Low-complexity three-dimensional AOA moving target positioning method capable of reducing deviation
By constructing a weighted least squares problem and simplified model in the three-dimensional AOA positioning method, the problem of low positioning accuracy under the influence of unknown sensor position and noise is solved, and efficient and robust mobile target positioning is achieved, which is suitable for complex scenarios.
Patent Information
- Application Number
- CN202411105879.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing AOA positioning method has low positioning accuracy and high computational complexity when the sensor location is unknown or the measurement noise is large, and has high requirements for the geometric distribution of the positioning scene, making it difficult to achieve effective without prior information. moving target positioning.
Using the three-dimensional AOA moving target positioning method, by establishing a reference coordinate system, building a weighted least squares problem, introducing auxiliary variables and Lagrangian functions, using angle measurement information to simplify the distance measurement model, reduce deviations and improve calculation efficiency, suitable for sensors The location part is known.
Without prior information, the positioning accuracy is significantly improved, the calculation complexity is reduced, the robustness of measurement noise is enhanced, and the good positioning performance is maintained in any positioning scenario, which can achieve Clariano-Royal Positioning accuracy of the lower bound.
Smart Images

Figure CN118938129A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a target positioning technology, and in particular to a three-dimensional AOA (Angle-of-Arrival) mobile target positioning method with a partially known sensor position and the ability to reduce deviation. Background Art
[0002] In the past few decades, target positioning technology has been widely used in many fields, including but not limited to target tracking, autonomous driving, search and rescue, and wireless communication. Among them, the Angle of Arrival (AOA) positioning method has become a hot research topic in the field of target positioning due to its low price and high positioning accuracy.
[0003] The existing AOA positioning method requires the model to be converted. However, a large amount of deviation is usually introduced during the conversion process, which will have a significant impact on the estimation of the position and velocity of the moving target. Existing research methods that consider reducing deviations usually use the maximum likelihood method or the semi-definite relaxation method to estimate the target position. However, the maximum likelihood method usually requires prior knowledge and will fall into a local solution when the variable dimension is large and the initial solution is inaccurate; while the semi-definite relaxation method often requires high computational complexity and is not applicable in practical engineering fields.
[0004] In existing research, it is usually considered that the positions of all sensors are known, which is applicable to fixed ground stations or situations where the positions can be obtained through technologies such as GPS. However, in practical application scenarios such as indoors or underwater, the lack of GPS signals may result in unknown position information of some sensors. Therefore, the positioning problem of unknown positions of some sensors deserves further in-depth exploration. Existing literature rarely studies the method of using AOA for positioning when the positions of some sensors are unknown. The AOA positioning method when all sensors are known is very sensitive to measurement noise and has high requirements on the geometric distribution of the positioning scene. Summary of the invention
[0005] The technical problem to be solved by the present invention is to provide a three-dimensional AOA mobile target positioning method with the ability to reduce deviation when the sensor position is partially known. Without the need for any prior information, the method uses the AOA measurement information to jointly locate the sensor and the mobile target at the unknown position when the sensor position is partially known. It has no requirements for the geometric distribution of the positioning scene and exhibits good robustness to measurement noise.
[0006] The technical solution adopted by the present invention to solve the above technical problems is: a low-complexity three-dimensional AOA moving target positioning method with the ability to reduce deviation, characterized by comprising the following steps:
[0007] Step 1: In the sensor network, establish a three-dimensional coordinate system as the reference coordinate system, and assume that there are J sensors and a moving target moving in a uniform linear motion. The positions of the 1st sensor to the Nth sensor are known, and the positions of the N+1th sensor to the Jth sensor are unknown. The initial position and moving speed of the moving target are unknown. The moving target is assumed to move for a total of K moments. The true position of any sensor with a known position in the reference coordinate system is set as the origin of the reference coordinate system. When 1≤i≤N, the true position of the sensor with the i-th known position in the reference coordinate system is recorded as s i ,s i =[s x,i ,s y,i ,s z,i ] T , when N+1≤i≤J, the true position of the sensor at the i-th unknown position in the reference coordinate system is recorded as The actual initial position of the moving target in the reference coordinate system is recorded as The actual moving speed of the moving target is recorded as v o , The actual position of the moving target in the reference coordinate system at the kth moment is recorded as Where K>1, i=1,...,J,s x,i ,s y,i ,s z,i Corresponding representation s i The horizontal coordinate, vertical coordinate, and vertical coordinate of , the superscript "T" indicates the transpose of a vector or matrix, Corresponding representation The horizontal, vertical and vertical coordinates of Corresponding representation The horizontal, vertical and vertical coordinates of Corresponding representation v o The velocity component along the x-axis of the reference coordinate system, the velocity component along the y-axis of the reference coordinate system, and the velocity component along the z-axis of the reference coordinate system, k = 1, ..., K, Indicates a time interval;
[0008] Step 2: In the sensor network, the signal transmitted by the mobile target at the kth time is received by the i-th sensor, and the i-th sensor obtains the AOA measurement value of the arrival signal contaminated by noise. The AOA measurement value includes the azimuth measurement value and the pitch angle measurement value, which are denoted by θ k,i and φ k,i ; Wherein, k=1,...,K, i=1,...,J;
[0009] Step 3: Construct the AOA measurement model of the arrival signal obtained after the signal emitted by the mobile target is received by the sensor at a known position, which is described as:
[0010]
[0011] And construct the AOA measurement model of the arrival signal obtained after the signal emitted by the mobile target is received by the sensor at the unknown position, which is described as:
[0012]
[0013] Among them, θ k,i That is, the azimuth measurement value of the arrival signal contaminated by noise obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor. It represents the true value of the azimuth of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. represents the azimuth measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor, φ k,i That is, the pitch angle measurement value of the arrival signal contaminated by noise obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. It represents the true value of the pitch angle of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. represents the pitch angle measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor. When 1≤i≤N, r k,i represents the two-dimensional distance between the moving target and the sensor at the i-th known position on the xoy plane of the reference coordinate system, When N+1≤i≤J represents the two-dimensional distance between the moving target and the sensor at the i-th unknown position on the xoy plane of the reference coordinate system,
[0014] Step 4: Preprocess equation (1): Ignore the noise terms in the two equations of equation (1) and take the tangent to obtain Then cross-multiply both sides of the two equations in formula (1.1) and sort them out to get
[0015] According to Perform a first-order Taylor expansion on the sine and cosine terms in the two equations of formula (1.2), and we get
[0016] Preprocess equation (2): Ignore the noise terms in the two equations of equation (2) and take the tangent to obtain Then cross-multiply both sides of the two equations in formula (2.1) and sort them out to get According to Perform a first-order Taylor expansion on the sine and cosine terms in the two equations of formula (2.2), and we get
[0018] Step 5: Combine equations (1.3) and (2.3) to construct a weighted least squares problem for joint positioning with partially known sensor locations, which can be described as:
[0019]
[0020] in, is the optimization variable of the weighted least squares problem, represents the initial position variable of the moving target, represents the moving speed variable of the moving target, represents the position variable of the sensor with the i-th unknown position, is the objective function of the weighted least squares problem, is the coefficient matrix, The kth sub-matrix in is The i-th row vector of The i-th row vector of 0 N×3M represents an all-zero matrix of dimension N×3M, where N represents the number of sensors with known positions, and M represents the number of sensors with unknown positions, M=JN, blkdiag() is a function for finding block diagonal matrices. The iNth block diagonal submatrix in is The iNth block diagonal submatrix in is is the introduced coefficient vector, The kth column vector of 0 1×M represents a vector of all zeros with a dimension of 1×M, The i-th element of The i-th element of is the introduced weight matrix, It means to find the mathematical expectation. is the introduced coefficient matrix, The kth block diagonal submatrix in is diag() is a function for finding diagonal matrices, || || is a binary norm operator, represents the measurement noise vector, The kth subvector of is Subject to the mean of 0 and the covariance matrix is Gaussian distribution, express The covariance matrix of The covariance matrix of
[0021] Step 6: Introduce new auxiliary variables into the weighted least squares problem constructed in step 5 Convert the objective function of the weighted least squares problem constructed in step 5 to Then define and Will Convert to Then Split into a matrix constructed from true values and errors, namely: Finally Introduction In, we get: in, θ k,i and φ k,i Replace with and After getting The kth sub-matrix in is The i-th row vector of The i-th row vector of The iNth block diagonal submatrix in is The iN block diagonal submatrices in are θ k,i and φ k,i Replace with and After getting The kth column vector of The i-th element of The i-th element of The i-th row vector of The i-th row vector of The i-th row vector of The iNth block diagonal submatrix in is The iNth block diagonal submatrix in is The iNth block diagonal submatrix in is V=blkdiag(V1,...,V K ), the kth block diagonal submatrix in V is V k , ⊙ is the Hadamard product operator symbol, express The submatrix consisting of the 1st to the Jth row and the 1st to the Jth column in , express The submatrix consisting of the J+1th row to the 2Jth row and the J+1th column to the 2Jth column in , express The submatrix consisting of the 1st to Jth rows and the J+1th to 2Jth columns in , Depend on get, express The kth block diagonal submatrix in , The Kth column vector of The i-th element of The i-th element of The i-th element of
[0022] Step 7: Based on step 6, a constrained weighted least squares problem with the ability to reduce bias is constructed when the sensor position is partially known, which is described as:
[0023]
[0024] in, is the optimization variable of the constrained weighted least squares problem, The dimension is (7+3M)×1, represents a constant;
[0025] Step 8: Construct the Lagrangian function for the constrained weighted least squares problem constructed in step 7, expressed as: Then, according to the Lagrangian function, the KKT condition is used to obtain The optimal solution Formula Again Using eigenvalue decomposition, we can decompose Finally, according to Get the initial position variable of the moving target Estimated value of Moving target's moving speed variable Estimated value of Position variable of the sensor with unknown position i Estimated value of Where L() represents the Lagrangian function, is the Lagrange multiplier, express The vector consisting of the first to third elements of express The 7+3Mth element of express The vector consisting of the 4th to 6th elements of express The vector consisting of the 4+3(iN)th element to the 6+3(iN)th element of .
[0026] Compared with the prior art, the advantages of the present invention are:
[0027] 1) In existing research, the location of sensor deployment is usually assumed to be known. However, in environments where GPS signals cannot cover, such as underwater, underground or indoor positioning scenarios, the location information of the sensor may not be available. It is worth noting that the acquisition of AOA measurement information does not require the exact absolute position of the sensor, which opens up new possibilities for future research.
[0028] 2) Traditional AOA positioning methods often produce large deviations during the model conversion process. The method of the present invention successfully reduces the deviation by introducing new auxiliary variables in the weighted least squares problem of joint positioning with partially known sensor positions, and considers the second-order deviation as a constraint condition, thereby significantly improving the positioning accuracy.
[0029] 3) The method of the present invention cleverly uses angle measurement information to construct the pitch angle without requiring any prior information, thereby greatly simplifying the originally complex and nonlinear distance measurement model. This method not only reduces the number of coupled variables, but also can obtain a closed-form solution by constructing Lagrangian functions and eigenvalue decomposition methods, with low computational complexity.
[0030] 4) The existing mobile target positioning method has high requirements on the geometric distribution of the positioning scene. When the geometric distribution of the positioning scene is not conducive to the estimation accuracy, the positioning performance of the method drops sharply. The method of the present invention can show good positioning performance for the geometric distribution of the positioning scene in any positioning scene.
[0031] 5) The experimental results of the method of the present invention show that it can achieve the positioning accuracy of the Cramer-Rao lower bound (CRLB) and show stable performance. It shows excellent robustness to measurement noise, and it can still show robust positioning performance even when the measurement noise power is large. In addition, it is worth mentioning that even if the number of sensors with unknown positions increases, the performance of the method of the present invention will not be negatively affected, but may further improve the positioning accuracy of existing mobile targets. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 It is a general implementation block diagram of the method of the present invention;
[0033] Figure 2 The mean square error (MSE) of the initial position estimation of the mobile target by the method of the present invention and the Cramer-Rao lower bound (CRLB) when N=3, M=2, K=20 varies with the average power of the AOA measurement noise σ 2 Comparative graph of changes;
[0034] Figure 3 Here is the mean square error (MSE) of the moving target speed estimation using the proposed method and the Cramer-Rao lower bound (CRLB) when N=3, M=2, and K=20 as the AOA measurement noise average power σ 2 Comparative graph of changes;
[0035] Figure 4 Here is the mean square error (MSE) of the position estimation of the unknown sensor by the method of the present invention and the Cramer-Rao lower bound (CRLB) when N=3, M=2, K=20 as the AOA measurement noise average power σ 2 Comparative graph of changes;
[0036] Figure 5 The deviation performance (Bias) of the initial position estimation of the mobile target by the method of the present invention and the theoretical deviation (Theoretical) when N = 3, M = 2, K = 20 varies with the average power σ of the AOA measurement noise 2 Comparative graph of changes;
[0037] Figure 6 The deviation performance (Bias) of the proposed method and the theoretical deviation (Theoretical) for the moving target moving speed estimation under the condition of N = 3, M = 2, K = 20 varies with the AOA measurement noise average power σ 2 Comparative graph of changes;
[0038] Figure 7 The deviation performance (Bias) of the position estimation of the sensor at an unknown position by the method of the present invention and the theoretical deviation (Theoretical) is plotted against the average power σ of the AOA measurement noise when N=3, M=2, and K=20. 2 Comparison chart of changes. DETAILED DESCRIPTION
[0039] The present invention is further described in detail below with reference to the accompanying drawings.
[0040] The present invention proposes a three-dimensional AOA mobile target positioning method with a partially known sensor position and the ability to reduce deviation. The overall implementation block diagram is as follows: Figure 1 As shown, it includes the following steps:
[0041] Step 1: In the sensor network, establish a three-dimensional coordinate system as the reference coordinate system, and assume that there are J sensors and a moving target moving in a uniform linear motion. The positions of the 1st sensor to the Nth sensor are known, and the positions of the N+1th sensor to the Jth sensor are unknown. The initial position and moving speed of the moving target are unknown. The moving target is assumed to move for a total of K moments. The true position of any sensor with a known position in the reference coordinate system is set as the origin of the reference coordinate system. When 1≤i≤N, the true position of the sensor with the i-th known position in the reference coordinate system is recorded as s i ,s i =[s x,i ,s y,i ,s z,i ] T , when N+1≤i≤J, the true position of the sensor at the i-th unknown position in the reference coordinate system is recorded as The actual initial position of the moving target in the reference coordinate system is recorded as The actual moving speed of the moving target is recorded as v o , The actual position of the moving target in the reference coordinate system at the kth moment is recorded as Among them, the number of sensors with known positions is greater than 1, the number of sensors with unknown positions is greater than or equal to 1, K>1, i=1,…,J,s x,i ,s y,i ,s z,i Corresponding representation s i The horizontal coordinate, vertical coordinate, and vertical coordinate of , the superscript "T" indicates the transpose of a vector or matrix, Corresponding representation The horizontal, vertical and vertical coordinates of Corresponding representation The horizontal, vertical and vertical coordinates of Corresponding representation v o The velocity component along the x-axis of the reference coordinate system, the velocity component along the y-axis of the reference coordinate system, and the velocity component along the z-axis of the reference coordinate system, k = 1, ..., K, Indicates the time interval, such as Second.
[0042] Step 2: In the sensor network, the signal transmitted by the mobile target at the kth time is received by the i-th sensor. The i-th sensor obtains the AOA measurement value of the arrival signal contaminated by noise through the antenna array. The AOA measurement value includes the azimuth measurement value and the pitch angle measurement value, which are denoted by θ k,i and φ k,i; Wherein, k=1,...,K, i=1,...,J, when 1≤i≤N, the i-th sensor is a sensor with a known position, and when N+1≤i≤J, the i-th sensor is a sensor with an unknown position.
[0043] Step 3: Construct the AOA measurement model of the arrival signal obtained after the signal emitted by the mobile target is received by the sensor at a known position, which is described as:
[0044]
[0045] And construct the AOA measurement model of the arrival signal obtained after the signal emitted by the mobile target is received by the sensor at the unknown position, which is described as:
[0046]
[0047] Among them, θ k,i That is, the azimuth measurement value of the arrival signal contaminated by noise obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor. It represents the true value of the azimuth of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. represents the azimuth measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor, φ k,i That is, the pitch angle measurement value of the arrival signal contaminated by noise obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. It represents the true value of the pitch angle of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. represents the pitch angle measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor. When 1≤i≤N, r k,i represents the two-dimensional distance between the moving target and the sensor at the i-th known position on the xoy plane of the reference coordinate system, When N+1≤i≤J represents the two-dimensional distance between the moving target and the sensor at the i-th unknown position on the xoy plane of the reference coordinate system,
[0048] Step 4: Preprocess equation (1): Ignore the noise terms in the two equations of equation (1) and take the tangent to obtain Then cross-multiply both sides of the two equations in formula (1.1) and sort them out to get
[0049] According to Perform a first-order Taylor expansion on the sine and cosine terms in the two equations of formula (1.2), and we get
[0050] Preprocess equation (2): Ignore the noise terms in the two equations of equation (2) and take the tangent to obtain Then cross-multiply both sides of the two equations in formula (2.1) and sort them out to get According to Perform a first-order Taylor expansion on the sine and cosine terms in the two equations of formula (2.2), and we get
[0051] Step 5: Combine equations (1.3) and (2.3) to construct a weighted least squares problem for joint positioning with partially known sensor locations, which can be described as:
[0052]
[0053] Among them, min is the minimum value function, is the optimization variable of the weighted least squares problem, represents the initial position variable of the moving target, represents the moving speed variable of the moving target, The position variable representing the sensor with the N+1th unknown position, represents the position variable of the sensor with the Jth unknown position, represents the position variable of the sensor with the i-th unknown position, is the objective function of the weighted least squares problem, is the coefficient matrix, arranged by sampling time, for The first sub-matrix in for The Kth submatrix in , The kth sub-matrix in is The i-th row vector of The i-th row vector of 0 N×3M represents an all-zero matrix of dimension N×3M, where N represents the number of sensors with known positions, and M represents the number of sensors with unknown positions, M=JN, blkdiag() is a function for finding block diagonal matrices. express The first block diagonal submatrix in , express The M=JNth block diagonal submatrix in, The iNth block diagonal submatrix in is express The first block diagonal submatrix in , express The M=JNth block diagonal submatrix in, The iNth block diagonal submatrix in is is the introduced coefficient vector, arranged according to the sampling time, express The first column vector of express The Kth column vector of The kth column vector of 0 1×M represents a vector of all zeros with a dimension of 1×M, The i-th element of The i-th element of is the introduced weight matrix, It means to find the mathematical expectation. is the introduced coefficient matrix, express The first block diagonal submatrix in , express The Kth block diagonal submatrix in , The kth block diagonal submatrix in is diag() is a function for finding diagonal matrices, || || is the binary norm operator, r k,1 Represents the two-dimensional distance between the moving target and the first known position of the sensor on the xoy plane of the reference coordinate system, r k,N Represents the two-dimensional distance between the moving target and the sensor at the Nth known position on the xoy plane of the reference coordinate system, Represents the two-dimensional distance between the moving target and the sensor at the N+1th unknown position on the xoy plane of the reference coordinate system, represents the two-dimensional distance between the moving target and the sensor at the Jth unknown position on the xoy plane of the reference coordinate system, s1 represents the true position of the sensor at the first known position in the reference coordinate system, and s N represents the true position of the Nth known position sensor in the reference coordinate system, Represents the true position of the sensor with the N+1th unknown position in the reference coordinate system, represents the true position of the sensor with the Jth unknown position in the reference coordinate system, represents the measurement noise vector, express The first subvector of express The Kth subvector of The kth subvector of is It represents the azimuth measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the first sensor, It represents the azimuth measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the Jth sensor, It represents the pitch angle measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the first sensor, It represents the pitch angle measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the Jth sensor, Subject to the mean of 0 and the covariance matrix is Gaussian distribution, express The covariance matrix of express The covariance matrix of express The covariance matrix of The covariance matrix of
[0054] Step 6: Introduce new auxiliary variables into the weighted least squares problem constructed in step 5 Convert the objective function of the weighted least squares problem constructed in step 5 to Then define and Will Convert to Then Split into a matrix constructed from true values and errors, namely: Finally Introduction In, we get: in, θ k,i and φ k,i Replace with and After getting express The first sub-matrix in express The Kth submatrix in , The kth sub-matrix in is The i-th row vector of The i-th row vector of express The first block diagonal submatrix in , express The M=JNth block diagonal submatrix in, The iNth block diagonal submatrix in is express The first block diagonal submatrix in , express The M=JNth block diagonal submatrix in, The iN block diagonal submatrices in are θ k,i and φ k,i Replace with and After getting express The first column vector of express The Kth column vector of The kth column vector of The i-th element of The i-th element of is the second-order deviation, The i-th row vector of The i-th row vector of The i-th row vector of express The first block diagonal submatrix in , express The M=JNth block diagonal submatrix in, The iNth block diagonal submatrix in is express The first block diagonal submatrix in , express The M=JNth block diagonal submatrix in, The iNth block diagonal submatrix in is express The first block diagonal submatrix in , express The M=JNth block diagonal submatrix in, The iNth block diagonal submatrix in is V=blkdiag(V1,…,V K ), V1 represents the first block diagonal submatrix in V, V K represents the Kth block diagonal submatrix in V, and the kth block diagonal submatrix in V is V k , ⊙ is the Hadamard product operator symbol, express The submatrix consisting of the 1st to the Jth row and the 1st to the Jth column in , express The submatrix consisting of the J+1th row to the 2Jth row and the J+1th column to the 2Jth column in , express The submatrix consisting of the 1st to Jth rows and the J+1th to 2Jth columns in , Depend on get, express The kth block diagonal submatrix in , express The first block diagonal submatrix in , express The Kth block diagonal submatrix in , express The first column vector of express The Kth column vector of The Kth column vector of The i-th element of The i-th element of The i-th element of
[0055] Step 7: Based on step 6, a constrained weighted least squares problem with the ability to reduce bias is constructed when the sensor position is partially known, which is described as:
[0056]
[0057] Among them, st means "subject to...", is the optimization variable of the constrained weighted least squares problem, The dimension is (7+3M)×1, Represents a constant.
[0058] Step 8: Construct the Lagrangian function for the constrained weighted least squares problem constructed in step 7, expressed as: Then, according to the Lagrangian function, the KKT condition is used to obtain The optimal solution Formula Again Using eigenvalue decomposition, we can decompose Finally, according to Get the initial position variable of the moving target Estimated value of Moving target's moving speed variable Estimated value of Position variable of the sensor with unknown position i Estimated value of Where L() represents the Lagrangian function, is the Lagrange multiplier, express The vector consisting of the first to third elements of express The 7+3Mth element of express The vector consisting of the 4th to 6th elements of express The vector consisting of the 4+3(iN)th element to the 6+3(iN)th element of .
[0059] In order to verify the feasibility and effectiveness of the method of the present invention, a simulation experiment was carried out on the method of the present invention.
[0060] Assume that there are 3 sensors with known positions, 2 sensors with unknown positions, and 1 moving target in a 3D space, i.e., a 3D coordinate system. The actual Cartesian coordinate positions of the sensors in the 3D space are randomly generated between [-500,500] 3 m 3 In the three-dimensional space, the moving target is generated at [-700,700] 3 m 3 In the three-dimensional space, its moving speed is [-20,20] 3 (m / s) 3 In order to prevent the influence of geometry, the minimum distance between the sensor and the mobile target is assumed to be 150m. The sampling interval of the AOA measurement value is 1s. For the AOA measurement noise, it is assumed that the azimuth measurement noise of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor is considered to be The power design is The pitch angle measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor The power design is Its relevance is Among them, k is the scaling factor for the kth moment,
[0061] diag() is a function for finding diagonal matrices. It represents the true value of the pitch angle of the arrival signal obtained after the signal emitted by the mobile target at the kth moment is received by the first sensor. It represents the true value of the pitch angle of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the Jth sensor, σ 2 represents the variance of AOA measurement noise, and the average power of AOA measurement noise σ 2 satisfy tr() means finding the trace of the matrix. 20 sampling moments are used, and the sampling interval is 1s.
[0062] Based on the above parameter settings, simulation experiments tested the performance of the method of the present invention under different conditions of the average power of the AOA measurement noise when the sensor position is partially known, and the performance of the method under different conditions of the number of sensors with unknown positions based on the use of 3 sensors with known positions. The mean square error (MSE) and bias performance (Bias) describing the positioning performance in both cases are the MSE and Bias obtained by running 1000 Monte Carlo tests in 10 random scenarios for each scenario. At the same time, the performance benchmark Cramer-Rao lower bound (CRLB) and theoretical bias (Theoretical) are introduced for comparison.
[0063] Figure 2 The mean square error (MSE) of the initial position estimation of the moving target by the proposed method and the Cramer-Rao lower bound (CRLB) when N=3, M=2, K=20 is given along with the AOA measurement noise average power σ 2 Comparison chart of changes, Figure 3 The mean square error (MSE) of the moving target speed estimation using the proposed method and the Cramer-Rao lower bound (CRLB) when N=3, M=2, K=20 is given along with the AOA measurement noise average power σ 2 Comparison chart of changes, Figure 4 The mean square error (MSE) of the position estimation of the unknown sensor with the AOA measurement noise average power σ is given when the proposed method and the Cramer-Rao lower bound (CRLB) are used in the case of N = 3, M = 2, K = 20. 2 When the sensor position is partially known, Figure 2 , Figure 3 and Figure 4 It can be seen that when the AOA measures the average noise power σ 2 When the average power of AOA measurement noise changes, the method of the present invention can achieve the positioning accuracy of the Cramer-Rao lower bound when the average power of AOA measurement noise is small and medium.
[0064] Figure 5 The bias performance (Bias) of the proposed method and the theoretical deviation (Theoretical) for the initial position estimation of the moving target is given when N = 3, M = 2, K = 20 and the average power of the AOA measurement noise σ 2 Comparison chart of changes, Figure 6 The deviation performance (Bias) of the proposed method and the theoretical deviation (Theoretical) for the moving target speed estimation under the condition of N = 3, M = 2, K = 20 is given as the average power σ of the AOA measurement noise 2Comparison chart of changes, Figure 7 The bias performance (Bias) of the position estimation of the unknown position sensor of the proposed method and the theoretical deviation (Theoretical) is given with the AOA measurement noise average power σ 2 When the sensor position is partially known, Figure 5 , Figure 6 and Figure 7 It can be seen that when the AOA measures the average noise power σ 2 When the AOA measurement noise average power is small or medium, the positioning accuracy of the method of the present invention can achieve the theoretical deviation.
Claims
1. A low-complexity three-dimensional AOA moving target positioning method with the ability to reduce deviation, characterized in that The following steps are involved: Step 1: In the sensor network, establish a three-dimensional coordinate system as the reference coordinate system, and assume that there are J sensors and a moving target moving in a uniform linear motion. The positions of the 1st sensor to the Nth sensor are known, and the positions of the N+1th sensor to the Jth sensor are unknown. The initial position and moving speed of the moving target are unknown. The moving target is assumed to move for a total of K moments. The true position of any sensor with a known position in the reference coordinate system is set as the origin of the reference coordinate system. When 1≤i≤N, the true position of the sensor with the i-th known position in the reference coordinate system is recorded as s i ,s i =[s x,i ,s y,i ,s z,i ] T , when N+1≤i≤J, the true position of the sensor at the i-th unknown position in the reference coordinate system is recorded as The actual initial position of the moving target in the reference coordinate system is recorded as The actual moving speed of the moving target is recorded as v o , The actual position of the moving target in the reference coordinate system at the kth moment is recorded as Where K>1, i=1,...,J,s x,i ,s y,i ,s z,i Corresponding representation s i The horizontal coordinate, vertical coordinate, and vertical coordinate of the superscript "T" represent the transpose of the vector or matrix. Corresponding representation The horizontal, vertical and vertical coordinates of Corresponding representation The horizontal, vertical and vertical coordinates of Corresponding representation v o The velocity component along the x-axis of the reference coordinate system, the velocity component along the y-axis of the reference coordinate system, and the velocity component along the z-axis of the reference coordinate system, k = 1, ..., K, Indicates a time interval; Step 2: In the sensor network, the signal transmitted by the mobile target at the kth time is received by the i-th sensor, and the i-th sensor obtains the AOA measurement value of the arrival signal contaminated by noise. The AOA measurement value includes the azimuth measurement value and the pitch angle measurement value, which are denoted by θ k,i and φ k,i ; Wherein, k=1,...,K, i=1,...,J; Step 3: Construct the AOA measurement model of the arrival signal obtained after the signal emitted by the mobile target is received by the sensor at a known position, which is described as: And construct the AOA measurement model of the arrival signal obtained after the signal emitted by the mobile target is received by the sensor at the unknown position, which is described as: Among them, θ k,i That is, the azimuth measurement value of the arrival signal contaminated by noise obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor. It represents the true value of the azimuth of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. represents the azimuth measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor, φ k,i That is, the pitch angle measurement value of the arrival signal contaminated by noise obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. It represents the true value of the pitch angle of the arrival signal obtained after the signal emitted by the mobile target at the kth time is received by the i-th sensor. represents the pitch angle measurement noise of the arrival signal obtained after the signal transmitted by the mobile target at the kth time is received by the i-th sensor. When 1≤i≤N, r k,i represents the two-dimensional distance between the moving target and the sensor at the i-th known position on the xoy plane of the reference coordinate system, When N+1≤i≤J represents the two-dimensional distance between the moving target and the sensor at the i-th unknown position on the xoy plane of the reference coordinate system, Step 4: Preprocess equation (1): Ignore the noise terms in the two equations of equation (1) and take the tangent to obtain Then cross-multiply both sides of the two equations in formula (1.1) and sort them out to get According to Perform a first-order Taylor expansion on the sine and cosine terms in the two equations of formula (1.2), and we get Preprocess equation (2): Ignore the noise terms in the two equations of equation (2) and take the tangent to obtain Then cross-multiply both sides of the two equations in formula (2.1) and sort them out to get According to Perform a first-order Taylor expansion on the sine and cosine terms in the two equations of formula (2.2), and we get Step 5: Combine equations (1.3) and (2.3) to construct a weighted least squares problem for joint positioning with partially known sensor locations, which can be described as: in, is the optimization variable of the weighted least squares problem, represents the initial position variable of the moving target, represents the moving speed variable of the moving target, represents the position variable of the sensor with the i-th unknown position, is the objective function of the weighted least squares problem, is the coefficient matrix, The kth sub-matrix in is The i-th row vector of The i-th row vector of 0 N×3M represents an all-zero matrix of dimension N×3M, where N represents the number of sensors with known positions, and M represents the number of sensors with unknown positions, M=JN, blkdiag() is a function for finding block diagonal matrices. The iNth block diagonal submatrix in is The iNth block diagonal submatrix in is is the introduced coefficient vector, The kth column vector of 0 1×M represents a vector of all zeros with a dimension of 1×M, The i-th element of The i-th element of is the introduced weight matrix, It means to find the mathematical expectation. is the introduced coefficient matrix, The kth block diagonal submatrix in is diag() is a function for finding diagonal matrices, || || is a binary norm operator, represents the measurement noise vector, The kth subvector of is Subject to the mean of 0 and the covariance matrix is Gaussian distribution, express The covariance matrix of The covariance matrix of Step 6: Introduce new auxiliary variables into the weighted least squares problem constructed in step 5 Convert the objective function of the weighted least squares problem constructed in step 5 to Then define and Will Convert to Then Split into a matrix constructed from true values and errors, namely: Finally Introduction In, we get: in, θ k,i and φ k,i Replace with and After getting The kth sub-matrix in is The i-th row vector of The i-th row vector of The iNth block diagonal submatrix in is The iN block diagonal submatrices in are θ k,i and φ k,i Replace with and After getting The kth column vector of The i-th element of The i-th element of The i-th row vector of The i-th row vector of The i-th row vector of The iNth block diagonal submatrix in is The iNth block diagonal submatrix in is The iNth block diagonal submatrix in is V=blkdiag(V1,...,V K ), the kth block diagonal submatrix in V is V k , ⊙ is the Hadamard product operator symbol, express The submatrix consisting of the 1st to the Jth row and the 1st to the Jth column in , express The submatrix consisting of the J+1th row to the 2Jth row and the J+1th column to the 2Jth column in , express The submatrix consisting of the 1st to Jth rows and the J+1th to 2Jth columns in , Depend on get, express The kth block diagonal submatrix in , The Kth column vector of The i-th element of The i-th element of The i-th element of Step 7: Based on step 6, a constrained weighted least squares problem with the ability to reduce bias is constructed when the sensor position is partially known, which is described as: in, is the optimization variable of the constrained weighted least squares problem, The dimension is (7+3M)×1, represents a constant; Step 8: Construct the Lagrangian function for the constrained weighted least squares problem constructed in step 7, expressed as: Then, according to the Lagrangian function, the KKT condition is used to obtain The optimal solution Formula Again Using eigenvalue decomposition, we can decompose Finally, according to Get the initial position variable of the moving target Estimated value of Moving target's moving speed variable Estimated value of Position variable of the sensor with the i-th unknown position Estimated value of Where L() represents the Lagrangian function, is the Lagrange multiplier, express The vector consisting of the first to third elements of express The 7+3Mth element of express The vector consisting of the 4th to 6th elements of express The vector consisting of the 4+3(iN)th element to the 6+3(iN)th element of .
Citation Information
Cited By
Underwater three-dimensional target fusion positioning method only using linear array AOA measurement
CN119881791A
