Multi-base-station geometric space construction unmanned aerial vehicle positioning method based on fitting algorithm

By deploying multiple base stations in complex environments and optimizing drone positioning using fitting algorithms and timing models, the problem of low accuracy in complex environments is solved, and high accuracy and robustness are achieved Differential positioning.

CN119064857APending Publication Date: 2024-12-03STATE GRID HUBEI ELECTRIC POWER RES INST
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411150912.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Traditional drone positioning methods have low positioning accuracy in complex environments and are susceptible to occlusion and electronic interference, resulting in insufficient robustness.

Method used

The geometric space drone positioning method is adopted based on the fitting algorithm. By deploying multiple base stations for ranging, combining timing model and polynomial regression, the positioning of the drone in three-dimensional space is optimized, and the distance measurement error is reduced. Improve robustness.

Benefits of technology

It realizes high-precision positioning of the drone in different environments, enhances the robustness of the system, and can work stably under a variety of environmental conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119064857A_ABST
    Figure CN119064857A_ABST
Patent Text Reader

Abstract

The invention provides a multi-base-station geometric space unmanned aerial vehicle positioning method based on a fitting algorithm, and the method comprises the steps: obtaining the distance data between an unmanned aerial vehicle and a base station through the deployment of a plurality of base stations, calculating the position of the unmanned aerial vehicle according to a formula, and introducing a model to assist the ranging calibration in combination with the fitting algorithm. According to the method, the error between the distance measurement data and the model prediction value is reduced, accurate positioning of the unmanned aerial vehicle in the three-dimensional space is realized, the measurement error and interference can be inhibited, the robustness of the positioning system is improved, and the undifferentiated high-precision positioning of the unmanned aerial vehicle in different environments is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle positioning, and in particular to a method for unmanned aerial vehicle positioning based on a fitting algorithm by constructing a geometric space with multiple base stations. Background Art

[0002] With the increasing demand for location services and the rapid development of drone technology, drone-based positioning methods have been widely used and understood. Traditional drone positioning methods usually rely on ground base stations, satellites and other hardware equipment for ranging and positioning, which are easily affected by buildings, trees, etc., resulting in inaccurate positioning or loss of signal. They are also prone to electronic interference or malicious human interference, resulting in low positioning accuracy in complex environments. Summary of the invention

[0003] The present invention provides a method for positioning a UAV in a geometric space constructed by multiple base stations based on a fitting algorithm. The method uses a distance measurement technology to obtain the distance data between the UAV and the base station through the deployed multiple base stations, calculates the position of the UAV according to the formula, and combines the fitting algorithm to introduce a model-assisted distance measurement calibration to reduce the error between the distance measurement data and the model prediction value, thereby realizing accurate positioning of the UAV in three-dimensional space. The method can suppress measurement errors and interference, improve the robustness of the positioning system, and realize indiscriminate high-precision positioning of the UAV in different environments.

[0004] A method for positioning unmanned aerial vehicles in geometric space using multiple base stations based on a fitting algorithm comprises the following steps:

[0005] Step S1: Deploy four base stations with uniform positions in a three-dimensional space, and determine the coordinate position A (X A ,Y A ,Z A )、B(X B ,Y B ,Z B )、C(X C ,Y C ,Z C )、D(X D ,Y D ,Z D );

[0006] Step S2: Design the flight path, flight altitude and flight speed of the UAV, and plan the flight trajectory of the UAV according to the actual application scenario and requirements;

[0007] Step S3: During the flight of the drone, the drone and each base station transmit data by sending pulse signals to each other; a signal receiver and recorder is used to record the sending time t of the drone sending the signal to the base station. A 1. t B1. t C 1. t D 1. The receiving time t when the base station receives the signal A 2. t B 2. t C 2. t D 2. The time t at which the base station sends a response signal to the drone A 3. t B 3. t C 3. t D 3. The receiving time t when the drone receives the response signal A 4. t B 4. t C 4. t D 4 and the flight altitude, flight speed, and flight acceleration of the drone at the corresponding time point;

[0008] Step S4: Calculate the round-trip time RT of the signal between each base station and the drone based on the data received and recorded by the recorder in step S3 and the signal propagation speed. A , RT B , RT C , RT D , one-way time ST A , ST B 、ST C 、ST D and the distance S between the drone and each base station A , S B , S C , S D , where the signal propagation speed is the speed of light c, and the position of the drone is calculated based on the coordinate position of each base station and the distance between the drone and each base station;

[0009] Step S5: designing a time series model LMM for learning the mapping relationship between input features and drone positions. The time series model LMM utilizes a deep neural network structure based on the time recursion principle, can process input data in time sequence, and captures the time dependency in the data by maintaining an internal state.

[0010] Step S6: Collect the sending and receiving times recorded in step S3 and the flight altitude, flight speed, and flight acceleration of the drone at the corresponding time points, the location information of the drone calculated in S4, the distance between the base station and the drone, and the unique identifier used to distinguish different base stations and the location coordinates of the base stations as a training set, sort the training set in chronological order, and then use the sliding window technology to construct the input sequence.

[0011] Step S7: Apply the input sequence generated in step S6 to the time series model LMM, and predict the position of the drone through the time series model LMM;

[0012] Step S8: According to the difference between the UAV position predicted by the time series model LMM and the reference position obtained by geometric calculation in S4, the time series model LMM is optimized using the objective function FMSE and the chain rule;

[0013] Step S9: Use the optimized time series model LMM to predict the new ranging data and generate the real-time position estimate of the UAV; then, combine the real-time position estimate with the known coordinates of the base station and use triangulation to preliminarily estimate the position p0(m,n,o) of the UAV in three-dimensional space;

[0014] After obtaining the preliminary estimated UAV position p0, polynomial regression is used to optimize p0(m,n,o) to improve the accuracy of position estimation.

[0015] Furthermore, in step S4, the round-trip time RT of the signal between each base station and the drone is calculated based on the data received by the signal in step S3 and recorded by the recorder and the signal propagation speed. A , RT B , RT C , RT D , one-way time ST A 、ST B , ST C , ST D and the distance S between the drone and each base station A , S B , S C , S D , the calculation formula is as follows:

[0016]

[0017] Calculate the drone position p(X,Y,Z) using the following formula:

[0018]

[0019] By solving the above matrix equation, the position information of the UAV can be obtained.

[0020] Further, step S8 optimizes the time series model LMM using the objective function FMSE and the chain rule according to the difference between the drone position predicted by the time series model LMM and the reference position obtained by geometric calculation in S4, specifically including:

[0021] Using FMSE as the error metric, compare the difference between the drone position predicted by the timing model LMM and the reference position obtained by geometric calculation in step S4; using the chain rule, starting from the output layer, back propagate layer by layer to calculate the derivative of FMSE with respect to the activation value of each layer, including the activation function and weight of the output layer and the hidden layer, and the derivative is further used to calculate the gradient value of the network weight and bias term, thereby optimizing the model parameters; the whole process is carried out layer by layer from the output layer to the input layer until the entire network parameters are optimized, and the optimized timing model LMM is obtained.

[0022] 6. Further, the objective function FMSE calculation formula is:

[0023]

[0024] Where n is the number of training samples, is the predicted position of the ith sample, X i , Y i , Z i is the corresponding real position;

[0025] The derivative of the final output of FMSE is decomposed into a series of derivatives of intermediate variables with respect to FMSE by using the chain rule. First, starting from the output layer, the derivative of FMSE with respect to the output layer activation value a is calculated. (L) The derivative of is calculated as follows:

[0026]

[0027] Using the chain rule, we first calculate the FMSE for the output layer linear combination Q (L) The derivative and FMSE of the hidden layer linear combination Q (L-1) The derivative of FMSE is used to obtain the activation value a of the hidden layer. (L-1) The derivative of is calculated as follows:

[0028]

[0029] in, is the derivative of the output layer activation function f, is the weight matrix from the hidden layer to the output layer, is the derivative of the hidden layer activation function g;

[0030] Then use these derivatives and the derivatives of the state and gating mechanism inside the model unit to further trace back to the weights and bias terms of the network and calculate the contribution of the weights and bias terms to the MSE, that is, the gradient value, as follows:

[0031]

[0032] in, is the gradient of the output layer activation value, is the gradient of the activation value of the hidden layer; the process is reversed layer by layer until the input layer. When calculating the derivative of the current layer with respect to FMSE, the activation value of the current layer and the gradient of the activation function as well as the derivative information of the next layer are used.

[0033] Furthermore, in step S9, the triangulation method is used to preliminarily estimate the position p0 (m, n, o) of the drone in three-dimensional space, and the calculation formula is as follows:

[0034]

[0035] Compared with the traditional UAV positioning method, the multi-base station UAV positioning method proposed in the present invention has the following characteristics:

[0036] 1. The reliability of the measurement data is ensured by evenly deploying four base stations in three-dimensional space and determining the base station coordinates through multiple measurements;

[0037] 2. Combining multi-base station ranging and fitting algorithms, high-precision positioning of drones in three-dimensional space is achieved, significantly improving the accuracy and robustness of drone positioning;

[0038] 3. The sliding window technology and FMSE are used as error metrics, and the model parameters are optimized through the back propagation algorithm. The trained LMM model can make real-time predictions for new ranging data, and further optimize the position estimation by combining polynomial regression, thus achieving indiscriminate high-precision positioning of UAVs in different environments and further improving the prediction accuracy.

[0039] In general, the present invention not only improves the positioning accuracy of the UAV, but also enhances the robustness of the system, enabling it to work stably under a variety of environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The present invention is a flowchart of a method for constructing a geometric space drone positioning method using multiple base stations based on a fitting algorithm according to an embodiment of the present invention.

[0041] Figure 2 This is a signal transmission interaction diagram of the present invention.

[0042] Figure 3 It is a structural schematic diagram of the LLM model of the present invention.

[0043] Figure 4 This is the main view of the base station deployment of the present invention.

[0044] Figure 5 This is a left view of the base station deployment of the present invention.

[0045] Figure 6 This is a top view of the base station deployment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] like Figure 1 and Figure 2 As shown, the embodiment of the present invention provides a method for positioning a UAV in a geometric space using multiple base stations based on a fitting algorithm, comprising the following steps:

[0048] S1: Determine the coordinates of the four base stations 1, 2, 3, and 4 by measuring (X A ,Y A ,Z A )、(X B ,Y B ,Z B )、(X C ,Y C ,Z C )、(X D ,Y D ,Z D ), the corresponding specific values ​​are (-300cm, 700cm, 150cm), (600cm, 600cm, 160cm), (600cm, -200cm, 160cm), (-200cm, -200cm, 160cm). The base station placement is as follows Figure 4 , 5 , as shown in Figure 6.

[0049] S2: Place the drone at the coordinates (300cm, 200cm, 0). After the take-off command is issued, the drone will fly to the coordinate point (300cm, 200cm, 70cm) and start to fly a circular trajectory according to the preset path, with the coordinate point (200cm, 200cm, 70cm) as the center and a radius of 100cm. The take-off speed is 50cm / s and the maximum flight speed is 100cm / s.

[0050] S3: During the flight of the drone, the drone and each base station transmit data within a very wide spectrum by sending extremely short pulse signals to each other; a signal receiver and recorder is used to record the time t when the drone sends the signal to the base station. A 1. t B 1. t C 1. tD 1. The receiving time t when the base station receives the signal A 2. t B 2. t C 2. t D 2. The time t at which the base station sends a response signal to the drone A 3. t B 3. t C 3. t D 3. The receiving time t when the drone receives the response signal A 4. t B 4. t C 4. t D 4 and the flight altitude, flight speed, and flight acceleration of the UAV at the corresponding time point.

[0051] S4: Calculate the round-trip time RT of the signal between each base station and the drone based on the data recorded by the signal receiving and recorder in S3 and the signal propagation speed A , RT B , RT C , RT D , one-way time ST A , ST B , ST C , ST D and the distance S between the drone and each base station A , S B , S C , S D , where the signal propagation speed is the speed of light c, and the specific calculation formula is as follows:

[0052]

[0053] Calculate the drone position p(X,Y,Z) using the following formula:

[0054]

[0055] By solving the above matrix equation, the position information of the UAV at a certain moment can be obtained.

[0056] S5: Design a time series model LMM to learn the mapping relationship between input features and drone positions. The model structure is as follows: Figure 3 As shown in Figure 2, the time series model LMM utilizes a deep neural network structure based on the principle of time recurrence, which can process input data in time order and capture the time dependency in the data by maintaining an internal state.

[0057] S6: Collect the sending and receiving times recorded in S3 and the flight altitude, flight speed, and flight acceleration of the drone at the corresponding time points, the drone location information calculated in S4, the distance between the base station and the drone, the unique identifier used to distinguish different base stations, and the location coordinates of the base stations as a training set, sort the training set in chronological order, and use the sliding window technology, that is, the window slides on the data with a certain step size, to generate multiple overlapping or non-overlapping input sequences. Each input sequence is a two-dimensional array with a shape of [sequence length, number of features], where each row represents the data at a time point.

[0058] S7: Input the input sequence generated in S6 into the time series model LMM. The input layer of the model receives these fixed-length window data sequences consisting of ranging data at multiple time points. The fixed-length window data sequence includes the distance measurement values ​​between the drone and each base station, flight altitude, flight speed, acceleration and other data. The time series model LMM processes each element in time step order and updates the internal state, including cell state and hidden state, to capture the time dependency of the time series. The cell state is responsible for long-term memory, and the hidden state is responsible for short-term memory. At each time step, the model combines the current ranging data and internal state to calculate the output, which will be used as the input for the next time step. After processing the entire sequence, the output layer of the LMM model integrates the information of all time steps to generate a prediction of the position of the drone at a specific time point in the sequence.

[0059] S8: Based on the difference between the UAV position predicted by the timing model LMM and the reference position obtained by geometric calculation in S4, the timing model LMM is optimized using the objective function FMSE and the chain rule.

[0060] Specifically, the objective function FMSE is used to measure the difference between the drone position predicted by the comparison model and the reference position obtained by geometric calculation in S4 to optimize the prediction result in S7. The calculation formula is:

[0061]

[0062] Where n is the number of training samples, is the predicted position of the ith sample, X i , Y i , Z i is the corresponding real position.

[0063] The chain rule is used to decompose the derivative of the final output of FMSE into a series of intermediate variables (including the activation values ​​of each layer in LLM, the state of the hidden layer, and the output of the gating mechanism, etc.) with respect to FMSE. First, starting from the output layer, calculate the derivative of FMSE with respect to the output layer activation value a (L) The derivative of is calculated as follows:

[0064]

[0065] Using the chain rule, we first calculate the FMSE for the output layer linear combination Q (L) The derivative and FMSE of the hidden layer linear combination Q (L-1) The derivative of FMSE can be used to obtain the activation value a of the hidden layer. (L-1) The derivative of is calculated as follows:

[0066]

[0067] in, is the derivative of the output layer activation function f, is the weight matrix from the hidden layer to the output layer, is the derivative of the hidden layer activation function g.

[0068] Then use these derivatives and the derivatives of the state and gating mechanism inside the model unit to further trace back to the weights and bias terms of the network and calculate their contribution to the MSE, that is, the gradient value, as follows:

[0069]

[0070] in, is the gradient of the output layer activation value, is the gradient of the activation value of the hidden layer. This process is reversed layer by layer until the input layer. When calculating the derivative of the current layer with respect to FMSE, the activation value of the current layer and the gradient of the activation function as well as the derivative information of the next layer are used.

[0071] S9: Based on the time series model LMM optimized in S8, the trained time series model LMM is used to predict the new ranging data to generate the real-time position estimate of the drone. Then, these real-time position estimates are combined with the known coordinates of the base station given in S1, and the position p0(m,n,o) of the drone is preliminarily estimated in three-dimensional space using triangulation. The calculation formula is as follows:

[0072]

[0073] After obtaining the preliminary estimated position p0 of the drone, polynomial regression is used to optimize p0(m,n,o) to improve the accuracy of position estimation. Polynomial regression optimization mainly includes the definition of the quadratic polynomial model and the calculation of the error function E(a,b,c) for the polynomial coefficient a i 、b i 、c i The gradient has two parts, the specific steps are as follows:

[0074] (1) Define a quadratic polynomial model where x(t), y(t), and z(t) represent the estimated position of the drone, and t is the time step. Then, the coefficients of the polynomial are optimized by minimizing the error function E(a, b, c). The formula is as follows:

[0075]

[0076] (2) Calculate the error function E(a, b, c) for the polynomial coefficient a i 、b i 、c i The gradient of the coefficient a is adjusted using the gradient descent method i 、b i 、c i , so that the error function E(a,b,c) is minimized, the formula is as follows:

[0077]

[0078] The multi-base station UAV positioning method proposed in the present invention ensures the reliability of measurement data by evenly deploying four base stations in three-dimensional space and determining the base station coordinates through multiple measurements; combines multi-base station ranging and fitting algorithms to achieve high-precision positioning of the UAV in three-dimensional space, and significantly improves the accuracy and robustness of UAV positioning; the present invention adopts sliding window technology and FMSE as error metrics, optimizes model parameters through back propagation algorithm, and further improves the prediction accuracy; finally, the trained LMM model can make real-time predictions on new ranging data, and further optimizes position estimation by combining polynomial regression, thereby achieving indiscriminate high-precision positioning of the UAV in different environments.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for positioning unmanned aerial vehicles in geometric space using multiple base stations based on a fitting algorithm, characterized in that: The steps include: Step S1: Deploy four base stations with uniform positions in a three-dimensional space, and determine the coordinate position A (X A ,Y A ,Z A )、B(X B ,Y B ,Z B )、C(X C ,Y C ,Z C )、D(X D ,Y D ,Z D ); Step S2: Design the flight path, flight altitude and flight speed of the UAV, and plan the flight trajectory of the UAV according to the actual application scenario and requirements; Step S3: During the flight of the drone, the drone and each base station transmit data by sending pulse signals to each other; a signal receiver and recorder is used to record the sending time t of the drone sending the signal to the base station. A 1. t B 1. t C 1. t D 1. The receiving time t when the base station receives the signal A 2. t B 2. t C 2. t D 2. The time t at which the base station sends a response signal to the drone A 3. t B 3. t C 3. t D 3. The receiving time t when the drone receives the response signal A 4. t B 4. t C 4. t D 4 and the flight altitude, flight speed, and flight acceleration of the drone at the corresponding time point; Step S4: Calculate the round-trip time RT of the signal between each base station and the drone based on the data received and recorded by the recorder in step S3 and the signal propagation speed. A , RT B , RT C , RT D , one-way time ST A , ST B , ST C , ST D and the distance S between the drone and each base station A , S B , S C , S D , where the signal propagation speed is the speed of light c, and the position of the drone is calculated based on the coordinate position of each base station and the distance between the drone and each base station; Step S5: designing a time series model LMM for learning the mapping relationship between input features and drone positions. The time series model LMM utilizes a deep neural network structure based on the time recursion principle, can process input data in time sequence, and captures the time dependency in the data by maintaining an internal state. Step S6: Collect the sending and receiving times recorded in step S3 and the flight altitude, flight speed, and flight acceleration of the drone at the corresponding time points, the location information of the drone calculated in S4, the distance between the base station and the drone, and the unique identifier used to distinguish different base stations and the location coordinates of the base stations as a training set, sort the training set in chronological order, and then use the sliding window technology to construct the input sequence. Step S7: Apply the input sequence generated in step S6 to the time series model LMM, and predict the position of the drone through the time series model LMM; Step S8: According to the difference between the UAV position predicted by the time series model LMM and the reference position obtained by geometric calculation in S4, the time series model LMM is optimized using the objective function FMSE and the chain rule; Step S9: Use the optimized time series model LMM to predict the new ranging data and generate the real-time position estimate of the UAV; then, combine the real-time position estimate with the known coordinates of the base station and use triangulation to preliminarily estimate the position p0(m,n,o) of the UAV in three-dimensional space; After obtaining the preliminary estimated UAV position p0, polynomial regression is used to optimize p0(m,n,o) to improve the accuracy of position estimation.

2. The method for unmanned aerial vehicle positioning using a multi-base station geometric space constructed based on a fitting algorithm as claimed in claim 1, characterized in that: In step S4, the round-trip time RT of the signal between each base station and the drone is calculated based on the data received by the signal in step S3 and recorded by the recorder as well as the signal propagation speed. A , RT B , RT C , RT D , one-way time ST A , ST B , ST C , ST D and the distance S between the drone and each base station A , S B , S C , S D , the calculation formula is as follows: Calculate the drone position p(X,Y,Z) using the following formula: By solving the above matrix equation, the position information of the UAV can be obtained.

3. The method for positioning unmanned aerial vehicles in geometric space constructed by multiple base stations based on a fitting algorithm as claimed in claim 1, characterized in that: Step S8 optimizes the time series model LMM using the objective function FMSE and the chain rule according to the difference between the UAV position predicted by the time series model LMM and the reference position obtained by geometric calculation in S4, specifically including: Using FMSE as the error metric, compare the difference between the drone position predicted by the timing model LMM and the reference position obtained by geometric calculation in step S4; using the chain rule, starting from the output layer, back propagate layer by layer to calculate the derivative of FMSE with respect to the activation value of each layer, including the activation function and weight of the output layer and the hidden layer, and the derivative is further used to calculate the gradient value of the network weight and bias term, thereby optimizing the model parameters; the whole process is carried out layer by layer from the output layer to the input layer until the entire network parameters are optimized, and the optimized timing model LMM is obtained.

4. The method for positioning unmanned aerial vehicles in geometric space constructed by multiple base stations based on a fitting algorithm as claimed in claim 3, characterized in that: The objective function FMSE calculation formula is: Where n is the number of training samples, is the predicted position of the ith sample, X i , Y i , Z i is the corresponding real position; The derivative of the final output of FMSE is decomposed into a series of derivatives of intermediate variables with respect to FMSE by using the chain rule. First, starting from the output layer, the derivative of FMSE with respect to the output layer activation value a is calculated. (L) The derivative of is calculated as follows: Using the chain rule, we first calculate the FMSE for the output layer linear combination Q (L) The derivative and FMSE of the hidden layer linear combination Q (L-1) The derivative of FMSE is used to obtain the activation value a of the hidden layer. (L-1) The derivative of is calculated as follows: in, is the derivative of the output layer activation function f, is the weight matrix from the hidden layer to the output layer, is the derivative of the hidden layer activation function g; Then use these derivatives and the derivatives of the state and gating mechanism inside the model unit to further trace back to the weights and bias terms of the network and calculate the contribution of the weights and bias terms to the MSE, that is, the gradient value, as follows: in, is the gradient of the output layer activation value, is the gradient of the activation value of the hidden layer; the process is reversed layer by layer until the input layer. When calculating the derivative of the current layer with respect to FMSE, the activation value of the current layer and the gradient of the activation function as well as the derivative information of the next layer are used.

5. The method for positioning unmanned aerial vehicle using geometric space constructed by multiple base stations based on a fitting algorithm as claimed in claim 2, characterized in that: In step S9, the triangulation method is used to preliminarily estimate the position p0 (m, n, o) of the drone in three-dimensional space, and the calculation formula is as follows:

Citation Information

Cited By

  • Linear base station array positioning method based on geometric modeling

    CN119854934A