variance was 0.0425, which is consistent with the offline measured PL index. The variances are not significantly different, indicating that the method of the present invention can adjust the PL index of the third LED light source when the user moves. Make a reliable estimate.

[0038] like Figure 6 As shown, the PL index of the fourth LED light source was measured offline. The variance is 0.0245, and the PL index of the fourth LED light source is estimated online based on Kalman filtering. The variance was 0.0303, which is consistent with the offline measured PL index. The variances are not significantly different, indicating that the method of the present invention can adjust the PL index of the fourth LED light source when the user is moving. Make a reliable estimate.

[0039] like Figure 7 As shown, the PL index is based on offline actual measurement and offline actual measurement. The 90% localization errors of the fingerprint features generated by the mean values ​​were 3.38 cm and 5.92 cm, respectively. The PL exponent was estimated online based on Kalman filtering. The 90% positioning error of the generated fingerprint features is approximately 13.55 cm, which avoids the PL index during the initial offline phase of system deployment. Repeated measurements can be performed to achieve the PL index of each LED light source even when the receiver is tilted during user movement. Estimation and centimeter-level positioning.

Claims

1. An indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database, characterized in that, The steps are as follows: Step 1, assuming there is indoor... One LED light source, initialize the PL index and obtain the estimated vertical RSS value of the user's first position; Step 2, initialize Kalman filter parameters, predict the first... The three-dimensional coordinates of the position are calculated and the first position is calculated. The position, the The positions are respectively related to the first... Predicted distance between individual LED light sources; Step 3: Use Kalman filtering to obtain the estimated RSS value of the current position; Step 4: Estimate and update the PL index of each LED light source based on the estimated vertical RSS values ​​of the current position and the previous position. Step 5, if the user moves to the... For each location, repeat steps 3-4 to continue estimating. If the user leaves the indoor positioning scene, end all estimations.

2. The indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database according to claim 1, characterized in that, The specific process of step 1 is as follows: Step 1.1: Initialize the power index (PL) of the light source based on its half-power angle. , , ; Step 1.2, calculate the vertical received power at the receiving position directly below each LED light source 1. ; The expression is: , （1） In equation (1), The transmission power of the LED light source, The effective detection area of ​​the photodetector. The vertical height between the LED and the receiver. For the first The initial value of the Lambert index of each LED light source. For the first The distance between each LED light source and the receiving position directly below the LED light source; Step 1.3, divide the receiving plane into The grid is a set of grids, with the center of each grid representing the location of a fingerprint point in the fingerprint database. The 1st grid is generated using formula (2). The fingerprint point received the first Vertical receiving power of each LED light source ; （2） In equation (2), For the first The distance between each LED light source and the receiving position directly below the LED light source For the first The LED light source and the first The distance between fingerprint points , ; Step 1.4, if the polar angle and azimuth angle of the receiver at the user's initial position are respectively and According to equation (3), the tilt reception power of each fingerprint point at the receiver under the current receiving direction is obtained. This forms the initial vertical RSS fingerprint database; ， ， （3） In equation (3), This represents the gain factor introduced by the polar angle and azimuth angle of the receiver at each fingerprint point for different LED light sources at the current position. （4） In equation (4), For the first The three-dimensional coordinates of an LED light source For the first The three-dimensional coordinates of the fingerprint point; Step 1.5: Estimate the two-dimensional coordinates of the user's initial position using the WKNN matching algorithm. ; Step 1.6, based on the two-dimensional coordinates of the user's initial position. Initialize the first expression (8) Vertical from the initial position of each LED light source , As the estimated vertical RSS value for the first location; ， （8） In equation (8), For the first The three-dimensional coordinates of an LED light source This represents the photoelectric conversion efficiency of the photodetector.

3. The indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database according to claim 2, characterized in that, The specific process of step 1.5 is as follows: Step 1.5.1, if the user is at the initial position, the receiver measures the signal from the first... The receiving power of each LED light source is Calculate using equation (5) With the fingerprint database Euclidean distance between fingerprint points ; The expression is: ， （5） Step 1.5.2, Euclidean distance Arrange the items from smallest to largest, and record the distance after the arrangement as . ,Record The corresponding horizontal coordinate of the fingerprint point and ordinate ; Step 1.5.3, calculate the weighting coefficients using equation (6). ; The expression is: ， （6） Step 1.5.4: Estimate the two-dimensional coordinates of the user's initial position using equation (7). ; The expression is: （7） In equation (7), To find the optimal number of nearest neighbors, we typically take... ; Weighting coefficients The former indivual, .

4. The indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database according to claim 3, characterized in that, The specific process of step 2 is as follows: Step 2.1: Initialize the posterior error covariance matrix based on experience. Process noise covariance and measurement noise covariance ; Step 2.2, assuming every If the PL index is estimated once per second, the number of estimations required is: Users with speed Continue moving to the 1st The position, according to the first Azimuth angle of the receiver measured at each location and the Estimated two-dimensional coordinates of each location Predicting the first The three-dimensional coordinates of each position ; Step 2.3, calculate the first... The position and the first Predicted distance between LED light sources Calculate the first The position and the first Predicted distance between LED light sources ; The expression is: 。 5. The indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database according to claim 4, characterized in that, In step 2.2, predict the first... The three-dimensional coordinates of each position The expression is: , （9）。 6. The indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database according to claim 5, characterized in that, The specific process of step 3 is as follows: Step 3.1, the first The vertical RSS value at each location is denoted as . Then for The dynamic linear model for prediction is represented as follows: ， （10） In equation (10), for The state transition matrix; Indicates the first The estimated vertical RSS values ​​at each location are of size [value missing]. ; for The control input matrix; Indicates size is Control input; Let be the process noise, with a mean of zero and a process noise covariance of . Gaussian distribution; in, （11） （12） In equation (12), Indicates the first The position comes from the first Control input factor of each LED light source ; For the first The position The Lambert index of an LED light source; （13） Step 3.2, using the prior error covariance matrix Indicates the first The uncertainty in estimating the vertical RSS value at each location, namely: ， （14） In equation (14), Let Q be the posterior error covariance matrix, representing the uncertainty of the estimated vertical RSS value at the previous position; k Represents the process noise covariance; Step 3.3, if the first The polar angle and azimuth angle measured by the receiver at each location are: and The calculation of the first LED light source is performed using equation (4). The polar angle of the receiver at each position and azimuth Gain factors introduced at each fingerprint point ;according to Formula (3) is used to recalculate the tilt reception power of each fingerprint point in the fingerprint database. This will obtain an updated RSS fingerprint database; if the user is in the first... The measurement from the terminal at the location is from the first location. The vertical RSS value of each LED light source is ,in accordance with Get the first Received power at each location Repeat steps 1.5 through 1.8 to estimate the user's [number]. Two-dimensional coordinates of each position ; Step 3.4, the The observation equation for the vertical RSS at each location is: ， （15） In equation (15), for The observation matrix Indicates the first The vertical RSS values ​​measured at each location, with a magnitude of [value missing]. ; The noise to be measured follows a mean of zero, and the measured noise covariance is... Gaussian distribution; in, , （16） In equation (16), Calculated using formula (4), it represents the value of different LED light sources in the first... The gain factor introduced by the polar angle and azimuth angle of the receiver at each fingerprint point at each location; Step 3.5, Calculate the Kalman gain : ， （17） Step 3.6, based on Kalman gain Calculate the first The optimal estimate of the vertical RSS value at each location, i.e.: ， （18） Step 3.7, using the posterior error covariance matrix This represents the uncertainty in estimating the vertical RSS value at the current location, i.e.: ， （19） In equation (19), I represents N × N The identity matrix.

7. The indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database according to claim 6, characterized in that, The specific process of step 4 is as follows: Step 4.1, calculate the first... The position and the first The distance between LED light sources , No. The position and the first Estimated distance between LED light sources ; The expression is: Step 4.2, calculate the first... The position is the first PL index of an LED light source The estimate is expressed as: ， （20） Step 4.3, according to Update the Lambert index .

8. The indoor visible light anti-tilt positioning method with self-generated and self-updated RSS fingerprint database according to claim 7, characterized in that, The specific process of step 5 is as follows: Step 5.1, the user moves to the... For the position, repeat step 3, for the position... Estimate the vertical RSS at the nth position, and estimate the vertical RSS at the nth position. Two-dimensional coordinates of each position Make an estimate; Step 5.2, repeat step 4, for the first... PL index at each position Make an estimate; Step 5.3: If the user leaves the indoor scene, end the estimation of the vertical RSS value, the estimation of the two-dimensional coordinates of the location, and the estimation of the PL index for the new location; Step 5.4: If the user does not leave the indoor scene, repeat steps 5.1 to 5.4.

Images