A Phase-coded Sequence Parallel Acceleration Descent Optimization Design Method

Through the parallel acceleration descent optimization design of simulated annealing algorithm and coordinate descent method, the problem of low waveform sequence efficiency and insufficient research on frequency-phase composite coded signals in radar waveform optimization is solved, the expansion and flexibility of radar waveform library is improved, and the The probability of obtaining the optimal solution.

CN113534087BActive Publication Date: 2024-12-06THE 724TH RESEARCH INSTITUTE OF CHINA STATE SHIPBUILDING CORP LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110751918.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-07-02
Publication Date
2024-12-06
Estimated Expiration
2041-07-02

AI Technical Summary

Technical Problem

In the radar waveform optimization, the existing technology has problems such as low waveform sequence optimization efficiency, limited waveform library scale and insufficient research on frequency-phase composite coded signals in radar waveform optimization, which makes it difficult to effectively expand and flexibly adjust the radar waveform library, and the intelligent algorithm and coordinate descent method are at a high level. There are challenges such as sensitivity and large feasible solutions on the problem of dimension optimization.

Method used

A simulated annealing algorithm is used to generate a new solution and use it as the initial parameter of the coordinate descent method. Through the parallel acceleration descent optimization design method, the multiphase coding sequence is optimized, combined with the minimization of autocorrelation and cross-correlation functions, the optimal solution is improved. Probability.

Benefits of technology

The probability of obtaining the optimal solution that meets the needs is significantly improved, the problem of coordinate descent method being sensitive to initial values ​​is solved, and the reduction of the objective function is accelerated through the parallel acceleration descent algorithm, which improves the expansion and flexibility of the waveform library.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113534087B_ABST
    Figure CN113534087B_ABST
Patent Text Reader

Abstract

The present invention designs a phase-coded sequence parallel accelerated descent optimization design method. For non-convex optimization problems, the coordinate descent method is sensitive to initial values ​​and cannot guarantee that a satisfactory optimal solution will be obtained in the end. In view of the problems of the above two optimization methods, the present invention proposes a phase-coded sequence parallel accelerated descent optimization design method, and the specific steps include: according to the set initial coding parameters, a simulated annealing algorithm is used to generate a new solution, and the new solution is used as the initial parameter of the coordinate descent method to update the variable, and it is determined whether the objective function increment corresponding to the updated solution meets the acceptance condition, and when the new solution meets the convergence condition, the calculation is terminated. The present invention combines the simulated annealing algorithm and the coordinate descent method, and uses the simulated annealing solution as the input of the coordinate descent, which not only accelerates the descent of the objective function, but also solves the problem of the coordinate descent method being sensitive to the initial variables.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of radar waveform optimization. Background Art

[0002] As the electromagnetic environment becomes more and more complex, the intensity of electronic countermeasures based on RF storage functions is high and diverse, and the low efficiency of waveform sequence optimization has seriously constrained the expansion of the radar waveform library and the flexibility of waveform truncation. For the optimization of discrete multi-phase coded sequences, usually only intelligent algorithms can be used for global random optimization. Therefore, on the one hand, due to the limitation of waveform optimization efficiency, the scale of the waveform library of current radar equipment is limited; on the other hand, domestic research on frequency-phase composite coded signals is extremely difficult, and it is impossible to learn from foreign superior technologies to upgrade our waveform library.

[0003] Regarding the optimization problem of multi-phase coding sequence groups, the objective function must consider both the autocorrelation of each sequence and the cross-correlation between pulses. This is a multi-objective optimization problem and a Pareto optimization problem. Intelligent algorithms and coordinate descent methods can be used to solve the optimization design problem of multi-phase coding sequence groups. When the entire coding sequence group is used as an optimization variable, its dimension is high, the feasible solution space is large, and the random search is difficult. Therefore, the probability of obtaining the optimal solution that meets the requirements is very low if the entire intelligent algorithm is used for global optimization. For non-convex optimization problems, the coordinate descent method is sensitive to the initial value, and the coordinate descent method alone cannot guarantee that a satisfactory optimal solution can be obtained in the end. Therefore, an intelligent algorithm such as the simulated annealing algorithm can be used to obtain a set of coding sequences to achieve a preliminary descent of the objective function, and then the coordinate descent method can be used to use the coding sequence group as the initial value to further optimize the objective function. In the end, the probability of obtaining the optimal solution that meets the requirements can be greatly increased. Summary of the invention

[0004] The present invention proposes a multi-phase coding sequence parallel accelerated descent optimization design method to optimize the multi-phase coding sequence and improve the probability of obtaining the optimal solution that meets the requirements.

[0005] The technical solution to realize the present invention is: according to the set initial encoding parameters, a new solution is generated by using a simulated annealing algorithm, and the new solution is used as the initial parameter of the coordinate descent method to update the variable, and it is determined whether the increment of the objective function corresponding to the updated solution meets the acceptance condition. When the new solution meets the convergence condition, the calculation is terminated. The specific steps are:

[0006] Step 1: Set the initial temperature, cooling times M, and initial multi-phase encoding group.

[0007] Step 2: Generate multiple new solutions based on the current temperature.

[0008] (1) Calculate the peak sidelobe level of the autocorrelation function of the frequency-phase coded signal;

[0009] (2) Calculating the peak level of the inter-pulse cross-correlation function of the frequency-phase coded signal;

[0010] (3) Solve the Pareto optimization problem and generate a new solution X'.

[0011] Step 3: Use the new solution in step 2 as the initial variable of the coordinate descent method, set the convergence constant, and update the variables.

[0012] (1) Calculate the objective function value based on the initialization parameters;

[0013] (2) Solve to minimize the objective function to obtain the optimal solution;

[0014] (3) Update the pulse index that needs to be optimized;

[0015] (4) Determine whether the termination condition is met.

[0016] Step 4: Determine the current optimal solution based on the objective function increment.

[0017] Step 5: Determine the updated state or output solution based on the convergence criterion.

[0018] Compared with the prior art, the present invention has the following significant advantages: compared with using the simulated annealing algorithm alone, after using the simulated annealing algorithm to update the new solution, the present invention uses the coordinate descent method to further optimize the new solution, so that the objective function is accelerated to decrease, and the probability of obtaining the optimal solution that meets the requirements is greatly improved; compared with using the coordinate descent algorithm alone, the present invention uses a group of coding sequences obtained by the simulated annealing algorithm as the initial parameters of the coordinate descent method, which greatly solves the problem that the coordinate descent method is sensitive to the initial value. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is the principle block diagram of the parallel accelerated descent algorithm. DETAILED DESCRIPTION

[0020] The present invention is further described below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited by the implementation regulations. The present invention calculates a new solution based on the criterion of minimizing the autocorrelation peak sidelobe level, and uses the coordinate descent method to optimize the new solution, verifies whether the new solution is acceptable, and outputs the optimal solution after the termination condition is met. The specific steps are:

[0021] Step 1: Set the initial temperature, cooling times M, and initial multi-phase encoding group.

[0022] Step 2: Generate multiple new solutions based on the current temperature.

[0023] (1) Calculate the peak sidelobe level of the autocorrelation function of the frequency-phase coded signal

[0024]

[0025] in k=0,1,2,…,N s -1 is the autocorrelation function of the frequency-phase coded signal, S represents the set of delays corresponding to the side lobes of the autocorrelation function, s i and i+k represents the i-th and i+k-th components in the set S;

[0026] (2) Calculate the peak level of the cross-correlation function between pulses of the frequency-phase coded signal

[0027]

[0028] in k=-N+1,…,-1,0,1,2,…,N-1 is the mth and lth groups of phase coding sequences x m and x l The cross-correlation function of the corresponding signal, N is the length of the phase encoding sequence

[0029] (3) Solve the Pareto optimization problem and generate a new solution X':

[0030]

[0031] Where P g represents the Pareto optimization problem, x m,i represents the i-th component of the m-th phase-coded sequence, f2(x1,x2,…,x Np ) is the objective function:

[0032]

[0033] Where α is a positive constant, which indicates the degree of minimization of PSL and CPL in the optimization problem. When α<1, the degree of minimization of PSL is lower than that of CPL; when α>1, the degree of minimization of CPL is lower than that of PSL.

[0034] Step 3: Use the new solution in step 2 as the initial variable of the coordinate descent method, set the convergence constant, and update the variables.

[0035] (1) By initialization parameters Calculate the objective function value Let d = 1, t = 1, n = 1, where i = 1, 2, ..., N p ;

[0036] (2) Solve to minimize the objective function to obtain the optimal solution

[0037]

[0038] stx t (d)∈X'

[0039] Solve the above formula to get Update solution If d=N, execute (3), otherwise d=d+1, execute (2), where N is the number of polyphase codes;

[0040] (3) Update the pulse index to be optimized: If t = N p , execute (4); otherwise, t = t + 1, d = 1, execute (2);

[0041] (4) Determine whether the termination condition is met: If |f2(x {n} )-f2(x {n-1} )|<ε, output the optimal solution: X * =X {n} , Otherwise, n=n+1, d=1, and execute step (2).

[0042] Step 4: Determine the current optimal solution based on the objective function increment.

[0043] Calculate the increment of the objective function △F=f2(X * )-f2(X0); if △F≤0, accept the newly generated optimal solution X * is the current optimal solution X0=X * ; If △F>0, calculate p=exp(△F); if p>rand(0,1), accept the newly generated optimal solution X' as the current optimal solution X0=X * ; Otherwise X0 does not change.

[0044] Step 5: Determine the updated state or output solution based on the convergence criterion.

[0045] Calculate the objective function increment of the current new solution. If |f2(X * )-f2(X0)| does not meet the convergence criteria, then update the temperature according to the temperature update function, the number of cooling times M=M+1, and go to step 1. If |f2(X * )-f2(X0)| satisfies the convergence criterion, then the current optimal solution is output.

Claims

1. A phase-coded sequence parallel accelerated descent optimization design method, Features: Step 1: Set the initial temperature, the number of cooling times M, and the initial multi-phase encoding group; Step 2: Generate multiple new solutions based on the current temperature; (1) Calculate the peak sidelobe level of the autocorrelation function of the frequency-phase coded signal; (2) Calculating the peak level of the inter-pulse cross-correlation function of the frequency-phase coded signal; (3) Solve the Pareto optimization problem and generate a new solution X'; Step 3: Use the new solution X' in step 2 as the initial variable of the coordinate descent method, set the convergence constant, and update the variable; (1) By initialization parameter x i {0} ∈X' calculates the objective function value f 2 (x 1 {0},x 2 {0},…,x Np {0}), and let d = 1, t = 1, n = 1, where i = 1, 2, ..., N p , N p Indicates the number of phase coding sequences; (2) Solve to minimize the objective function to obtain the optimal solution x t {n}* (d), update the solution x t {n} (d) = x t {n}* (d), if d = N, execute (3), otherwise d = d + 1, execute (2), where N is the length of the phase encoding sequence; (3) Update the pulse index to be optimized: If t = N p , execute (4); otherwise, t = t + 1, d = 1, execute (2); (4) Determine whether the termination condition is met: If |f 2 (x {n} )-f 2 (x {n-1} )|<ε, output the optimal solution: X * =X {n} , Otherwise, n=n+1, d=1, and execute step (2); Step 4: Determine the current optimal solution based on the objective function increment; Step 5: Determine the updated state or output solution based on the convergence criterion.

2. According to claim 1, a phase-coded sequence parallel accelerated descent optimization design method, Features: The method for calculating the peak sidelobe level of the autocorrelation function of the intermediate frequency phase coded signal in step 2 (1) includes: In the formula k = -N+1, ..., -1, 0, 1, 2, ..., N-1 is the autocorrelation function of the frequency-phase coded signal, S represents the set of delays corresponding to the side lobes of the autocorrelation function, s i and i+k represents the i-th and i+k-th components in the set S, and N is the length of the phase encoding sequence.

3. The phase-coded sequence parallel accelerated descent optimization design method according to claim 2, Features: The method for calculating the peak level of the cross-correlation function between pulses of the intermediate frequency phase coded signal in step 2 (2) includes: In the formula k=-N+1,…,-1,0,1,2,…,N-1 is the mth and lth groups of phase coding sequences x m and x l The corresponding signal cross-correlation function, N is the length of the phase encoding sequence.

4. The phase-coded sequence parallel accelerated descent optimization design method according to claim 3, Features: The calculation method for minimizing the Pareto optimization problem in step 2(3) includes: Where P g represents the Pareto optimization problem, x m,i represents the i-th component of the m-th phase-coded sequence, is the objective function, N p Indicates the number of phase encoding sequences.

5. The phase-coded sequence parallel accelerated descent optimization design method according to claim 1, Features: In step 4, the objective function increment ΔF is compared with 0. If ΔF ≤ 0, the newly generated optimal solution X is accepted. * is the current optimal solution X 0 =X * ; If ΔF>0, calculate p=exp(ΔF); if p>rand(0,1), accept the newly generated optimal solution X' as the current optimal solution X 0 =X * ; otherwise X 0 No change.

6. The phase-coded sequence parallel accelerated descent optimization design method according to claim 1, Features: In step 5, the objective function increment of the current new solution is calculated. If |f 2 (X * )-f 2 (X 0 )| does not meet the convergence criteria, then update the temperature according to the temperature update function, the number of cooling times M = M + 1, go to step 1, if |f 2 (X * )-f 2 (X 0 )|If the convergence criterion is met, the current optimal solution is output.