Joint design of constant modulus waveform and receiving filter based on Riemannian product manifold framework

By employing a relaxation-free optimization method based on the Riemannian product manifold framework, the computational complexity and performance loss issues in waveform and filter design in radar systems are addressed, achieving efficient interference suppression and target detection.

CN119026359APending Publication Date: 2024-11-26UNIV OF ELECTRONICS SCI & TECH OF CHINA +1
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411141351.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Existing radar systems suffer from high computational complexity, approximation errors due to performance relaxation, and signal-to-noise ratio loss when designing constant-mode waveforms and receiving filters. In particular, it is difficult to effectively improve interference suppression and target detection capabilities during joint optimization.

Method used

A relaxation-free optimization problem is constructed using the Riemann product manifold framework. The transmit waveform and receive filter are jointly designed on the Riemann product manifold using a parallel conjugate gradient algorithm, which avoids the approximation error caused by relaxation processing and satisfies constant mode constraints and energy constraints.

Benefits of technology

It achieves joint optimization without relaxation, improves radar interference suppression capability and target detection performance, reduces signal-to-noise ratio loss, and improves the clarity and efficiency of target detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026359A_ABST
    Figure CN119026359A_ABST
Patent Text Reader

Abstract

The invention discloses a joint design of a constant modulus waveform and a receiving filter based on a Riemannian product manifold framework, and belongs to the field of radars. According to the method, a Riemannian product manifold framework is constructed, relaxation processing can be directly carried out on the optimization problem of joint design of the receiving filter and the transmitted waveform without relaxation, so that approximate errors caused by relaxation processing are avoided, and when joint design needs to be carried out on the receiving filter and the transmitted waveform according to an expected AF graph, the optimization problem of joint design of the receiving filter and the transmitted waveform is solved. The method is implemented based on a Riemannian product manifold method of fuzzy function forming, and the implementation mode is a non-relaxation method based on mismatched filtering. According to the method, firstly, a Riemannian product manifold (RPM) is constructed to meet constant modulus constraint and energy constraint at the same time; then, an optimization problem of joint design of a receiving filter and a transmitted waveform is converted into an unconstrained optimization problem on RPM; and finally, solving through a parallel conjugate gradient algorithm of RPM, and obtaining a final optimal transmitting waveform and an optimal receiving filter based on a solving result. The transmitted waveform designed by the invention can obtain a deeper fuzzy function notch, and the weak target detection performance is better.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of radar technology, and in particular to a joint design of a constant modulus waveform and a receiving filter based on a Riemann product manifold framework. Background Art

[0002] Constant modulus constraint (CMC) waveform design based on ambiguity function (AF) shaping is a key technology in radar systems. Shaping the AF with a deep zero line can improve the radar's resolution, anti-interference, and stealth capabilities. In addition, since nonlinear power amplifiers usually operate in saturation mode to maximize efficiency, it is preferred to enforce CMC on the waveform. Therefore, AF shaping with CMC has received widespread attention.

[0003] In radar systems, AF is determined by waveforms and receiving filters, where receiving filters can be divided into matched filters and unmatched filters. According to the type of receiving filters, existing methods can be mainly divided into two categories. The first category is to design a single-mode waveform based on autofocus shaping using a matched filter. Due to the limited degree of freedom, the interference suppression performance of this type of method is limited. The second category is to design a single-mode waveform based on AF shaping using a mismatched filter. By jointly designing the mismatched filter and the transmitted waveform, the degree of freedom for suppressing interference is increased at an acceptable signal-to-noise ratio (SNR) loss.

[0004] For the first type of methods, the problem is formulated as a complex quartic problem. To solve this problem, Aubry A, De Maio A, Jiang B, et al. Ambiguity function shaping for cognitive radar via complex quartic optimization [J]. IEEE Transactions on Signal Processing, 2013, 61 (22): 5603-5619. proposed a two-stage maximum block improvement (MBI) method to transform the quartic problem into a multilinear problem to solve. However, this method requires the generation of a large number of filters and waveform samples, and has high computational complexity. In order to reduce the computational complexity, the paper Wu L, Babu P, Palomar D P. Cognitive radar-based sequence design via SINR maximization [J]. IEEE Transactions on Signal Processing, 2016, 65 (3): 779-793. proposed a method based on maximization-minimization (MM). It simplifies the complex quartic problem by deriving a surrogate function. However, constructing a surrogate function is a challenge.

[0005] For the second type of methods, the problem is formulated as a coupled quadratic problem with CMC and complex sphere constraints. In general, existing methods solve this non-convex problem by relaxation. T, Koivunen V. Mismatched filter design and interference mitigation for MIMO radars [J]. IEEE transactions on signal processing, 2016, 65 (2): 454-466. " proposed a semidefinite relaxation (SDR) method, which relaxes both the objective function and the constraints, resulting in performance degradation. In order to further improve the performance, the paper "Zhou K, Li D, Su Y, et al. Joint design of transmit waveform and mismatch filter in the presence of interrupted sampling repeater jamming [J]. IEEE Signal Processing Letters, 2020, 27: 1610-1614. " proposed an alternating direction multiplication (ADMM) method, which only relaxes the objective function, but it is challenging to determine the relaxation parameters. To solve this problem, the paper "Wang F, Feng S, Yin J, et al. Unimodular sequence and receiving filter design for local ambiguity function shaping[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-12." proposed a minimization (MM)-based method that avoids the selection of relaxation parameters by deriving a surrogate function. However, the surrogate function still inevitably leads to approximation errors. Summary of the invention

[0006] The purpose of the present invention is to provide a joint design of a constant modulus waveform and a receiving filter based on a Riemann product manifold framework, which is a joint design method of a constant modulus waveform and a receiving filter based on a Riemann product manifold method, which is used to improve the radar's detection capability for targets at specific distances and speeds.

[0007] The technical solution adopted by the present invention is:

[0008] The joint design of the constant modulus waveform and the receiving filter based on the Riemann product manifold framework includes the following steps:

[0009] Step 1: Construct the Riemann product manifold space based on the transmit waveform and receive filter of the single-input single-output radar system:

[0010] M=M x ×M h ={(x,h):x∈M x ,h∈M h}

[0011]

[0012] Where M is the Riemann product manifold, and the emission waveform x = [x(1), x(2), …, x(N)] T , x(n) is the transmitted symbol of the transmitted waveform x at the nth symbol time, n is the symbol time, N is the symbol length of the transmitted waveform, and x(n)=e jφ(n) , j is the imaginary unit, e is the natural base, φ(n) is the phase of the transmitted waveform x, φ(n)∈[0,2π]; the receiving filter sequence h n represents the nth element of the receive filter sequence, M x is the complex circular manifold about the emission waveform x, M h is the complex spherical manifold about the receiving filter h, N h represents the energy constraint value of the receiving filter, (·) T represents the transpose operation, ||·|| F , |·| respectively represent the matrix bi-norm and vector element-wise modulus, represents a complex domain;

[0013] Step 2: Construct a joint optimization problem on the Riemann product manifold M:

[0014]

[0015] sth H h=N h

[0016] |x n |=1,n=1,...,N

[0017] Where Γ(x,h) represents the objective function of the joint optimization problem, ε represents the weighting parameter (whose value range can be set to ε∈[0,1]), WISL(x,h) represents the weighted integrated sidelobe level (WISL) of the discrete local ambiguity function about x and h, and the peak cost function g(x,h)=|h H x max |2 , a max represents the preset control parameter, which is used to control the peak value of the local ambiguity function with the expected value μ of the signal-to-noise ratio loss (SNRL), (·) H represents the conjugate transpose operation;

[0018] Specifically, Where, w(n,f) is the weight at the nth symbol time and Doppler frequency shift f, a(f) is the normalized Doppler frequency vector, Diag(·) represents the diagonal matrix, and the auxiliary quantity J n Elements Among them, element index p, q = 1, ..., N, n = 1-N, ..., N-1; f 1 ,f 2 The boundary value of the Doppler frequency range is set;

[0019] It is also possible to calculate the Doppler frequency range [f 1 ,f 2 ] is divided into L adjacent intervals of Δf=(f 2 -f 1 ) / (L-1) is a discrete Doppler frequency value, and its corresponding weighted integrated sidelobe level is expressed as:

[0020]

[0021] Among them, l represents the discrete point index, f l =f 1 +lΔf is the discrete Doppler frequency.

[0022] Specifically, a max The specific settings are

[0023] Step 3: Use the parallel conjugate gradient algorithm to solve the constructed objective function Γ(x,h), and obtain the optimal transmit waveform and receive filter based on the solution results.

[0024] Specifically, in step 3, the parallel conjugate gradient algorithm is used to solve the constructed objective function f(x,h), which specifically includes:

[0025] (1) Compute the parallel Riemann gradient gradf(x,h) on the Riemann product manifold M based on the current values ​​of x and h:

[0026] gradf(x,h)=Proj (x,h) (Gradf(x,h))

[0027] =[Proj x (Grad x f(x,h));Proj h(Grad h f(x,h))]

[0028] =[grad x f(x,h);grad h f(x,h)]

[0029] Where Gradf(x,h) is the Euclidean gradient of f(x,h), Proj (x,h) (Gradf(x,h)) is the linear tangent space T of Gradf(x,h) on the Riemann product manifold M (x,h) Orthogonal projection on M; and are the Riemann gradients of x and h respectively;

[0030] (2) Use the Polak-Ribiere conjugate gradient method to determine the descent direction of the transmit waveform and the receive filter and the descending direction The subscript k is used to identify the current number of iterations;

[0031] (3) Based on the step size η corresponding to the current iteration k , update x and h on the Riemann product manifold M to obtain the emission waveform x of the k+1th iteration k+1 and the receive filter sequence h k+1 :

[0032]

[0033] in, and They are respectively the scaling of x and h on RPM. The specific operations are:

[0034]

[0035] Zoom formula Ret C (·), Ret E The expression of (·) is:

[0036]

[0037] Among them, e and e' are respectively the scaling formula Ret C (·), Ret E (·) input parameters;

[0038] (3) Based on x k+1 and h k+1 , the Armijo linear search algorithm is used to adaptively update the initial step size (η) of the next iteration k+1), and then in the next iteration, the step length update result is obtained based on the update of the initial step length, such as the update result η new =c n′ η k+1 , where c (0<c<1) is the Armijo parameter, n′ is the number of Armijo searches; after the update is completed, η can be directly new Assigns a value to the parameter symbol η k+1 , participate in subsequent calculation and processing;

[0039] (4) Determine whether the function value of the objective function f(x,h) converges. If so, the current emission waveform x k+1 and the receive filter sequence h k+1 As the final solution result; otherwise, based on the current x k+1 and h k+1 Continue with steps (1) to (4).

[0040] The technical solution provided by the present invention brings at least the following beneficial effects:

[0041] In the present invention, by constructing a Riemann product manifold framework, the optimization problem of the joint design of the receiving filter and the transmitting waveform can be directly relaxed without relaxation, thereby avoiding the approximate error caused by the relaxation process. When the receiving filter and the transmitting waveform need to be jointly designed according to the desired AF map, the present invention is implemented based on the Riemann product manifold method of fuzzy function type assignment, and this implementation method is a non-relaxation method based on mismatch filtering. The present invention first constructs a Riemann product manifold (RPM) to simultaneously satisfy the constant modulus constraint and the energy constraint; then, the optimization problem of the joint design of the receiving filter and the transmitting waveform is converted into an unconstrained optimization problem on the RPM; finally, the parallel conjugate gradient algorithm of the RPM is used to solve it, and the final optimal transmitting waveform and the optimal receiving filter are obtained based on the solution result. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 It is a CIAFIS (matched filter strategy) fuzzy function diagram of scenario 1 in the embodiment;

[0044] Figure 2 MMSR (mismatched filter strategy) fuzzy function diagram of scenario 1 in the embodiment;

[0045] Figure 3 The PMAFS (Riemann product manifold method of fuzzy function modeling) fuzzy function diagram of the present invention for scenario 1 in the embodiment;

[0046] Figure 4 It is a zero Doppler ambiguity function slice diagram of the method proposed in the present invention, the existing CIAFIS method and the MMSR method in scenario 1 of the embodiment;

[0047] Figure 5 is a cost function comparison curve diagram of the PMAFS method of the present invention and the MMSR method in scenario 1 of the embodiment;

[0048] Figure 6 This is a target detection effect diagram of the MMSR method in scenario 2 of the embodiment;

[0049] Figure 7 This is a diagram showing the target detection effect of the PMAFS method of the present invention in Scenario 2 of the embodiment. DETAILED DESCRIPTION

[0050] To make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be described in detail and completely in conjunction with the drawings in the implementation of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings can be arranged and designed using different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present invention.

[0051] The problem modeling of the joint design of constant modulus waveform and filter shaped by local ambiguity function (AF) can be described as:

[0052] Consider the transmit waveform x of a SISO (single-input single-output) radar system, whose length is N, expressed as:

[0053] x=[e jφ(1) ,e jφ(2) ,…,e jφ(N) ] T ,φ(n)∈[0,2π],n=1,...,N

[0054] Where e represents the natural base, n is used to represent the symbol time, and φ(n) represents the phase of the transmitted waveform x. Consider a receiving filter whose corresponding receiving filter sequence is in,(·) T represents the transpose operation, ||h|| 2 =N h , Nh represents the energy constraint value of the receiving filter, and the discrete AF corresponding to x and h is:

[0055] A x,h (n,f)=h H J n Diag(a(f))x

[0056] in

[0057]

[0058] p,q=1,…,N,n=1-N,…,N-1

[0059] Among them, J -n =(J n ) T , Diag(·) represents the diagonal function.

[0060] a(f) is the normalized Doppler frequency vector:

[0061]

[0062] Where f is the Doppler frequency shift, f∈[-N / 2,N / 2).

[0063] The Doppler frequency range of interest is [f 1 ,f 2 ], local fuzzy function A x,h The weighted integrated sidelobe level (WISL) of

[0064]

[0065] where w(n,f) is the weight at distance n and Doppler frequency shift f. 1 ,f 2 ] is divided into L adjacent intervals of Δf=(f 2 -f 1 ) / (L-1) is the discretized Doppler frequency value, and the discretized WISL is:

[0066]

[0067] In the formula, f l =f 1 +lΔf is the discrete Doppler frequency.

[0068] Unmatched filters inevitably cause signal-to-noise ratio loss (SNRL), namely:

[0069]

[0070] Since ||x||2 and ||h|| 2 is a constant, SNRL is given by |h H x| determines, which is the peak value of AF. The present invention considers a peak cost function to control SNRL, and its expression is:

[0071] g(x,h)=|h H x max | 2

[0072] in, The peak value of AF is controlled by the expected value μ of SNRL, so the joint optimization problem can be expressed as:

[0073]

[0074] sth H h=N h

[0075] |x n |=1,n=1,...,N

[0076] Where Γ(x,h) represents the objective function of the joint optimization problem, and ε∈[0,1] is the weighting parameter.

[0077] Based on the above description, the joint design of a constant modulus waveform and a receiving filter based on a Riemann product manifold framework provided by an embodiment of the present invention is a joint design using a Riemann product manifold method (PMAFS) with fuzzy function modeling, and the specific implementation steps include:

[0078] Step 1: Riemann product manifold space construction:

[0079] In order to satisfy both the constant modulus constraint (CMC) and the complex sphere constraint (CSC), an RPM space is constructed, which is defined as:

[0080] M=M x ×M h ={(x,h):x∈M x ,h∈M h}

[0081]

[0082] Among them, M is the product manifold, M x is a complex circular manifold, M h is a complex spherical manifold.

[0083] Then, the problem can be transformed into an unconstrained optimization problem:

[0084] f(x,h)=εWISL(x,h)+(1-ε)g(x,h)

[0085] st(x,h)∈M

[0086] Among them, f(x,h) is the corresponding objective function.

[0087] Since RPM is a nonlinear space, the gradient needs to be calculated in a linear space. Therefore, a linear tangent space is established around (x,h)∈M to solve the problem on RPM, defined as:

[0088]

[0089] in, and are the tangent vectors of x and h, respectively, N is an N-dimensional 0 vector, ⊙ is the Hadamard product, denote the real and imaginary parts of the elements of a complex vector, respectively.

[0090] Step 2: Compute the parallel Riemann gradient

[0091] According to this problem model, we define The directional derivative in the direction (x, h) is:

[0092]

[0093] Where Gradf(x,h) is the Euclidean gradient of f(x,h), Proj (x,h) (Gradf(x,h)) is the value of Gradf(x,h) in T (x,h) Orthogonal projection onto M.

[0094] Then, the parallel Riemann gradient on the product manifold M It is expressed as:

[0095] gradf(x,h)=Proj (x,h) (Gradf(x,h))

[0096] =[Proj x (Grad x f(x,h));Proj h (Grad h f(x,h))]

[0097] =[grad x f(x,h);grad h f(x,h)]

[0098] in, and is the Riemann gradient (Euclidean gradient) of x and h.

[0099] The Euclidean gradient of x is expressed as:

[0100]

[0101] Among them, the auxiliary amount b = h H J n Diag(a(f l )).

[0102] Similarly, the Euclidean gradient of h is expressed as:

[0103]

[0104] Among them, the auxiliary amount c = J n Diag(a(f l ))x.

[0105] Get the Riemann gradient grad x f(x,h) and grad h After f(x,h), the Euclidean gradient is projected onto the tangent space, expressed as:

[0106]

[0107] Step 3: Update the feasible solution:

[0108] In order to achieve superlinear convergence, the descent direction is determined by the Polak-Ribiere conjugate gradient method. and It is expressed as:

[0109]

[0110] Among them, β k ≥0 is the Polak-Ribieres conjugation parameter, which is calculated as:

[0111]

[0112] Among them, gradf k =gradf(x k ,h k )and Trans(·)←(·) is used to establish the relationship between tangent vectors in different tangent spaces. Transferring from (x,h)∈M to another point (x′,h′)∈M is expressed as:

[0113]

[0114] in

[0115]

[0116] Based on the above construction, update (x k ,h k ) to (x k+1 ,h k+1 ) process is expressed as:

[0117]

[0118] The Armijo linear search algorithm will then be used to determine the step size η k , the scaling operation is expressed as:

[0119]

[0120] Right now

[0121] in, and Define the scaling operations for x and h on RPM respectively:

[0122]

[0123] Based on the above construction and for:

[0124]

[0125] To simplify notation, the overall descent direction is defined as:

[0126]

[0127] Step 4: Adaptive step size update:

[0128] This embodiment uses the Armijo linear search algorithm to adaptively update the step size η k At the same time, this strategy ensures the monotonicity of the algorithm and improves the convergence speed of the algorithm, which can be expressed as:

[0129]

[0130] The initial step size of the next iteration is η k+1 =κ 2 η k , κ 1 and κ 2 ∈(0,1).

[0131] In order to further speed up the convergence, the step size η k Make appropriate adjustments based on the change in the number of linear searches n:

[0132] n = 1, indicating that the step size is too small, the initial step size of the next iteration should be increased to

[0133] n = 2, indicating that the initial step size is reasonable, and the step size of the next iteration should remain η k+1 =η k ;

[0134] n>2, indicating that the step size of the current point is too small, and the initial step size of the next iteration should be increased to

[0135] Example

[0136] In order to verify the solution proposed in this embodiment, the waveform is optimized by using the existing solution and the proposed solution.

[0137] The simulation parameters are set as follows: sequence length N = 50 and normalized Doppler frequency discrete number L = 50. ε is set to 0.1, and the signal-to-noise ratio loss is set to μ = 0.5 dB. In order to avoid loss of generality, this embodiment sets the same random normalized phase sequence x and h as the initial input of the two methods. When the number of iterations is greater than 600, the optimization will stop.

[0138] Existing solutions:

[0139] (1) CIAFIS method (matched filter strategy), reference: Wu L, Babu P, Palomar DP. Cognitive radar-based sequence design via SINR maximization [J]. IEEE Transactions on Signal Processing, 2016, 65(3): 779-793.

[0140] (2) MMSR method (mismatched filter strategy), reference: Wang F, Feng S, Yin J, et al. Unimodular sequence and receiving filter design for local ambiguity function shaping[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-12.

[0141] First, consider suppressing the side lobes near zero Doppler and zero distance (n,f) = (0,0). Scenario 1 is set as:

[0142]

[0143] Figure 1 , Figure 2 and Figure 3 The fuzzy function diagrams of CIAFIS, MMSR and PMAFS of the present invention are shown respectively. It can be seen that PMAFS of the present invention obtains a deeper notch than CIAFIS and MMSR. Figure 4 The zero Doppler ambiguity function slicing effects of the three methods are shown. It can be seen that the notch value of the PMAFS of the present invention is about 30dB and 20dB lower than that of CIAFIS and MMSR, respectively. This result shows that the scheme proposed by the present invention has a strong interference suppression performance. Figure 5 The relationship curves of WISL and different SNRL (μ) are shown. As μ increases, the WISL values ​​obtained by both methods are lower. For different μ, PMAFS performs better than MMSR. At the same time, for the proposed scheme, each iteration is more effective than MMSR. In general, for μ = 0, at the 600th iteration, the WISL of PMAFS reaches 0.00224, which is much lower than MMSR (0.2084) at 5001 iterations for μ = 1.0.

[0144] In order to evaluate the ability of the sequence designed by the proposed solution to detect moving targets, two weak targets are considered, located at distances of 2 and 21, respectively, with normalized Doppler frequencies of 2 and 12, respectively, and the ratios of zero Doppler zero-distance response to target echo are 10 dB and 15 dB, respectively. In order to simplify the model, an effective method is to minimize the sidelobe level of AF in the area where the target is expected to appear. On this basis, scenario 2 is set as:

[0145]

[0146] Figure 6 The AF of the original sequence, the sequence obtained by the MMSR method, and the sequence obtained by the PMAFS method are shown. Figure 7 It can be seen that MMSR can successfully identify target 1, but the side lobe of sub-area 2 cannot be effectively suppressed, resulting in target 2 being covered and unable to be detected. Different from the MMSR method, the scheme proposed in the present invention can clearly detect two targets, and has a lower side lobe level in sub-area 1 where target 1 is located, and has better detection performance.

[0147] Through the above simulations, the performance advantage of the sequence of the joint design of the constant modulus waveform and the receiving filter based on the Riemann product manifold framework proposed in the present invention is verified. Compared with the existing methods, the present invention achieves lower side lobes in the range-Doppler region of interest. The target detection simulation results confirm that the designed mismatched filter and waveform have good target detection performance.

[0148] 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0149] The above are only some embodiments of the present invention. For those skilled in the art, several modifications and improvements can be made without departing from the creative concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. Joint design of constant modulus waveform and receiving filter based on Riemann product manifold framework, characterized by: The following steps are involved: Step 1: Construct the Riemann product manifold space based on the transmit waveform and receive filter of the single-input single-output radar system: M=M x ×M h ={(x,h):x∈M x ,h∈M h } Where M is the Riemann product manifold, and the emission waveform x = [x(1), x(2), …, x(N)] T , x(n) is the transmitted symbol of the transmitted waveform x at the nth symbol time, n is the symbol time, N is the symbol length of the transmitted waveform, and the receiving filter sequence h n represents the nth element of the receive filter sequence, M x is the complex circular manifold about the emission waveform x, M h is the complex spherical manifold about the receiving filter h, N h represents the energy constraint value of the receiving filter, (·) T represents the transpose operation, represents a complex domain; Step 2: Construct a joint optimization problem on the Riemann product manifold M: s.t.h H h=N h |x n |=1,n=1,...,N Where f(x,h) represents the objective function of the joint optimization problem, ε represents the weighting parameter, WISL(x,h) represents the weighted integrated sidelobe level of the discrete local ambiguity function about x and h, and the peak cost function g(x,h) = |h H x max | 2 , a max represents the preset control parameter, which is used to control the peak value of the local fuzzy function with the expected value μ of the signal-to-noise ratio loss, (·) H represents the conjugate transpose operation; Step three, use the parallel conjugate gradient algorithm to solve the constructed objective function f(x,h), and obtain the optimal transmit waveform and receive filter based on the solution results.

2. The joint design of constant modulus waveform and receiving filter based on Riemann product manifold framework as claimed in claim 1, characterized in that: Weighting parameter ε∈[0,1].

3. The joint design of constant modulus waveform and receiving filter based on Riemann product manifold framework as claimed in claim 1, characterized in that: The expression of WISL(x,h) is set as: Where w(n,f) is the weight at the nth symbol time and Doppler frequency shift f, a(f) is the normalized Doppler frequency vector, Diag(·) represents the diagonal matrix, and the auxiliary quantity J is n Elements Among them, the element index p, q = 1, ..., N, n = 1-N, ..., N-1; f1, f2 are the boundary values ​​of the set Doppler frequency range.

4. The joint design of constant modulus waveform and receiving filter based on Riemann product manifold framework as claimed in claim 1, characterized in that: The expression of WISL(x,h) is set as: The Doppler frequency range [f1, f2] is divided into L discrete Doppler frequency values ​​with adjacent intervals of Δf = (f2-f1) / (L-1), and the corresponding weighted integrated sidelobe level is expressed as: Where L represents the number of discrete Doppler frequency values ​​that divide the Doppler frequency range [f1, f2] into adjacent values ​​with an interval of Δf = (f2-f1) / (L-1), l represents the discrete point index, and the discrete Doppler frequency f l =f1+lΔf,w(n,f l ) is the nth symbol time and the discrete Doppler frequency f l The weight at a(f l ) corresponds to the discrete Doppler frequency f l The normalized Doppler frequency vector, Diag(·) represents the diagonal matrix, and the auxiliary quantity J n Elements Among them, the element index p, q = 1, ..., N, n = 1-N, ..., N-1.

5. The joint design of a constant modulus waveform and a receiving filter based on a Riemann product manifold framework according to any one of claims 1 to 4, characterized in that: A max The specific settings are 6. The joint design of constant modulus waveform and receiving filter based on Riemann product manifold framework as claimed in claim 1, characterized in that: In step 3, the parallel conjugate gradient algorithm is used to solve the constructed objective function f(x,h), specifically including: (1) Based on the current values ​​of x and h, calculate the parallel Riemann gradient gradf(x,h) on the Riemann product manifold M: gradf(x,h)=Proj (x,h) (Gradf(x,h)) =[Proj x (Grad. x f(x,h));Proj h (Grad. h f(x,h))] =[grad x f(x,h);grad h f(x,h)] Where Gradf(x,h) is the Euclidean gradient of f(x,h), Proj (x,h) (Gradf(x,h)) is the linear tangent space T of Gradf(x,h) on the Riemann product manifold M (x,h) Orthogonal projection on M; and are the Riemann gradients of x and h respectively; (2) Use the Polak-Ribiere conjugate gradient method to determine the descent direction of the transmit waveform and the receive filter and descending direction The subscript k is used to identify the number of iterations; (3) Based on the step size η corresponding to the current number of iterations k , update x and h on the Riemann product manifold M to obtain the emission waveform x of the k+1th iteration k+1 and the receive filter sequence h k+1 : in, and They are respectively the scaling of x and h on RPM. The specific operations are: Zoom formula Ret C (·), Ret E The expression of (·) is: Among them, e and e' are respectively the scaling formula Ret C (·), Ret E (·) input parameters; (3) Based on x k+1 and h k+1 , the Armijo linear search algorithm is used to adaptively update the initial step size of the next iteration; (4) Determine whether the function value of the objective function f(x,h) converges. If so, the current emission waveform x k+1 and the receive filter sequence h k+1 As the final solution result; otherwise, based on the current x k+1 and h k+1 Continue with steps (1) to (4).

7. The joint design of constant modulus waveform and receiving filter based on Riemann product manifold framework as claimed in claim 6, characterized in that: In step (3), the update strategy for the initial step size of the next iteration is set to: If the number of step searches is 1, the initial step size for the next iteration is Among them, the parameters The value of is greater than 1; If the number of step searches is 2, the initial step size for the next iteration is η k ; If the number of step searches is greater than 1, the initial step size for the next iteration is Among them, the parameters The value of is greater than 1.

Citation Information

Cited By

  • Manifold optimization low-sidelobe constant modulus waveform design method for low-altitude unmanned aerial vehicle detection

    CN119291616A

  • Anti-interference waveform design method based on directional diagram matching under joint constraint

    CN119861339A

  • An anti-jamming waveform design method based on direction pattern matching under joint constraints

    CN119861339B