A high-precision outdoor positioning method for UAVs based on visible light

Abstract

The invention provides a high-precision outdoor positioning method for an unmanned aerial vehicle based on visible light, comprising the following steps: calculating the intensity of non-interference optical signals received by the target point from three surrounding unmanned aerial vehicles according to a visible light communication channel model; obtaining the total optical power received by the target point according to an outdoor ambient light noise model; calculating the actual distance between the unmanned aerial vehicle and the target point according to the non-interference received optical power; calculating the measured distance between the unmanned aerial vehicle and the target point according to the received optical power interfered by ambient light noise; changing the target position and the unmanned aerial vehicle allocation position, repeating the above four steps, obtaining a large number of actual distance values ​​and distance measurement values, using them as training data, and correcting the error between the two groups of distance values ​​by using a BP neural network; optimizing the distance value measured in the previous step by using the trained BP neural network; and calculating the position of the target according to the positions of the three unmanned aerial vehicles around the target and the optimized distance value.

Description

Technical Field

[0001] The invention relates to a high-precision outdoor positioning method for an unmanned aerial vehicle based on visible light, and belongs to the field of wireless optical communications. Background Art

[0002] Since traditional radio frequency communication is easily affected by electromagnetic interference and electromagnetic silence, resulting in unstable performance, visible light communication can provide high-speed, large-capacity communication coverage for designated areas with its rich broadband resources and advantages of not being susceptible to external electromagnetic interference. At the same time, drones have the advantages of small size, high maneuverability, and coordinated deployment. The visible light positioning technology assisted by drones has broad research prospects. In some specific environments such as disaster-stricken areas, when traditional radio frequency communication may be interrupted, as a supplementary technology, we can combine drones with visible light communication to achieve communication coverage of the area.

[0003] In outdoor environments, ambient light interference such as natural light sources and artificial light sources is the main factor affecting the positioning performance in outdoor visible light communication. If the background light radiation is strong (for example, when exposed to direct sunlight), that is, when the average light power from natural light sources and artificial light sources is stronger than the light signal power radiated by the drone, the visible light communication performance will be seriously impaired. At present, researchers in the industry mainly focus on visible light positioning technology based on indoor environments and indoor ambient light noise elimination technology. However, considering that the impact of ambient light noise on the positioning effect in outdoor environments is difficult to ignore, it is very important to propose a positioning technology with intelligent outdoor ambient light noise elimination function based on traditional visible light positioning technology. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to propose a high-precision outdoor positioning method for UAVs based on visible light in response to the defects or problems of the prior art.

[0005] A high-precision outdoor positioning method for UAVs based on visible light randomly generates n positioning target points (x k ,y k )(k=1,2,...n), three points (x k 1 ,y k 1 ,h),(x k 2 ,y k 2 ,h),(x k 3 ,y k 3,h)(k=1,2,...n), h is the height, which is the distribution position of the three UAVs when communicating with the target point.

[0006] The high-precision outdoor positioning method of UAV based on visible light includes the following steps:

[0007] Step 1: According to the visible light communication channel model, calculate the interference-free received optical power received by the target point from the three surrounding drones

[0008] Step 2: According to the outdoor ambient light noise model, the interference-free received optical power obtained in step 1 is Ambient light noise power Add together to get the total optical power received by the target point affected by ambient light noise

[0009] Step 3: According to the interference-free received optical power obtained in step 1 Calculate the actual distance between the drone and the target point

[0010] Step 4: According to the total optical power received by the target point in step 2 and interfered by the ambient light noise The distance between the three drones and the target point is calculated as the distance measurement value.

[0011] Step 5: Change the position of the target point and the positions of the three surrounding drones, repeat the above four steps, and obtain a large number of actual distance values And the distance value (r = 1, 2, 3, t is the number of repetitions), and use it as training data to correct the actual distance value using the BP neural network And the distance value The error between

[0012] Step 6: Use the trained BP neural network to measure the distance value in step 4 Optimize and get the optimized distance value

[0013] Step 7: The positions of the three drones around the target point (x k 1 ,y k 1 ,h),(x k 2 ,y k 2 ,h),(x k3 ,y k 3 ,h) and the optimized distance value obtained in step 6 According to the three-point positioning model, the position of the target (x k’ ,y k’ )(k=1,2,...n).

[0014] Further, in step 1, the visible light communication channel model is defined as:

[0015]

[0016] Where H(0) represents the channel DC gain, A represents the physical area of ​​the detector in the photodetector, and D d represents the distance between the ground user point and the UAV, represents the irradiation angle, ψ represents the incident angle, Ts(ψ) represents the gain of the optical filter, g(ψ) represents the gain of the optical concentrator, Ψ c represents the receiver field of view, and m is the order of the Lambert light source model.

[0017] Further, in step 1, the gain g(ψ) of the optical concentrator is defined as:

[0018]

[0019] Where n is the refractive index, ψ is the incident angle, c Indicates the receiver field of view.

[0020] Further, in step 2, the outdoor ambient light noise model is defined as:

[0021]

[0022] in represents the shot noise power, represents thermal noise, It is expressed as:

[0023]

[0024] Where q is the electron charge, B s is the equivalent noise bandwidth, R is the photoelectric conversion coefficient, P r is the optical power received by the receiver, I b is the background current intensity, and I2 is the noise bandwidth factor.

[0025] Furthermore, in step 2, the thermal noise It is expressed as:

[0026]

[0027] Where G is the open-loop voltage gain, k is the Boltzmann constant, T k is the absolute temperature, η is the capacitance per unit area of ​​the photodetector, A is the physical area of ​​the photodetector, I2 is the noise bandwidth factor, B s is the equivalent noise bandwidth, Γ is the field effect tube noise factor, g is the transconductance of the field effect tube, and I3 is the compensation factor.

[0028] Furthermore, in step 5, the actual distance value is corrected by using the BP neural network And the distance value The error between

[0029] First, calculate the hidden layer Hout in the BP neural network o The output of the output layer is r ,

[0030]

[0031]

[0032] in is the transfer function, w io represents the connection weight from the input layer to the hidden layer, v or represents the connection weight from the hidden layer to the output layer, b o represents the bias value of each neuron in the hidden layer, l r It represents the bias value of each neuron in the output layer, and s is the number of neurons in the hidden layer.

[0033] Secondly, calculate the mean square error between the output result and the expected result

[0034]

[0035] Use the gradient descent algorithm to continuously update the connection weights and bias values ​​in the BP neural network to reduce the mean square error e, that is,

[0036]

[0037] Where N represents the number of training times, μ represents the learning efficiency, and w io represents the connection weight from the input layer to the hidden layer, v or represents the connection weight from the hidden layer to the output layer, b o represents the bias value of each neuron in the hidden layer, l r Represents the bias value of each neuron in the output layer. When the mean square error e is controlled within the specified limit, the training is considered complete, that is,

[0038]

[0039] Among which Yout r,trained is the output of the output layer at the end of training, and e required represents the specified error limit.

[0040] Furthermore, in step seven, the three-point positioning model is defined as:

[0041]

[0042]

[0043]

[0044] Among which, is the distance between three drones around the target. By solving the above three equations simultaneously, the position [x k,solved , y k,solved , z k,solved of the target point will have two solutions, one is the true solution and the other is a discardable solution judged according to the flight altitude of the drone. By adding the constraint condition z k,solved <h, the position solution (x k’ , y k’ ) of the final target point can be obtained.

[0045] Beneficial effects:

[0046] Compared with the prior art, the present invention has the following advantages:

[0047] 1. Relying on the advantages of visible light communication technology such as being immune to external electromagnetic interference and having a vast amount of spectrum resources, and combining with the high mobility and rapid deployability of drones, the present invention combines visible light communication with drone positioning and innovatively proposes an outdoor visible light positioning technology assisted by drones. The present invention can not only provide lighting functions for the area, but also provide high-speed and large-capacity communication coverage, and achieve precise positioning of targets within the area.

[0048] 2. The present invention breaks through the existing research on visible light positioning technology based on indoor scenarios and realizes visible light positioning in outdoor scenarios. At the same time, considering the inevitable influence of ambient light noise in the outdoor environment on the positioning performance, the influence is quantified through the modeling of ambient light noise. On the basis of the classical three-point positioning method, a BP neural network is used to eliminate the ambient light noise, thereby realizing a high-precision outdoor visible light positioning method. Description of the drawings

[0049] Figure 1 is a schematic diagram of the positioning scenario involved in the present invention;

[0050] Figure 2This is a diagram showing the experimental simulation results of the relationship between positioning error and target quantity of different positioning algorithms of the present invention;

[0051] Figure 3 This is a simulation result diagram of the relationship between the positioning error and the receiver field of view FoV of different positioning algorithms of the present invention;

[0052] Figure 4 This is a diagram of experimental simulation results of the relationship between positioning error and number of drones for different positioning algorithms of the present invention. DETAILED DESCRIPTION

[0053] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments of the specification. The implementation methods of the present invention include but are not limited to the following embodiments.

[0054] The invention provides a high-precision outdoor positioning method for unmanned aerial vehicles based on visible light, which randomly generates n positioning target points (x k ,y k )(k=1,2,...n), 3 points (x k 1 ,y k 1 ,h),(x k 2 ,y k 2 ,h),(x k 3 ,y k 3 ,h)(k=1,2,...n), as the assigned positions of the three drones when communicating with the target. The method for drone-assisted high-precision outdoor visible light positioning based on intelligent algorithm to eliminate ambient light noise includes the following steps:

[0055] Step 1: According to the visible light communication channel model, calculate the interference-free received optical power received by the target point from the three surrounding drones

[0056]

[0057] Where P t represents the transmitted optical power, and H(0) is the channel DC gain:

[0058]

[0059] Where A represents the physical area of ​​the detector in the photodetector, and D d represents the distance between the ground user point and the UAV, represents the irradiation angle, ψ represents the incident angle, Ts(ψ) represents the gain of the optical filter, g(ψ) represents the gain of the optical concentrator, Ψ c represents the receiver field of view (FoV), and m is the order of the Lambert light source model.

[0060] The gain g(ψ) of an optical concentrator is defined as:

[0061]

[0062] Where n represents the refractive index.

[0063] Step 2: According to the outdoor ambient light noise model, the interference-free received optical power obtained in step 1 is Ambient light noise power Add together to get the total optical power received by the target point

[0064] The outdoor ambient light noise model is defined as:

[0065]

[0066] in represents the shot noise power, represents thermal noise. It can be expressed as:

[0067]

[0068] Where q is the electron charge, B s represents the equivalent noise bandwidth, I b is the background current intensity, I2 is the noise bandwidth factor, P r is the optical power received by the receiver, and R represents the photoelectric conversion coefficient.

[0069] Thermal Noise It can be expressed as:

[0070]

[0071] Where G is the open-loop voltage gain, k is the Boltzmann constant, T k is the absolute temperature, η is the capacitance per unit area of ​​the photodetector, A is the physical area of ​​the photodetector, I2 is the noise bandwidth factor, B s represents the equivalent noise bandwidth, Γ is the field effect tube noise factor, g is the transconductance of the field effect tube, which is about tens of milli-conductors, and I3 is the compensation factor, whose value is 0.868.

[0072] Step 3: Based on the interference-free received optical power obtained in step 1 Calculate the actual distance between the drone and the target point

[0073]

[0074] Where m is the order of the Lambert light source model, represents the irradiation angle, ψ represents the incident angle, Ts(ψ) represents the gain of the optical filter, g(ψ) represents the gain of the optical concentrator, P t Indicates the transmitted optical power.

[0075] Step 4: Based on the received optical power interfered by the ambient light noise in step 2 Calculate the measured distance between the drone and the target point

[0076]

[0077] Where m is the order of the Lambert light source model, represents the irradiation angle, ψ represents the incident angle, Ts(ψ) represents the gain of the optical filter, g(ψ) represents the gain of the optical concentrator, P t Indicates the transmitted optical power.

[0078] Step 5: Change the target position and the drone assigned position, repeat the above four steps to obtain a large number of actual distance values And the distance value (r = 1, 2, 3, t is the number of repetitions), and use it as training data to correct the error between the two sets of distance values ​​using the BP neural network. First, calculate the hidden layer Hout in the neural network o The output of the output layer is r ,

[0079]

[0080]

[0081] in is the transfer function, w io represents the connection weight from the input layer to the hidden layer, v or represents the connection weight from the hidden layer to the output layer, b o represents the bias value of each neuron in the hidden layer, l r It represents the bias value of each neuron in the output layer, and s is the number of neurons in the hidden layer.

[0082] Then find the mean square error between the output result and the expected result

[0083]

[0084] And use the gradient descent algorithm to continuously update the connection weights and bias values ​​in the neural network to reduce the mean square error e, that is,

[0085]

[0086] Where N represents the number of training times, μ represents the learning efficiency, and w io represents the connection weight from the input layer to the hidden layer, v or represents the connection weight from the hidden layer to the output layer, b o represents the bias value of each neuron in the hidden layer, l r Represents the bias value of each neuron in the output layer. When the mean square error e is controlled within the specified limit, the training is considered complete, that is,

[0087]

[0088] Yout ,trained is the output of the output layer at the end of training, e required Indicates the specified error limit.

[0089] Step 6: Use the trained BP neural network to calculate the distance value measured in step 4 Optimize and get the optimized distance value

[0090]

[0091]

[0092]

[0093] Where Hout o ' represents the hidden layer output of the trained neural network, Yout r ' represents the output layer output of the trained neural network, w io , , b o ' represents the connection weights in the neural network at the end of training, s is the number of neurons in the hidden layer, v ir ,,l r ' represents the bias value in the neural network at the end of training.

[0094] Step 7: The positions of the three drones around the target (x k 1 ,y k 1 ,h),(x k 2 ,y k 2 ,h),(xk 3 , y k 3 , h) and the optimized distance value obtained in step 6 According to the three-point positioning model, calculate the position (x k’ , y k’ ) (k = 1, 2,... n).

[0095]

[0096]

[0097]

[0098] By simultaneously solving the above three equations for the position of the target [x k,solved , y k,solved , z k,solved will obtain two solutions, one is the true solution, and the other is a discardable solution that can be judged according to the flight altitude of the UAV. Therefore, a constraint condition z k,solved < h can be added to obtain the final solution of the target position (x k’ , y k’ ).

[0099] Figure 2 Shows the variation trend of the positioning error of the positioning algorithm (BP optimization + three-point positioning method) of the present invention with the number of targets in the region increasing from 3 to 8 compared with the positioning errors of the other two positioning algorithms: the least squares method and the three-point positioning method without BP optimization. We carried out simulation in MATLAB software. The simulation results show that compared with the three-point positioning method without BP optimization, the least squares method has higher positioning accuracy, which is due to the higher anti-error ability of the least squares method. And the three-point positioning method combined with BP optimization reaches the highest positioning accuracy because the optimized distance value used in its positioning is approximately equal to the true distance value.

[0100] Figure 3 Using MATLAB software to simulate the variation trend of the positioning errors of the three positioning algorithms with the change of FoV. It can be seen from the simulation results that the positioning error of the three-point positioning method without BP optimization is greatly affected by the change of FoV and reaches a peak when FoV is 90°. The positioning error of the least squares method also changes with the change of FoV, but the change range is small. While the positioning error of the three-point positioning method combined with BP optimization remains basically unchanged, indicating that the stability of this positioning method is relatively high. This is because the influence of noise on positioning can always be controlled within the expected range through the optimization of the BP neural network. Therefore, no matter how FoV changes, the three-point positioning is always based on the approximate value of the actual distance.

[0101] Figure 4 The MATLAB software was used to simulate the changing trend of the positioning errors of the three positioning algorithms with the number of drones. Unlike the usual positioning with three drones, the positioning with more drones requires the selection of the three drones that receive the maximum optical power of the target. From the simulation results, it can be seen that the positioning errors of the three algorithms all show a downward trend with the increase in the number of drones. However, when the number of drones reaches 5 or more, the downward trend becomes less obvious. Therefore, considering the cost and positioning accuracy, using 5 drones for positioning is the best choice.

[0102] The above embodiments are only preferred implementation modes of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and equivalent substitutions without departing from the principles of the present invention. These technical solutions after improvements and equivalent substitutions to the claims of the present invention all fall within the protection scope of the present invention.

Claims

1. A high-precision outdoor positioning method for UAV based on visible light, characterized by: Randomly generate n positioning target points (x k ,y k )(k=1,2,...n), three points (x k 1 ,y k 1 ,h),(x k 2 ,y k 2 ,h),(x k 3 ,y k 3 ,h)(k=1,2,...n), h is the height, which is the distribution position of the three UAVs when communicating with the target point. The method for high-precision outdoor positioning of a UAV based on visible light comprises the following steps: Step 1: According to the visible light communication channel model, calculate the interference-free received optical power received by the target point from the three surrounding drones Step 2: According to the outdoor ambient light noise model, the interference-free received optical power obtained in step 1 is Ambient light noise power Add together to get the total optical power received by the target point affected by ambient light noise Step 3: According to the interference-free received optical power obtained in step 1 Calculate the actual distance between the drone and the target point Step 4: According to the total optical power received by the target point in step 2 and interfered by the ambient light noise The distance between the three drones and the target point is calculated as the distance measurement value. Step 5: Change the position of the target point and the positions of the three surrounding drones, repeat the above four steps, and obtain a large number of actual distance values And the distance value (r = 1, 2, 3, t is the number of repetitions), and use it as training data to correct the actual distance value using the BP neural network And the distance value The error between Step 6: Use the trained BP neural network to measure the distance value in step 4 Optimize and get the optimized distance value Step 7: The positions of the three drones around the target point (x k 1 ,y k 1 ,h),(x k 2 ,y k 2 ,h),(x k 3 ,y k 3 ,h) and the optimized distance value obtained in step 6 According to the three-point positioning model, the position of the target (x k ',y k ')(k=1,2,...n).

2. The method for high-precision outdoor positioning of a UAV based on visible light according to claim 1, characterized in that: In step 1, the visible light communication channel model is defined as: Where H(0) represents the channel DC gain, A represents the physical area of ​​the detector in the photodetector, and D d represents the distance between the ground user point and the UAV, represents the irradiation angle, ψ represents the incident angle, Ts(ψ) represents the gain of the optical filter, g(ψ) represents the gain of the optical concentrator, Ψ c represents the receiver field of view, and m is the order of the Lambert light source model.

3. The method for high-precision outdoor positioning of a UAV based on visible light according to claim 2 is characterized in that: In step 1, the gain g(ψ) of the optical concentrator is defined as: Where n is the refractive index, ψ is the incident angle, c Indicates the receiver field of view.

4. The method for outdoor high-precision positioning of a UAV based on visible light according to claim 1, characterized in that: In step 2, the outdoor ambient light noise model is defined as: in represents the shot noise power, represents thermal noise, It is expressed as: Where q is the electron charge, B s is the equivalent noise bandwidth, R is the photoelectric conversion coefficient, P r is the optical power received by the receiver, I b is the background current intensity, and I2 is the noise bandwidth factor.

5. The method for high-precision outdoor positioning of a UAV based on visible light according to claim 4 is characterized in that: In step 2, the thermal noise It is expressed as: Where G is the open-loop voltage gain, k is the Boltzmann constant, T k is the absolute temperature, η is the capacitance per unit area of ​​the photodetector, A is the physical area of ​​the photodetector, I2 is the noise bandwidth factor, B s is the equivalent noise bandwidth, Γ is the field effect tube noise factor, g is the transconductance of the field effect tube, and I3 is the compensation factor.

6. The method for high-precision outdoor positioning of a UAV based on visible light according to claim 1, characterized in that: In step 5, the actual distance value is corrected by using the BP neural network And the distance value The error between First, calculate the hidden layer Hout in the BP neural network o The output layer outputs Yout r , in is the transfer function, w io represents the connection weight from the input layer to the hidden layer, v or represents the connection weight from the hidden layer to the output layer, b o represents the bias value of each neuron in the hidden layer, l r represents the bias value of each neuron in the output layer, s is the number of neurons in the hidden layer, Secondly, calculate the mean square error between the output result and the expected result Use the gradient descent algorithm to continuously update the connection weights and bias values ​​in the BP neural network to reduce the mean square error e, that is, Where N represents the number of training times, μ represents the learning efficiency, and w io represents the connection weight from the input layer to the hidden layer, v or represents the connection weight from the hidden layer to the output layer, b o represents the bias value of each neuron in the hidden layer, l r Represents the bias value of each neuron in the output layer. When the mean square error e is controlled within the specified limit, the training is considered complete, that is, Yout r,trained is the output of the output layer at the end of training, e required Indicates the specified error limit.

7. The method for high-precision outdoor positioning of a UAV based on visible light according to claim 1, characterized in that: In step 7, the three-point positioning model is defined as: Among them, is the distance between three drones around the target and the target. By联立上述三式计算目标点的位置[x k,solved ,y k,solved ,z k,solved , two solutions will be obtained. One is the true solution, and the other is a discardable solution judged according to the flight altitude of the drone. Adding the constraint condition z k,solved <h to obtain the position solution (x k ’, y k ’) of the final target point. It should be noted that the part "联立上述三式计算目标点的位置" in the original text seems to be an incomplete or incorrect expression. It might need to be further clarified in the original context for a more accurate translation.

Images