Non-cooperative unmanned aerial vehicle target state estimation method based on MGMEE-SVSF under lattice sampling

Abstract

The invention discloses a non-cooperative unmanned aerial vehicle target state estimation method based on MGMEE-SVSF under lattice sampling. The method comprises the following steps: S1, constructing a non-cooperative unmanned aerial vehicle target motion model with an uncertain turning rate and a non-linear measurement model under non-Gaussian noise interference; s2, on the basis of a lattice sampling theory and a smooth variable structure filtering theory, a time-varying smooth boundary layer under turning rate uncertainty and non-Gaussian measurement noise is obtained through a prediction step; s3, quantifying a non-cooperative unmanned aerial vehicle target state estimation error through a time-varying smooth boundary layer, and selecting a mixed generalized minimum error entropy gain or a smooth variable structure filtering gain to perform state updating; and S4, outputting the target state estimation and the state estimation error covariance of the unmanned aerial vehicle. On the premise of ensuring the stability of the estimation system, the state estimation precision is further improved through a gain switching strategy, and the problem of high-precision robust state estimation of the non-cooperative unmanned aerial vehicle target when the turning rate is uncertain and the sensor is in non-Gaussian noise interference can be effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of non-cooperative UAV target state estimation, and in particular relates to a non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling. Background Art

[0002] With the rapid development of the drone industry, the number of drones held in society has increased dramatically. This is because the controllers usually do not have a historical behavior database of non-cooperative drones, and it is difficult to accumulate trajectory data that can be used for intent identification in a short period of time.

[0003] In the tracking of non-cooperative UAV targets, the random uniform speed, acceleration or turning motion of the tracked UAV target is difficult to describe by an accurate motion model, and the environmental disturbance and sensor failure interference will make the measurement noise have non-Gaussian characteristics such as heavy tails. Under the interference of non-Gaussian noise such as uncertain turning rate and heavy tails, the non-linear state estimation of non-cooperative UAV targets is inaccurate. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes a non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling. This method designs a new robust state estimation update method for non-cooperative UAV targets based on the mixed generalized minimum error entropy (MGMEE) gain and smooth variable structure filter (SVSF) gain switching strategy under lattice sampling. When there is non-Gaussian noise interference such as motion model parameter uncertainty and heavy tail in the UAV target tracking system, the algorithm has higher state estimation accuracy.

[0005] The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling of the present invention comprises the following steps:

[0006] S1. Construct a non-cooperative UAV target motion model with uncertain turning rate and a non-linear measurement model under non-Gaussian noise interference;

[0007] S2. Based on the lattice sampling theory and the smooth variable structure filtering theory, the time-varying smooth boundary layer under the uncertainty of turning rate and non-Gaussian measurement noise is obtained through the prediction step;

[0008] S3. Quantify the target state estimation error of non-cooperative UAV through time-varying smooth boundary layer, and select hybrid generalized minimum error entropy gain or smooth variable structure filter gain for state update;

[0009] S4. Output the UAV target state estimation and state estimation error covariance.

[0010] The non-cooperative UAV target motion model in S1 is:

[0011] x k =f k (x k-1 )+B k w k (1)

[0012] In formula (1), represents the state variable of the drone in three-dimensional space at time k, x k ,y k and z k are the north direction, east direction and altitude respectively, and are the velocities in the north, east and height directions respectively, and f k (·) is the state transfer matrix, f k (·)Depend on and Two state transfer functions, B k is the process noise amplitude, The mean is 0 and the variance is Q k The process noise is n, and n is the dimension of the state space;

[0013] The nonlinear measurement model under non-Gaussian noise interference is:

[0014] z k =h k (x k )+v k (4)

[0015] In formula (4), represents the distance, elevation angle and horizontal angle measured by the radar sensor at time k, and m is the measurement dimension; is the measurement noise, the measurement noise v k It is a non-Gaussian noise that is a mixture of Gaussian noise, Rayleigh distribution noise and α steady-state distribution noise.

[0016] The prediction step in S2 includes: estimating the initial state of the drone Initial state estimation error covariance P xx,k-1|k-1 , calculate the drone state prediction State prediction error covariance P xx,k|k-1 , measure the prediction error covariance P zz,k|k-1 , and the cross-covariance of state and measurement prediction errors P xz,k|k-1 ;

[0017] The formula for state prediction is:

[0018]

[0019] In the formula, is the state prediction, N is the number of lattice points, X j,k|k-1 Transfer lattice points for states;

[0020] The state prediction error covariance P xx,k|k-1 for

[0021]

[0022] Where N is the number of lattice points, X j,k|k-1 is the state transfer lattice point, B k is the process noise amplitude, is the state prediction, Q k is the process noise variance;

[0023] The measurement prediction error covariance P zz,k|k-1 for

[0024]

[0025] The cross-covariance of state and measurement prediction errors P xz,k|k-1 for

[0026]

[0027] In the formula, Z j,k|k-1 represents the measurement transfer lattice point, represents the measured predicted value, X j,k|k-1 is the state transfer lattice point, R k represents the approximate variance of the measurement noise.

[0028] The time-varying smooth boundary layer ψ obtained by the prediction step under the uncertain turning rate and non-Gaussian measurement noise k for:

[0029]

[0030] In formula (18), diag(Α) represents the diagonal form of vector Α; represents the measurement prediction error, represents the new information measured at time k-1, represents the estimated value of the measurement at time k-1; 0<γ′<1 is the error convergence factor; H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) - 1 P xz,k|k-1 ]T , R k represents the approximate variance of the measurement noise.

[0031] The S3 of quantifying the non-cooperative UAV target state estimation error through the time-varying smooth boundary layer means that the time-varying smooth boundary layer is compared with a pre-set boundary layer threshold. When the time-varying smooth boundary layer is less than or equal to the set threshold, the mixed generalized minimum error entropy gain is used to update the state; when the time-varying smooth boundary layer is greater than the set threshold, the smooth variable structure filter gain containing the error function is used to update the state.

[0032] The state estimate after the hybrid generalized minimum error entropy gain update is

[0033]

[0034] In formula (30), is the state estimate after the tth iteration, For state prediction, represents the mixed generalized minimum error entropy gain, z k is the measured value.

[0035] The state estimation error covariance after the hybrid generalized minimum error entropy gain is updated as follows:

[0036]

[0037] In formula (32), P xx,k|k is the state estimation error covariance update, I is the identity matrix, represents the mixed generalized minimum error entropy gain, H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) -1 P xz,k|k-1 ] T , P xx,k|k-1 is the state prediction error covariance, P xz,k|k-1 is the cross-covariance of state and measurement prediction errors, R k represents the approximate variance of the measurement noise.

[0038] The state estimate after the smooth variable structure filter gain is updated is:

[0039]

[0040] In formula (33), is the state estimation, is the state prediction, K k represents the smooth variable structure filter gain, e z,k|k-1 represents the measurement prediction error;

[0041] The state estimation error covariance after the smooth variable structure filter gain is updated is:

[0042] P xx,k|k =(IK k H k ) xx,k|k-1 (IK k H k ) T +K k R k K k T (36)

[0043] In formula (36), P xx,k|k is the state estimation error covariance update, I is the identity matrix, K k is the smooth variable structure filter gain, H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) -1 P xz,k|k-1 ] T , P xx,k|k-1 is the state prediction error covariance, P xz,k|k-1 is the cross-covariance of state and measurement prediction errors, R k represents the approximate variance of the measurement noise.

[0044] The hybrid generalized minimum error entropy gain for

[0045]

[0046] In formula (20), α i >0,i=1,2 and β i >0, i=1, 2 are shape parameter and bandwidth parameter respectively, η i ≥0, i=1, 2 is the proportional parameter. k is the linearized form of the nonlinear measurement function and c=1,2 are the related covariances respectively.

[0047] The smooth variable structure filter gain K k for

[0048]

[0049] In formula (34), the superscript “+” indicates pseudo-inverse, diag(Α) indicates the diagonal form of the appropriate amount of Α; e z,k|k-1 represents the measurement prediction error, e z,k-1|k-1 represents the new information measured at time k-1, and 0<γ′<1 is the error convergence factor; is the element-wise product; the error function And erf(∞)=1,erf(-∞)=-1, then the error function in equation (34) is It can be written as:

[0050] The beneficial effects of the present invention are:

[0051] (1) The present invention overcomes the problem that the state estimation method based on the traditional precise preset model may diverge when a non-cooperative UAV target with inaccurate model parameters such as unknown turning rate randomly maneuvers by introducing lattice sampling theory and smooth variable structure filter prediction theory.

[0052] (2) The present invention proposes a hybrid generalized minimum error entropy estimation method, which overcomes the problem that the state estimation accuracy of the traditional method decreases when the non-cooperative UAV target motion model is accurate but the sensor measurement noise is non-Gaussian measurement noise such as heavy tail.

[0053] (3) The present invention designs a state estimation method based on a gain switching strategy under lattice sampling. The time-varying smooth boundary layer is used to quantify the error in the drone target state estimation. When the time-varying smooth boundary layer is greater than the threshold, the smooth variable structure filter gain is used to update the state to ensure the stability of the system; when the time-varying smooth boundary layer is less than the threshold, the mixed generalized minimum error entropy gain is used to update the state to further improve the accuracy of the drone target state estimation.

[0054] (4) Aiming at the target state estimation of non-cooperative UAV, the present invention proposes a mixed generalized minimum error entropy-smooth variable structure filter (MGMEE-SVSF) method under lattice sampling, which can effectively solve the problem of robust state estimation under non-Gaussian measurement noise such as random maneuvers and heavy tails in the motion model of non-cooperative UAV. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is an overall block diagram of the non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling of the present invention.

[0056] Figure 2 It is the algorithm flow chart of MGMEE-SVSF of the present invention.

[0057] Figure 3 It is the pseudo code of the MGMEE-SVSF algorithm flow of the present invention. DETAILED DESCRIPTION

[0058] Embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to be used to explain the present invention, but should not be understood as limiting the present invention.

[0059] like Figure 2 As shown, an embodiment of the present invention provides a non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling. The state estimation of a non-cooperative UAV during random maneuvering motion with unknown turning rate is taken as an example for explanation.

[0060] The specific steps include:

[0061] S1. Construct a non-cooperative UAV target motion model with uncertain turning rate and a non-linear measurement model under non-Gaussian noise interference;

[0062] The motion model of non-cooperative UAV target with uncertain turning rate is:

[0063] x k =f k (x k-1 )+B k w k (1)

[0064] In formula (1), represents the state variable of the drone in three-dimensional space at time k, x k ,y k and z k are the north direction, east direction and altitude respectively, and are the velocities in the north, east and altitude directions respectively, is the process noise amplitude, The mean is 0 and the variance is Q k The process noise is n, and n is the dimension of the state space. The motion of non-cooperative UAV targets can be a mixture of three-dimensional uniform linear motion and constant rate turning motion. Then the state transfer matrix f in the motion model is k (·)can be and Two types of state transfer functions are composed as follows:

[0065]

[0066] In the formula, Δt represents the sampling period, Represents the turning rate in the north, east and altitude directions respectively.

[0067] The nonlinear measurement model under non-Gaussian noise interference is:

[0068] z k =h k(x k )+v k (4)

[0069]

[0070] represents the distance, pitch angle and horizontal angle measured by the radar sensor at time k, m is the measurement dimension, and (x, y, z) is the initial coordinate position of the radar. is the measurement noise, R k Denotes the approximate variance of the measurement noise, the measurement noise v k It is a non-Gaussian noise mixed with Gaussian noise, Rayleigh distribution noise and α steady-state distribution noise. The specific form is shown in the formula.

[0071]

[0072] in, Indicates that the mean is a and the variance is Gaussian noise, R(σ) represents Rayleigh distribution noise with parameter σ, is α steady-state distributed noise, a represents the characteristic index, b represents the symmetry parameter, γ represents the scale parameter, and ω′ represents the location parameter.

[0073] S2. Based on the lattice sampling theory and the smooth variable structure filtering theory, the time-varying smooth boundary layer under the uncertainty of turning rate and non-Gaussian measurement noise is obtained through the prediction step;

[0074] Prediction step: Estimation from the initial state of the drone Initial state estimation error covariance P xx,k-1|k-1 , calculate the drone state prediction State prediction error covariance P xx,k|k-1 , measure the prediction error covariance P zz,k|k-1 , and the cross-covariance of state and measurement prediction errors P xz,k|k-1 , the specific method is as follows.

[0075] S201. Calculation state prediction

[0076]

[0077] Where N is the number of lattice points, X j,k|k-1 The lattice point is transferred for the state, which is calculated as follows:

[0078]

[0079] In formula (8), represents the state transfer function f k The estimated value of (·), χ j,k-1|k-1Expressed as an initial lattice point, the calculation is as follows:

[0080]

[0081] In formula (9), S k-1|k-1 By P xx,k-1|k-1 Obtained by Cholesky decomposition, that is Y j represents the lattice point weight, which is expressed by the inverse mapping of the normal cumulative distribution of the lattice sampling points as follows

[0082]

[0083] In formula (10), To transform the lattice points, the initial lattice points After Cranley-Patterson transformation, it can be expressed as:

[0084]

[0085] In formula (11), the initial lattice point It is generated by the rank 1 lattice rules as follows:

[0086]

[0087] In formula (12), g = [a 0 ,a 1 ,···,a n-1 ] T is a vector containing n integers, a and n are relatively prime, and jmodN represents the remainder after j is divided by N.

[0088] S202. Calculate the state prediction error covariance P xx,k|k-1 :

[0089]

[0090] S203. Calculate the measurement prediction error covariance P zz,k|k-1 and the error cross-covariance P xz,k|k-1 :

[0091]

[0092] In formula (14) and formula (15), Z j,k|k-1 Represents the measured transfer lattice point, Represents the measured predicted value, which is calculated as follows:

[0093] Z j,k|k-1 =h k (x j,k|k-1 ), j = 0, 1, · · ·, N-1 (16)

[0094]

[0095] S204. Calculate the time-varying smooth boundary layer ψ k as follows:

[0096]

[0097] In formula (18), diag(Α) represents the diagonal form of Α; represents the measurement prediction error, represents the new information measured at time k-1, represents the estimated value of the measurement at time k-1; 0<γ′<1 is the error convergence factor; H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) -1 P xz,k|k-1 ] T .

[0098] S3. Quantify the target state estimation error of non-cooperative UAVs through a time-varying smooth boundary layer, which is represented by the time-varying smooth boundary layer ψ k and the pre-set boundary layer threshold ψ con For comparison, when the time-varying smooth boundary layer is less than or equal to the set threshold, the mixed generalized minimum error entropy gain is used to update the state; when the time-varying smooth boundary layer is greater than the set threshold, the smooth variable structure filter gain containing the error function is used to update the state.

[0099] S31. If ψ k ≤ψ con , then the hybrid generalized minimum error entropy gain can be used to update the state to improve the estimation accuracy. The specific method is as follows:

[0100] S311. Combine the state prediction and the initial measurement model and expand the measurement model into the following form:

[0101]

[0102] In the formula, To expand the measurement noise, and is the state prediction error, is the measurement noise. Extended measurement noise The variance P ωω,k|k It can be expressed as Among them, S k|k-1 Available through P xx,k|k-1 Obtained by Cholesky decomposition, that is make Then S r,k Available through γk Obtained by Cholesky decomposition, that is

[0103] S312. Hybrid generalized minimum error entropy gain It can be expressed as:

[0104]

[0105] In formula (20), α i >0,i=1,2 and β i >0, i=1, 2 are shape parameter and bandwidth parameter respectively, η i ≥0, i=1, 2 is the scale parameter. The relevant covariance can be calculated as follows:

[0106]

[0107] And Θ c,xx ,Θ c,xz ,Θ c,zx ,Θ c,zz (c=1,2) can be expressed by decomposing Θ1 and Θ2,

[0108]

[0109] In formulas (22) and (23),

[0110]

[0111] In formulas (24)-(25), L = n + m, u, v = 1, 2, ..., L. Let e k =[e 1,k ,···,e L,k ] T represents the error, and the Gaussian mixture function G of the error e α,β (e) can be expressed as follows,

[0112]

[0113] In formula (26), the function z>0 is the Gamma function.

[0114] S313. Based on the extended measurement model, the error e k =[e 1,k ,···e L,k ] T It can be expressed as follows,

[0115]

[0116] In formula (27), S kAvailable through P ωω,k|k Obtained by Cholesky decomposition, that is

[0117] The cost function J of state estimation L (x k ) can be expressed as follows,

[0118]

[0119] By maximizing the cost function,

[0120]

[0121] The optimal state update can be obtained. Let t be the number of fixed-point iterations, then the tth state iteration Update the value to:

[0122]

[0123] Compare the current iteration results and the previous iteration result If the following inequality (31) is satisfied, then let the state update value Otherwise, continue to execute the state update formula (30) until the required accuracy is met or a specific number of iterations is reached.

[0124]

[0125] S314. State estimation error covariance P xx,k|k Updated as follows:

[0126]

[0127] S32. If ψ k >ψ con , then the smooth variable structure gain can be used to update the state and state estimation error covariance to ensure stability, as follows;

[0128] S321. State estimation update for:

[0129]

[0130] In formula (33), K k Expressed as gain:

[0131]

[0132] Among them, 0<γ′<1; the superscript “+” indicates pseudo-inverse; represents element-wise product; error function And erf(∞)=1,erf(-∞)=-1, then the error function in equation (34) is It can be expressed as follows:

[0133]

[0134] S322. The state estimation error covariance is updated to

[0135] P xx,k|k =(IK k H k ) xx,k|k-1 (IK k H k ) T +K k R k K k T (36)

[0136] In formula (36), I is the unit matrix.

[0137] S4. Output UAV target state estimation and the state estimation error covariance P xx,k|k .

[0138] Although the above embodiments have been shown and described, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. Changes, modifications, substitutions and variations of the above embodiments by those of ordinary skill in the art are all within the scope of protection of the present invention.

Claims

1. A non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling, characterized in that: The following steps are involved: S1. Construct a non-cooperative UAV target motion model with uncertain turning rate and a non-linear measurement model under non-Gaussian noise interference; S2. Based on the lattice sampling theory and the smooth variable structure filtering theory, the time-varying smooth boundary layer under the uncertainty of turning rate and non-Gaussian measurement noise is obtained through the prediction step; S3. Quantify the target state estimation error of non-cooperative UAV through time-varying smooth boundary layer, and select hybrid generalized minimum error entropy gain or smooth variable structure filter gain for state update; S4. Output the UAV target state estimation and state estimation error covariance.

2. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 1 is characterized in that: The non-cooperative UAV target motion model in S1 is: x k =f k (x k-1 )+B k w k (1) In formula (1), represents the state variable of the drone in three-dimensional space at time k, x k ,y k and z k are the north direction, east direction and altitude respectively, and are the velocities in the north, east and height directions respectively, and f k (·) is the state transfer matrix, f k (·)Depend on and Two state transfer functions, B k is the process noise amplitude, The mean is 0 and the variance is Q k The process noise is n, and n is the dimension of the state space; The nonlinear measurement model under non-Gaussian noise interference is: z k =h k (x k )+v k (4) In formula (4), represents the distance, elevation angle and horizontal angle measured by the radar sensor at time k, and m is the measurement dimension; is the measurement noise, the measurement noise v k It is a non-Gaussian noise that is a mixture of Gaussian noise, Rayleigh distribution noise and α steady-state distribution noise.

3. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 1 is characterized in that: The prediction step includes: estimating the initial state of the drone Initial state estimation error covariance P xx,k-1|k-1 , calculate the drone state prediction State prediction error covariance P xx,k|k-1 , measure the prediction error covariance P zz,k|k-1 , and the cross-covariance of state and measurement prediction errors P xz,k|k-1 ; The formula for state prediction is: In the formula, is the state prediction, N is the number of lattice points, X j,kk-1 Transfer lattice points for states; The state prediction error covariance P xx,k|k-1 for Where N is the number of lattice points, X j,kk-1 is the state transfer lattice point, B k is the process noise amplitude, is the state prediction, Q k is the process noise variance; The measurement prediction error covariance P zz,k|k-1 for The cross-covariance of state and measurement prediction errors P xz,k|k-1 for In the formula, Z j,kk-1 represents the measurement transfer lattice point, represents the measured predicted value, X j,kk-1 is the state transfer lattice point, R k represents the approximate variance of the measurement noise.

4. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 1 is characterized in that: The time-varying smooth boundary layer ψ obtained by the prediction step under the uncertain turning rate and non-Gaussian measurement noise k for: In formula (18), diag(Α) represents the diagonal form of vector Α; represents the measurement prediction error, represents the new information measured at time k-1, represents the estimated value of measurement at time k-1; 0<γ′<1 is the error convergence factor; H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) -1 P xz,k|k-1 ] T , R k represents the approximate variance of the measurement noise.

5. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 1 is characterized in that: The S3 of quantifying the non-cooperative UAV target state estimation error through the time-varying smooth boundary layer means that the time-varying smooth boundary layer is compared with a pre-set boundary layer threshold. When the time-varying smooth boundary layer is less than or equal to the set threshold, the mixed generalized minimum error entropy gain is used to update the state; when the time-varying smooth boundary layer is greater than the set threshold, the smooth variable structure filter gain containing the error function is used to update the state.

6. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 5 is characterized in that: The state estimate after the hybrid generalized minimum error entropy gain update is In formula (30), is the state estimate after the tth iteration, For state prediction, represents the mixed generalized minimum error entropy gain, z k is the measured value; The state estimation error covariance after the hybrid generalized minimum error entropy gain is updated as follows: In formula (32), P xx,k|k is the state estimation error covariance update, I is the identity matrix, represents the mixed generalized minimum error entropy gain, H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) -1 P xz,k|k-1 ] T , P xx,k|k-1 is the state prediction error covariance, P xz,k|k-1 is the cross-covariance of state and measurement prediction errors, R k represents the approximate variance of the measurement noise.

7. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 5 is characterized in that: The state estimate after the smooth variable structure filter gain is updated is: In formula (33), is the state estimation, is the state prediction, K k represents the smooth variable structure filter gain, e z,k|k-1 represents the measurement prediction error; The state estimation error covariance after the smooth variable structure filter gain is updated is: P xx,kk =(I-K k H k )P xx,kk-1 (I-K k H k ) T +K k R k K k T (36) In formula (36), P xx,k|k is the state estimation error covariance update, I is the identity matrix, K k is the smooth variable structure filter gain, H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) -1 P xz,k|k-1 ] T , P xx,k|k-1 is the state prediction error covariance, P xz,k|k-1 is the cross-covariance of state and measurement prediction errors, R k represents the approximate variance of the measurement noise.

8. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 6 is characterized in that: The hybrid generalized minimum error entropy gain for In formula (20), α i >0,i=1,2 and β i >0, i=1, 2 are shape parameter and bandwidth parameter respectively, η i ≥0, i=1, 2 is the proportional parameter; H k is the linearized form of the nonlinear measurement function and H k =[(P xx,k|k-1 ) -1 P xz,k|k-1 ] T , are the relevant covariances respectively.

9. The non-cooperative UAV target state estimation method based on MGMEE-SVSF under lattice sampling according to claim 7 is characterized in that: The smooth variable structure filter gain K k for In formula (34), the superscript "+" indicates pseudo-inverse, diag(Α) indicates the diagonal form of the appropriate amount of Α; e z,kk-1 represents the measurement prediction error, e z,k-1|k-1 represents the new information measured at time k-1, and 0<γ′<1 is the error convergence factor; is the element-wise product; the error function And erf(∞)=1,erf(-∞)=-1, then the error function in equation (34) is It can be written as:

Images