An antenna design optimization method based on migration optimization and the antenna

By employing migration optimization methods and covariance alignment and joint distribution adaptation techniques, the problems of long design cycles and low efficiency in antenna design are solved, achieving more efficient and accurate antenna optimization.

CN116776735BActive Publication Date: 2026-11-17HEBEI GEO UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310762622.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-06-27
Publication Date
2026-11-17
Estimated Expiration
2043-06-27

AI Technical Summary

Technical Problem

Existing antenna design methods rely on manual experience, have long design cycles, low efficiency, and are difficult to meet complex antenna optimization requirements, such as wide bandwidth, low VSWR, high energy gain in the relevant direction, small size, light weight, and low cost.

Method used

A migration optimization-based approach is adopted, utilizing covariance alignment and joint distribution adaptation techniques. Through mathematical modeling and optimization problem modeling, combined with differential evolution algorithm, antenna design optimization is achieved.

Benefits of technology

This allows for obtaining a better antenna design in a shorter time, improving design efficiency and accuracy, and solving the problems of low efficiency and insufficient accuracy in existing methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116776735B_ABST
    Figure CN116776735B_ABST
Patent Text Reader

Abstract

This application provides an antenna design optimization method and antenna based on transfer optimization, comprising: mathematical modeling based on the structural characteristics and performance requirements of the antenna to be optimized to determine the optimization problem model; determining the decision vector and decision space of the optimization problem model based on the structural parameters and parameter value ranges of the antenna to be optimized; determining the constraints of the optimization problem model based on the performance requirements of the antenna to be optimized; and solving the optimization problem model using a transfer optimization method based on covariance alignment and joint distribution adaptation, outputting the optimal result, thereby achieving the design optimization of the antenna to be optimized. By meeting the task requirements in antenna design optimization, the transfer optimization method based on covariance alignment and joint distribution adaptation is used to achieve antenna design optimization.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of antenna design technology, and more specifically, to an antenna design optimization method and antenna based on migration optimization. Background Technology

[0002] Wireless communication plays an increasingly important role in various fields, and antennas, as the core component of wireless communication, have increasingly high and diverse design requirements. Previous antenna design methods, primarily manual with simulation software as a supplement, relied heavily on engineers' experience, resulting in long design cycles, low efficiency, and difficulty in adjustments—methods that are no longer suitable for current needs. Evolutionary algorithms, due to their inherent intelligence, have demonstrated outstanding capabilities in solving antenna optimization problems.

[0003] Antenna optimization design is a complex problem that requires satisfying numerous conditions, including wide bandwidth, low VSWR, high energy gain in the relevant direction, small size, light weight, and low cost. The resulting mathematical model is a nonlinear optimization problem with many variables and complex constraints. This type of optimization problem is prevalent in other fields and requires high-performance evolutionary algorithms for solving; however, existing evolutionary algorithms are insufficient to meet the optimization needs of antenna design. Summary of the Invention

[0004] The purpose of this application is to provide an antenna design optimization method and antenna based on migration optimization, so as to achieve antenna design optimization by using migration optimization methods of covariance alignment and joint distribution adaptation to meet the task requirements in antenna design optimization.

[0005] To achieve the above objectives, the embodiments of this application are implemented in the following manner:

[0006] In a first aspect, embodiments of this application provide an antenna design optimization method based on migration optimization, comprising: performing mathematical modeling based on the structural characteristics and performance requirements of the antenna to be optimized to determine the optimization problem model; determining the decision vector and decision space of the optimization problem model based on the structural parameters and parameter value ranges of the antenna to be optimized; determining the constraints of the optimization problem model based on the performance requirements of the antenna to be optimized; and solving the optimization problem model using a migration optimization method based on covariance alignment and joint distribution adaptation, outputting the optimal result, thereby achieving the design optimization of the antenna to be optimized.

[0007] In conjunction with the first aspect, in the first possible implementation of the first aspect, a migration optimization method based on covariance alignment and joint distribution adaptation is used to solve the optimization problem model, including: determining the task OP1 associated with task OP2 and the objective function f2 in task OP2 based on the task OP2 to be solved by the optimization problem model, where the objective function f1 in task OP1 is determined, and the objective function of task OP2 and task OP1 is both 1; setting the basic parameters of the differential evolution algorithm to obtain the initial population corresponding to task OP2, and evaluating the individuals in the initial population to obtain each The fitness values ​​of individuals are used to generate a basic task evolution optimizer, whose basic parameters include the optimization termination condition, the migration condition, and the number of solutions N required to construct the search space mapping. The covariance alignment algorithm is used to determine the mapping relationship between the search spaces corresponding to tasks OP1 and OP2. The joint distribution adaptation method is used to determine the mapping relationship between the search spaces corresponding to tasks OP1 and OP2, resulting in the target problem population after migration. The task evolution optimizer is then used to evolve and optimize tasks OP1 and OP2 until the optimization termination condition is met, yielding the final optimized solution for task OP2.

[0008] In conjunction with the first possible implementation of the first aspect, in the second possible implementation of the first aspect, a covariance alignment algorithm is used to determine the mapping relationship between the search spaces corresponding to tasks OP1 and OP2, including:

[0009] For the optimized population corresponding to task OP1, individual genes are extracted to obtain the source domain data vector D. S For the initial population corresponding to task OP2, extract individual genes to obtain the target domain data vector D. T ;

[0010] For source domain data vector D S and target domain data vector D T Perform feature normalization to make:

[0011] μ s =μ t =0, C S ≠C T ,

[0012] Where, μ s μ t The source domain data vectors D are respectively S and target domain data vector D T The mean, C S C T The source domain data vectors D are respectively S and target domain data vector D T The covariance matrix;

[0013] For source domain data vector D S Applying a linear transformation A, and using the Frobenius norm as a measure of matrix distance, i.e.:

[0014]

[0015] Among them, China The transformed source domain feature D S The covariance of A, Denotes the Frobenius norm of a matrix;

[0016] Source domain data D S Target domain data D T As input to the covariance alignment method, the two sets of data are subjected to unsupervised domain adaptation covariance alignment, i.e.:

[0017] C S =cov(D S )+eye(size(D S ,2)),

[0018] C T =cov(D T )+eye(size(D T ,2)),

[0019]

[0020]

[0021] Output adjusted target domain data

[0022] Combining the second possible implementation of the first aspect, in the third possible implementation of the first aspect, the joint distribution adaptation method is used to determine the mapping relationship between the search spaces corresponding to task OP1 and task OP2, resulting in the target problem population after migration, including:

[0023] For the optimized population corresponding to task OP1, individual genes are extracted to obtain the source domain data X. S And transform the fitness values ​​of individuals into source domain data X. S The tag y s ;

[0024] target domain data As the initial population for task OP2, the task evolution optimizer is used to perform evolutionary optimization of task OP2 for g1 iterations, resulting in the optimized target domain data X. T ;

[0025] Using joint distribution fitting, the objective function of joint distribution fitting is obtained:

[0026]

[0027] in: It is a regularization term;

[0028] Source domain data X S The label y of the source domain data s Target domain data X T As input to the joint distribution adaptation algorithm, using the objective function of joint distribution adaptation, X is transformed through the transformation matrix A. T The transformation is performed to obtain the new target domain data X. T_new ;

[0029] Transformed target domain data X T_new As individual genes construct a new population, replacing the target problem population, which then becomes the target problem population after migration.

[0030] In conjunction with the third possible implementation of the first aspect, in the fourth possible implementation of the first aspect, the task evolution optimizer is used to perform evolution optimization on tasks OP1 and OP2 until the optimization termination condition is met, and the final optimized solution of task OP2 is obtained. This includes: using the task evolution optimizer to perform evolution optimization on tasks OP1 and OP2, and completing the evolution task as required; until the optimization termination condition is met, that is, evolution through generations, and the final optimized solution of task OP2 is obtained.

[0031] In conjunction with the first aspect, in the fifth possible implementation of the first aspect, a transfer optimization method based on covariance alignment and joint distribution adaptation is used to solve the optimization problem model. This includes: determining the objective function f2 in tasks OP1 and OP2 that are to be solved by the optimization problem model, identifying the task OP1 associated with task OP2, and the objective function f1 in task OP1, where the objective function of both task OP2 and task OP1 is 1; setting the basic parameters of the differential evolution algorithm to obtain the initial population corresponding to task OP2, evaluating the individuals in the initial population to obtain the fitness value of each individual, and generating a basic task evolution optimizer. The basic parameters include the optimization termination condition, the migration condition, and the number of solutions N required to construct the search space mapping. For the initial population of task OP2, the task evolution optimizer is used for g1 iterations to obtain the existing population corresponding to task OP2. The covariance alignment algorithm is used to determine the mapping relationship between the search spaces corresponding to tasks OP1 and OP2. The joint distribution adaptation method is used to determine the mapping relationship between the search spaces corresponding to tasks OP1 and OP2, obtaining the migrated target problem population. The task evolution optimizer is used to perform evolutionary optimization on tasks OP1 and OP2 until the optimization termination condition is met, obtaining the final optimized solution for task OP2.

[0032] In conjunction with the fifth possible implementation of the first aspect, the sixth possible implementation of the first aspect utilizes a covariance alignment algorithm to determine the mapping relationship between the search spaces corresponding to tasks OP1 and OP2, including:

[0033] For the optimized population corresponding to task OP1, individual genes are extracted to obtain the source domain data vector D. S For the existing population corresponding to task OP2, extract individual genes to obtain the target domain data vector D. T ;

[0034] For source domain data vector D S and target domain data vector D T Perform feature normalization to make:

[0035] μ s =μ t =0, C S ≠C T ,

[0036] Where, μ s μ t The source domain data vectors D are respectively S and target domain data vector D T The mean, C S C T The source domain data vectors D are respectively Sand target domain data vector D T The covariance matrix;

[0037] For source domain data vector D S Applying a linear transformation A, and using the Frobenius norm as a measure of matrix distance, i.e.:

[0038]

[0039] Among them, China The transformed source domain feature D S The covariance of A, Denotes the Frobenius norm of a matrix;

[0040] Source domain data D S Target domain data D T As input to the covariance alignment method, the two sets of data are subjected to unsupervised domain adaptation covariance alignment, i.e.:

[0041] C S =cov(D S )+eye(size(D S ,2)),

[0042] C T =cov(D T )+eye(size(D T ,2)),

[0043]

[0044]

[0045] Output adjusted target domain data

[0046] Combining the sixth possible implementation of the first aspect, in the seventh possible implementation of the first aspect, the joint distribution adaptation method is used to determine the mapping relationship between the search spaces corresponding to task OP1 and task OP2, resulting in the target problem population after the transition, including:

[0047] For the optimized population corresponding to task OP1, individual genes are extracted to obtain the source domain data X. S And transform the fitness values ​​of individuals into source domain data X. S The tag y s ;

[0048] target domain data As part of the existing population for task OP2, individual genes are extracted to obtain the target domain data X. T ;

[0049] Using joint distribution fitting, the objective function of joint distribution fitting is obtained:

[0050]

[0051] in: It is a regularization term;

[0052] Source domain data X S The label y of the source domain data s Target domain data X T As input to the joint distribution adaptation algorithm, using the objective function of joint distribution adaptation, X is transformed through the transformation matrix A. T The transformation is performed to obtain the new target domain data X. T_new ;

[0053] Transformed target domain data X T_new As individual genes construct a new population, replacing the target problem population, which then becomes the target problem population after migration.

[0054] In conjunction with the seventh possible implementation of the first aspect, in the eighth possible implementation of the first aspect, the task evolution optimizer is used to perform evolution optimization on tasks OP1 and OP2 until the optimization termination condition is met, and the final optimized solution of task OP2 is obtained. This includes: using the task evolution optimizer to continue to perform evolution optimization on tasks OP1 and OP2, and completing the evolution task as required; until the optimization termination condition is met, that is, evolving to generation g2, and obtaining the final optimized solution of task OP2.

[0055] Secondly, embodiments of this application provide an antenna, which is designed and optimized based on the antenna design optimization method based on migration optimization, as described in any one of the first aspects or possible implementations of the first aspect.

[0056] Beneficial effects:

[0057] 1. Based on the structural characteristics and performance requirements of the antenna to be optimized, mathematical modeling is performed to determine the optimization problem model. Using the structural parameters and parameter ranges of the antenna to be optimized, the decision vector and decision space of the optimization problem model are determined. Using the performance requirements of the antenna to be optimized, the constraints of the optimization problem model are determined. Then, a transfer optimization method based on covariance alignment and joint distribution adaptation is used to solve the optimization problem model and output the optimal result, achieving the design optimization of the antenna to be optimized. The transfer optimization method based on covariance alignment and joint distribution adaptation uses the optimal population of the source problem (which is related to the target problem) and the randomly generated initial population of the target problem. First, the covariance of the source data and the target data is aligned. Then, the joint distribution adaptation method is used to map to a high-dimensional Hilbert space. The new population obtained after alignment and mapping replaces the original population of the target problem to complete the subsequent evolutionary operation. This allows for a simpler computational method when solving the antenna design optimization problem model, with no special requirements on the objective function of the problem model. Furthermore, by utilizing prior knowledge, it can, to some extent, escape local optima and obtain a better solution in a faster solution time, thus achieving antenna design optimization.

[0058] 2. In the process of solving the problem model of antenna design optimization, the direct one-way transfer differential evolution algorithm is used to perform transfer in the early stage, which helps the target problem learn common features from a source problem that is related to it. The existing data and knowledge in the source task are used to help solve the current problem. Therefore, it effectively solves the problems of low efficiency, insufficient accuracy and low utilization of relevant information in existing methods, thereby improving the solution efficiency of the algorithm and realizing antenna design optimization more efficiently.

[0059] 3. In solving the antenna design optimization problem model, the one-way migration differential evolution algorithm can be used in the iterative process. Migration occurs during the iteration process. At the beginning of the target problem solution, the differential evolution algorithm performs a certain number of random searches to fully explore the possible regions. Then, knowledge learned from the source problem is incorporated to guide the population to converge as quickly as possible in the remaining iterations. This effectively solves the problems of low efficiency, insufficient accuracy, and low utilization of relevant information in existing methods, thereby improving the solution efficiency of the algorithm and achieving antenna design optimization more efficiently.

[0060] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0061] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart illustrating an antenna design optimization method based on migration optimization, provided as an embodiment of this application.

[0063] Figure 2 This is a schematic diagram of an antenna.

[0064] Figure 3 This is a diagram of the patch structure on the antenna.

[0065] Figure 4 This is a diagram of the antenna under-patch structure.

[0066] Figure 5 This is a framework diagram of the solution process using a migration optimization method based on covariance alignment and joint distribution adaptation. Detailed Implementation

[0067] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0068] Please see Figure 1 , Figure 1 This is a flowchart illustrating an antenna design optimization method based on migration optimization, provided in an embodiment of this application. In this embodiment, to optimize the antenna design, the antenna design optimization method based on migration optimization may include steps S10, S20, S30, and S40.

[0069] For antenna design optimization, the structural characteristics and performance requirements of the antenna need to be considered. Therefore, when optimizing the antenna design, steps S10 to S30 can be executed to construct an optimization problem model, determine the decision vector and decision space of the optimization problem model, as well as the constraints of the optimization problem model.

[0070] Step S10: Based on the structural characteristics and performance requirements of the antenna to be optimized, perform mathematical modeling to determine the optimization problem model.

[0071] Step S20: Based on the structural parameters and parameter value range of the antenna to be optimized, determine the decision vector and decision space of the optimization problem model.

[0072] Step S30: Based on the performance requirements of the antenna to be optimized, determine the constraints of the optimization problem model.

[0073] Antenna design optimization is a complex constrained optimization problem with the goal of finding an antenna structure that meets specific requirements. The first step is to analyze the antenna's structural characteristics and performance requirements, thereby creating a mathematical model to form the optimization problem model.

[0074] Problem modeling requires analysis of the following aspects:

[0075] Structural parameters: Analyze the structural parameters of the antenna and the range of their values ​​to form the decision vector and decision space for the problem.

[0076] Performance requirements: Analyze the performance requirements of the antenna to form the constraints of the problem.

[0077] Optimization objective: Analyze the optimization direction of the antenna and formulate the objective function of the problem.

[0078] For example, taking the design optimization of a circularly polarized wide-beam microstrip antenna as an example, the basic structure of the antenna is a square profile plate with two pairs of parallel dipoles, such as... Figure 2-4 As shown, Figure 2 This is a schematic diagram of the antenna structure. Figure 3 This is a diagram of the patch structure on the antenna. Figure 4 This is a diagram of the antenna's bottom patch structure. The antenna structure consists of three layers: a dielectric layer, an upper patch layer, and a lower patch layer. The antenna structure parameters are as follows:

[0079] The dielectric layer is 250 mm long and wide, with a height of Ls. The material is Rogers RO3010(tm), and the dielectric constant is ε. r =10.2.

[0080] The upper patch layer consists of four microstrip lines with phase differences of 0°, 90°, 180°, and 270°, and four L-shaped microstrip sheets, as shown. Figure 2 As shown. The actual length of the microstrip line AB is λ. g / 4, width is W1; actual length of microstrip line BC is λ. g / 4, width is W2; actual length of microstrip line DE is λ. g / 4, width is W2; actual length of microstrip line AD is λ. g *3 / 4, width is W1; where λ g It is the equivalent wavelength with a center frequency of 425MHz.

[0081] The lower patch layer consists of a square patch with a side length of 80mm, four rectangular patches with a width of W3 and a length of L1, and four L-shaped microstrips.

[0082] The L-shaped microstrip patch, with a total length of approximately λ0 / 4, is composed of L2 + Ld = λ0 / 4. The width of the long L2 patch is W2, and the width of the long Ld patch is Wd; where λ0 is the vacuum wavelength at the center frequency of 425MHz. The range of values ​​for the antenna structure parameters is shown in Table 1 below.

[0083] Table 1: Range of Antenna Structure Parameters

[0084]

[0085]

[0086] The antenna performance requirements are shown in Table 2 below:

[0087] Table 2: Antenna Performance Requirements

[0088]

[0089] Based on this, the optimization problem model can be constructed as follows:

[0090]

[0091]

[0092]

[0093]

[0094]

[0095] where 400MHz≤frq≤450MHz,step 12.5MHz, (6)

[0096] -70°≤θ≤70°, step 5°, (7)

[0097]

[0098] The optimization objectives are to maximize gain and minimize VSWR and axial ratio. The objective function (f) is the sum of the weighted regularized values ​​of gain, VSWR, and axial ratio. The constraint (g) consists of the requirements listed in the antenna performance indicators, forming four constraints, each of which has also been weighted.

[0099] Therefore, it is possible to construct an optimization problem model for antenna design optimization, as well as to determine the decision vector and decision space of the optimization problem model, and the constraints of the optimization problem model.

[0100] Then, step S40 can be executed.

[0101] Step S40: Using the migration optimization method based on covariance alignment and joint distribution adaptation, solve the optimization problem model, output the optimal result, and realize the design optimization of the antenna to be optimized.

[0102] In this embodiment, a transfer optimization method based on covariance alignment and joint distribution adaptation is used to solve the optimization problem model. The framework of the solution process is as follows: Figure 5 As shown in the figure. This embodiment provides two different methods for solving the optimization problem model: a direct one-way migration differential evolution method based on covariance alignment and joint distribution adaptation, and a procedural one-way migration differential evolution method based on covariance alignment and joint distribution adaptation.

[0103] First, the solution process of the direct one-way transfer differential evolution method based on covariance alignment and joint distribution adaptation is introduced. This method requires initial transfer (i.e., ...) Figure 5 The source population in the problem migrates in the initial stage.

[0104] Specifically, the direct one-way migration differential evolution method based on covariance alignment and joint distribution adaptation may include steps S411, S412, S413, S414, and S415.

[0105] Step S411: Based on the task OP2 to be solved by the optimization problem model and the objective function f2 in task OP2, determine the task OP1 associated with task OP2 and the objective function f1 in task OP1, where the objective number of both task OP2 and task OP1 is 1.

[0106] Step S412: Set the basic parameters of the differential evolution algorithm to obtain the initial population corresponding to task OP2, evaluate the individuals in the initial population to obtain the fitness value of each individual, and generate the basic task evolution optimizer. The basic parameters include the optimization termination condition, the migration condition, and the number of solutions N required to construct the search space mapping.

[0107] Step S413: Use the covariance alignment algorithm to determine the mapping relationship between the search spaces corresponding to task OP1 and task OP2.

[0108] Step S414: Using the joint distribution adaptation method, determine the mapping relationship between the search spaces corresponding to task OP1 and task OP2, and obtain the target problem population after migration.

[0109] Step S415: Use the task evolution optimizer to perform evolution optimization on tasks OP1 and OP2 until the optimization termination condition is met, and obtain the final optimized solution for task OP2.

[0110] First, based on the task OP2 to be solved in the optimization problem model and the objective function f2 in task OP2, we can determine the associated task OP1 and the objective function f1 in task OP1. The objective function of both task OP2 and task OP1 is 1. Then, we set the basic parameters of the differential evolution algorithm to obtain the initial population corresponding to task OP2. We then evaluate the individuals in the initial population to obtain the fitness value of each individual, generating a basic task evolution optimizer. The basic parameters include the optimization termination condition, the migration condition, and the number of solutions N required to construct the search space mapping.

[0111] It should be noted that, since this is applied to antenna design optimization, the evaluation method needs to be carried out in the following way: First, the individual decodes to form the structural parameter values ​​of the antenna; then, the structural parameter values ​​are input into the control script to generate FEKO files such as *.pre and *.cfx; then, the FEKO antenna simulation is run to obtain the antenna simulation data. After processing by the control script, the antenna performance parameter values ​​are obtained, and the objective function value and constraint condition value are further calculated to obtain the fitness value of the individual.

[0112] Based on this, the specific process of step S413 is as follows:

[0113] For the optimized population corresponding to task OP1, individual genes are extracted to obtain the source domain data vector D. S Furthermore, for the initial population corresponding to task OP2, individual genes are extracted to obtain the target domain data vector D. T .

[0114] Then, the source domain data vector D can be processed. S and target domain data vector D T Perform feature normalization to make:

[0115] μ s =μ t =0, C S ≠C T (9)

[0116] Where, μ s μ t The source domain data vectors D are respectively S and target domain data vector D T The mean, C S C T The source domain data vectors D are respectively S and target domain data vector D T The covariance matrix.

[0117] Then, the source domain data vector D can be processed. SApplying a linear transformation A, and using the Frobenius norm as a measure of matrix distance, i.e.:

[0118]

[0119] Among them, China The transformed source domain feature D S The covariance of A, This represents the Frobenius norm of the matrix.

[0120] Then transfer the source domain data D S Target domain data D T As input to the covariance alignment method, the two sets of data are subjected to unsupervised domain adaptation covariance alignment, i.e.:

[0121] C S =cov(D S )+eye(size(D S ,2)), (11)

[0122] C T =cov(D T )+eye(size(D T ,2)), (12)

[0123]

[0124]

[0125] Therefore, the adjusted target domain data can be output.

[0126] In order to obtain the target problem population after migration, step S414 is as follows:

[0127] For the optimized population corresponding to task OP1, individual genes are extracted to obtain the source domain data X. S And transform the fitness values ​​of individuals into source domain data X. S The tag y s Then the target domain data As the initial population for task OP2, the task evolution optimizer is used to perform evolutionary optimization of task OP2 for g1 iterations, resulting in the optimized target domain data X. T .

[0128] Joint distribution adaptation requires that both marginal distribution adaptation and conditional distribution adaptation between the source and target domain data be reduced to a certain extent. Therefore, the two adaptation methods can be combined to obtain the objective function of joint distribution adaptation.

[0129]

[0130] in: For regularization terms, and M c satisfy:

[0131]

[0132] Among them, D S For source domain data, D T Let n be the target domain data, and m be the number of samples in the source domain data and the target domain data, respectively.

[0133] Based on this, the source domain data X can be... S The label y of the source domain data s Target domain data X T As input to the joint distribution adaptation algorithm, using the objective function of joint distribution adaptation, X is transformed through the transformation matrix A. T The transformation is performed to obtain the new target domain data X. T_new The transformed target domain data X T_new By constructing a new population using individual genes and replacing the target problem population, the migrated target problem population can be obtained.

[0134] After obtaining the target problem population after migration, the task evolution optimizer can be used to perform evolutionary optimization on tasks OP1 and OP2, completing the evolutionary task as required until the optimization termination condition is met, i.e., after generations, to obtain the final optimized solution for task OP2. Finally, the optimal result can be output, thereby realizing the design optimization of the antenna to be optimized.

[0135] In this embodiment, another method can be provided to achieve the optimized design of the antenna, namely the process of one-way migration differential evolution based on covariance alignment and joint distribution adaptation.

[0136] Based on steps S10 to S30, this section introduces the solution process of the process-oriented one-way migration differential evolution method based on covariance alignment and joint distribution adaptation. This method requires migration during the iteration process (i.e., Figure 5 The source problem population in the algorithm migrates only after a certain number of iterations.

[0137] Specifically, the process-oriented unidirectional migration differential evolution method based on covariance alignment and joint distribution adaptation may include steps S421, S422, S423, S424, S425, and S426.

[0138] Step S421: Based on the objective function f2 in task OP1 and task OP2 that the optimization problem model needs to solve, determine the objective function f1 in task OP1 associated with task OP2, where the objective number of both task OP2 and task OP1 is 1.

[0139] Step S422: Set the basic parameters of the differential evolution algorithm to obtain the initial population corresponding to task OP2, evaluate the individuals in the initial population to obtain the fitness value of each individual, and generate the basic task evolution optimizer. The basic parameters include the optimization termination condition, the migration condition, and the number of solutions N required to construct the search space mapping.

[0140] Step S423: For the initial population of task OP2, use the task evolution optimizer to iterate for generation g1 to obtain the current population corresponding to task OP2.

[0141] Step S424: Use the covariance alignment algorithm to determine the mapping relationship between the search spaces corresponding to task OP1 and task OP2.

[0142] Step S425: Using the joint distribution adaptation method, determine the mapping relationship between the search spaces corresponding to task OP1 and task OP2, and obtain the target problem population after migration.

[0143] Step S426: Use the task evolution optimizer to perform evolution optimization on tasks OP1 and OP2 until the optimization termination condition is met, and obtain the final optimized solution for task OP2.

[0144] Steps S421 and S422 are no different from steps S411 and S412 described above, so please refer to the previous description and they will not be repeated here.

[0145] After generating the basic task evolution optimizer, the initial population for task OP2 can be iterated for g1 generations using the task evolution optimizer to obtain the current population corresponding to task OP2. Then, the covariance alignment algorithm needs to be used to determine the mapping relationship between the search spaces corresponding to tasks OP1 and OP2, as follows:

[0146] We need to extract individual genes from the optimized population corresponding to task OP1 to obtain the source domain data vector D. S Furthermore, for the existing population corresponding to task OP2, individual genes are extracted to obtain the target domain data vector D. T .

[0147] Therefore, the source domain data vector D can be processed. S and target domain data vector D T Perform feature normalization to make:

[0148] μ s =μ t =0, C S ≠C T (17)

[0149] Where, μ s μ t The source domain data vectors D are respectively S and target domain data vector D T The mean, C S C T The source domain data vectors D are respectively S and target domain data vector D T The covariance matrix.

[0150] Then, the source domain data vector D can be processed. S Applying a linear transformation A, and using the Frobenius norm as a measure of matrix distance, i.e.:

[0151]

[0152] Among them, China The transformed source domain feature D S The covariance of A, This represents the Frobenius norm of the matrix.

[0153] Then, the source domain data D can be... S Target domain data D T As input to the covariance alignment method, the two sets of data are subjected to unsupervised domain adaptation covariance alignment, i.e.:

[0154] C S =cov(D S )+eye(size(D S ,2)), (19)

[0155] C T =cov(D T )+eye(size(D T ,2)), (20)

[0156]

[0157]

[0158] Based on this, the adjusted target domain data can be output.

[0159] Then, the joint distribution adaptation method can be used to determine the mapping relationship between the search spaces corresponding to task OP1 and task OP2, and obtain the target problem population after migration. The specific process is as follows:

[0160] First, we need to extract individual genes from the optimized population corresponding to task OP1 to obtain the source domain data X. S And transform the fitness values ​​of individuals into source domain data X. S The tag y s ; and, target domain data As part of the existing population for task OP2, individual genes are extracted to obtain the target domain data X. T .

[0161] Then, using joint distribution adaptation, the objective function of joint distribution adaptation is obtained:

[0162]

[0163] in: This is a regularization term.

[0164] Therefore, the source domain data X can be... S The label y of the source domain data s Target domain data X T As input to the joint distribution adaptation algorithm, using the objective function of joint distribution adaptation, X is transformed through the transformation matrix A. T The transformation is performed to obtain the new target domain data X. T_new Then transform the target domain data X T_new By constructing a new population using individual genes and replacing the target problem population, the migrated target problem population can be obtained.

[0165] After obtaining the target problem population after migration, the task evolution optimizer can be used to continue evolutionary optimization of tasks OP1 and OP2, completing the evolutionary tasks as required until the optimization termination condition is met, i.e., evolving through generation g2, to obtain the final optimized solution for task OP2. Here, g1 and g2 satisfy the following relationship:

[0166] g1 + g2 = generation, (24)

[0167] Here, generation refers to the total number of iterations, which is the final number of iterations required in the direct unidirectional migration differential evolution method based on covariance alignment and joint distribution adaptation.

[0168] Similarly, the optimal result can be output in the end, thereby achieving the design optimization of the antenna to be optimized.

[0169] To verify the effectiveness of the scheme, this embodiment uses the test functions in Table 3 to evaluate the performance of the migration optimization method based on covariance alignment and joint distribution adaptation. Each set of test functions consists of two common single-objective optimization functions. In each set of test functions, Q1 is a simple function and Q2 is a complex function. For each set of functions, the function type and the degree of intersection and similarity between functions are given; a higher similarity value indicates a higher degree of similarity between functions.

[0170] Table 3. Test Question Attributes

[0171]

[0172]

[0173] The specific parameter settings for the experiment are shown in Table 4:

[0174] Table 4. Specific Parameter Settings for the Experiment

[0175]

[0176] The average target value and standard deviation of the migration optimization method based on covariance alignment and joint distribution adaptation and the comparison algorithm obtained in the test function experiment are shown in Table 5. The optimal result of the algorithm in solving the problem is indicated in bold.

[0177] Table 5. Average objective value and standard deviation of each algorithm on a single objective problem.

[0178]

[0179]

[0180] Table 5 shows that in 7 out of the 9 test functions, the transfer optimization method based on covariance alignment and joint distribution adaptation yields better solutions. Experimental results demonstrate that using this method improves the algorithm's optimization performance. In the 7 sets of experiments yielding better solutions, for problems 1, 2, 4, 5, 7, and 8, the solutions obtained using the proposed transfer optimization method based on covariance alignment and joint distribution adaptation (including the direct one-way transfer differential evolution method and the procedural one-way transfer differential evolution method based on covariance alignment and joint distribution adaptation) are all superior to those obtained without the transfer learning algorithm. The experimental comparison further demonstrates the effectiveness of the proposed transfer optimization method based on covariance alignment and joint distribution adaptation, which is particularly suitable for solving single-objective optimization problems in antenna design optimization.

[0181] This application also provides an antenna, which is designed and optimized by an antenna design optimization method based on migration optimization.

[0182] In summary, this application provides an antenna design optimization method and antenna based on migration optimization:

[0183] 1. Based on the structural characteristics and performance requirements of the antenna to be optimized, mathematical modeling is performed to determine the optimization problem model. Using the structural parameters and parameter ranges of the antenna to be optimized, the decision vector and decision space of the optimization problem model are determined. Using the performance requirements of the antenna to be optimized, the constraints of the optimization problem model are determined. Then, a transfer optimization method based on covariance alignment and joint distribution adaptation is used to solve the optimization problem model and output the optimal result, achieving the design optimization of the antenna to be optimized. The transfer optimization method based on covariance alignment and joint distribution adaptation uses the optimal population of the source problem (which is related to the target problem) and the randomly generated initial population of the target problem. First, the covariance of the source data and the target data is aligned. Then, the joint distribution adaptation method is used to map to a high-dimensional Hilbert space. The new population obtained after alignment and mapping replaces the original population of the target problem to complete the subsequent evolutionary operation. This allows for a simpler computational method when solving the antenna design optimization problem model, with no special requirements on the objective function of the problem model. Furthermore, by utilizing prior knowledge, it can, to some extent, escape local optima and obtain a better solution in a faster solution time, thus achieving antenna design optimization.

[0184] 2. In the process of solving the problem model of antenna design optimization, the direct one-way transfer differential evolution algorithm is used to perform transfer in the early stage, which helps the target problem learn common features from a source problem that is related to it. The existing data and knowledge in the source task are used to help solve the current problem. Therefore, it effectively solves the problems of low efficiency, insufficient accuracy and low utilization of relevant information in existing methods, thereby improving the solution efficiency of the algorithm and realizing antenna design optimization more efficiently.

[0185] 3. In solving the antenna design optimization problem model, the one-way migration differential evolution algorithm can be used in the iterative process. Migration occurs during the iteration process. At the beginning of the target problem solution, the differential evolution algorithm performs a certain number of random searches to fully explore the possible regions. Then, knowledge learned from the source problem is incorporated to guide the population to converge as quickly as possible in the remaining iterations. This effectively solves the problems of low efficiency, insufficient accuracy, and low utilization of relevant information in existing methods, thereby improving the solution efficiency of the algorithm and achieving antenna design optimization more efficiently.

[0186] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0187] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for antenna design optimization based on transfer optimization, characterized in that, include: Based on the structural characteristics and performance requirements of the antenna to be optimized, mathematical modeling is performed to determine the optimization problem model; Based on the structural parameters and parameter value range of the antenna to be optimized, the decision vector and decision space of the optimization problem model are determined. Based on the performance requirements of the antenna to be optimized, the constraints of the optimization problem model are determined. The migration optimization method based on covariance alignment and joint distribution adaptation is used to solve the optimization problem model, output the optimal result, and realize the design optimization of the antenna to be optimized. The optimization problem model is solved using a transfer optimization method based on covariance alignment and joint distribution adaptation, including: The task to be solved based on the optimization problem model ,Task The objective function in Determine the relationship with the task Related tasks ,Task The objective function in Among them, the task and tasks The target quantity is 1 for all values; the basic parameters of the differential evolution algorithm are set to obtain the corresponding task. An initial population is established, and the individuals in the initial population are evaluated to obtain the fitness value of each individual. A basic task evolution optimizer is generated, where the basic parameters include the termination condition, the migration condition, and the number of solutions N required to construct the search space mapping. The task is then determined using a covariance alignment algorithm. and tasks The mapping relationship between the corresponding search spaces; the task is determined using the joint distribution adaptation method. and tasks The mapping relationship between the corresponding search spaces is used to obtain the target problem population after migration; the task evolution optimizer is used to optimize the task. and tasks The process continues until the optimization termination condition is met, resulting in the task. The final optimal solution; Alternatively, a transfer optimization method based on covariance alignment and joint distribution adaptation can be used to solve the optimization problem model, including: The task to be solved based on the optimization problem model ,Task The objective function in Determine the relationship with the task Related tasks ,Task The objective function in Among them, the task and tasks The target quantity is 1 for all values; the basic parameters of the differential evolution algorithm are set to obtain the corresponding task. An initial population is established, and the individuals in the initial population are evaluated to obtain the fitness value of each individual, generating a basic task evolution optimizer. The basic parameters include the optimization termination condition, the migration condition, and the number of solutions N required to construct the search space mapping. For the task... The initial population is iterated using the task evolution optimizer. On behalf of, I received the task. The corresponding existing population; using the covariance alignment algorithm, the task is determined. and tasks The mapping relationship between the corresponding search spaces; the task is determined using the joint distribution adaptation method. and tasks The mapping relationship between the corresponding search spaces is used to obtain the target problem population after migration; the task evolution optimizer is used to optimize the task. and tasks The process continues until the optimization termination condition is met, resulting in the task. The final optimized solution.

2. The antenna design optimization method based on migration optimization according to claim 1, characterized in that, Using the covariance alignment algorithm, the task is determined. and tasks The mapping relationships between the corresponding search spaces include: For the task For the corresponding optimized population, individual genes are extracted to obtain the source domain data vector. ; Regarding the task For the corresponding initial population, individual genes are extracted to obtain the target domain data vector. ; For source domain data vectors and target domain data vector Perform feature normalization to make: , , in, , Source domain data vectors and target domain data vector The mean, , Source domain data vectors and target domain data vector The covariance matrix; For source domain data vectors Applying a linear transformation A, and using the Frobenius norm as a measure of matrix distance, i.e.: , in, The transformed source domain features covariance, Denotes the Frobenius norm of a matrix; Source domain data Target domain data As input to the covariance alignment method, the two sets of data are subjected to unsupervised domain adaptation covariance alignment, i.e.: , , , , Output adjusted target domain data .

3. The antenna design optimization method based on migration optimization according to claim 2, characterized in that, The task is determined using the joint distributed adaptation method. and tasks The mapping relationships between the corresponding search spaces yield the target problem population after migration, including: For the task For the corresponding optimized population, individual genes are extracted to obtain source domain data. And transform the fitness values ​​of individuals into source domain data. tags ; target domain data As a task The initial population, using a task evolution optimizer for task optimization. The number of iterations is Through repeated evolutionary optimizations, the target domain data is obtained after iterative optimization. ; Using joint distribution fitting, the objective function of joint distribution fitting is obtained: , in: It is a regularization term; Source domain data Tags of source domain data Target domain data As input to the joint distribution adaptation algorithm, the objective function of the joint distribution adaptation is used to transform matrix A to... Perform the transformation to obtain new target domain data. ; Transformed target domain data As individual genes construct a new population, replacing the target problem population, which then becomes the target problem population after migration.

4. The antenna design optimization method based on migration optimization according to claim 3, characterized in that, Using the task evolution optimizer to optimize tasks and tasks The process continues until the optimization termination condition is met, resulting in the task. The final optimized solution includes: Using the task evolution optimizer to optimize tasks and tasks Perform evolutionary optimization and complete the evolutionary task as required; Until the optimization termination condition is met, i.e., the evolution continues through generations, the task is obtained. The final optimized solution.

5. The antenna design optimization method based on migration optimization according to claim 1, characterized in that, Using the covariance alignment algorithm, the task is determined. and tasks The mapping relationships between the corresponding search spaces include: For the task For the corresponding optimized population, individual genes are extracted to obtain the source domain data vector. ; Regarding the task For the corresponding existing population, extract individual genes to obtain the target domain data vector. ; For source domain data vectors and target domain data vector Perform feature normalization to make: , , in, , Source domain data vectors and target domain data vector The mean, , Source domain data vectors and target domain data vector The covariance matrix; For source domain data vectors Applying a linear transformation A, and using the Frobenius norm as a measure of matrix distance, i.e.: , in, The transformed source domain features covariance, Denotes the Frobenius norm of a matrix; Source domain data Target domain data As input to the covariance alignment method, the two sets of data are subjected to unsupervised domain adaptation covariance alignment, i.e.: , , , , Output adjusted target domain data .

6. The antenna design optimization method based on migration optimization according to claim 5, characterized in that, The task is determined using the joint distributed adaptation method. and tasks The mapping relationships between the corresponding search spaces yield the target problem population after migration, including: For the task For the corresponding optimized population, individual genes are extracted to obtain source domain data. And transform the fitness values ​​of individuals into source domain data. tags ; target domain data As a task From the existing population, extract individual genes to obtain target domain data. ; Using joint distribution fitting, the objective function of joint distribution fitting is obtained: , in: It is a regularization term; Source domain data Tags of source domain data Target domain data As input to the joint distribution adaptation algorithm, the objective function of the joint distribution adaptation is used to transform matrix A to... Perform the transformation to obtain new target domain data. ; Transformed target domain data As individual genes construct a new population, replacing the target problem population, which then becomes the target problem population after migration.

7. The antenna design optimization method based on migration optimization according to claim 6, characterized in that, Using the task evolution optimizer to optimize tasks and tasks The process continues until the optimization termination condition is met, resulting in the task. The final optimized solution includes: Using the task evolution optimizer to optimize tasks and tasks Continue to optimize the evolution and complete the evolutionary task as required; Until the optimization termination condition is met, i.e., evolution. On behalf of, I received the task. The final optimized solution.

8. An antenna, characterized in that, The antenna is designed and optimized based on the antenna design optimization method based on migration optimization as described in any one of claims 1 to 7.