Subarray-level distributed frequency-controlled array sidelobe suppression method based on improved genetic algorithm

Abstract

The present invention relates to a subarray-level distributed frequency-controlled array sidelobe suppression method based on an improved genetic algorithm, and belongs to the field of array antenna signal processing. The method uses an improved genetic algorithm to simultaneously suppress the grating lobes or high sidelobes of the distributed frequency-controlled array in the angle dimension and the distance dimension; the improved genetic algorithm is a niche genetic algorithm based on a pre-selection mechanism, and a competition mechanism is introduced between the parent and the offspring. Only when the fitness value of the offspring individual is higher than that of the parent individual, the offspring individual can replace its parent individual accordingly, otherwise the parent individual is still retained in the next generation group, and the local and global optimal solutions are optimized in the feasible solution space; and taboo search is introduced to speed up the convergence speed, avoid the genetic algorithm from being premature and stagnant, and thus jump out of the local optimal solution. The present invention improves the optimization results of the subarray position and subarray frequency deviation of the subarray-level distributed frequency-controlled array, and achieves better sidelobe suppression results from the distance dimension and the angle dimension.

Description

Technical Field

[0001] The invention belongs to the field of array antenna signal processing and relates to a subarray-level distributed frequency-controlled array sidelobe suppression method based on an improved genetic algorithm. Background Art

[0002] The driving force for the advancement, development and renewal of radar technology comes from the diversification of radar application scenarios and the complexity of actual functional requirements. In particular, in response to complex battlefield environments and threats from enemy targets, radar systems must have long-range detection capabilities and accurate target positioning capabilities, which means that radar antennas need to meet the conditions of their own high gain and large-scale physical array aperture. However, in actual engineering, due to the physical structure limitations of the carrier platform, it is difficult to obtain sufficient space for the effective installation of large-scale antenna arrays. Therefore, it is possible to consider using multiple distributed small-aperture subarrays to achieve coherent collaborative processing of signals, that is, the subarray units can scan simultaneously with the main array to expand the antenna array aperture, thereby equivalent to a high-performance, high-power radar that can meet the high requirements of long-range detection and high-precision tracking and identification. These excellent performances have made the application of distributed arrays one of the research hotspots in recent years.

[0003] The baseline length between sub-arrays in a distributed array structure is usually much longer than half a wavelength, which will lead to the appearance of grating lobes or high side lobes, greatly reducing the accuracy of angle measurement and making the radar more susceptible to other interference. Therefore, how to effectively solve the grating lobe or high side lobe suppression problem in distributed arrays becomes the key.

[0004] A large number of studies have shown that by optimizing the distributed array layout structure, the periodicity of the array element arrangement can be broken, thereby achieving the purpose of suppressing grating lobes or high side lobes, that is, optimizing the position of the subarray nodes in the distributed array so that the subarray node position presents non-periodic characteristics. The more well-known one is An Zhengshuai from the 20th Institute of China Electronics Technology Group Corporation, who first established the topological structure of the distributed array antenna and proposed an improved genetic algorithm with a perturbation strategy to optimize the position of the distributed array at the subarray level, effectively suppressing the grating lobes of the distributed array, and proved the feasibility and effectiveness of the algorithm in the field of array antenna signal processing. Wang Lu from the same unit combined the distributed array structure with the directional pattern product theorem and proposed a distributed array two-way directional pattern design method. By integrating the array factor and the subarray directional pattern function, and using the Minimax criterion to design the transmission directional pattern weight to effectively suppress grating lobes and side lobes, the simulation verified the correctness and practicality of the proposed method. In the same year, the author proposed a distributed array directional pattern design method based on non-periodic arrays and random arrays, which more effectively suppressed grating lobes and side lobes. Chen Kesong from the University of Electronic Science and Technology of China proposed a modified genetic algorithm. Under the premise of comprehensive array element number, array aperture constraints and array element minimum spacing constraints, this algorithm was used to integrate sparse antenna arrays to effectively suppress peak sidelobe levels, greatly meeting the requirements of engineering design.

[0005] However, most of the current research is only for distributed phased arrays, and the integrated beam pattern of distributed phased arrays is only related to the angle dimension. Therefore, when it comes to applications related to distance-related interference or clutter suppression, distributed phased arrays are still helpless. These application scenarios are indispensable for the analysis of distance parameters, which requires that the radar's transmit beam pattern can have good spatial resolution in the distance dimension. In essence, it is necessary to change the radar system from the transmission system so that it can obtain the ability of distance resolution, so as to form an angle-distance joint beam pattern, so as to achieve the suppression of clutter and interference in the distance dimension. Therefore, it is of great significance to introduce a new type of array radar with distance dependence, the frequency-controlled array radar, into the field of radar detection. The main difference between the frequency-controlled array and the traditional phased array is that the introduction of a small frequency offset causes each array element to have a different transmission frequency. When the frequency offset is zero, the frequency-controlled array is simplified to a phased array, so the frequency-controlled array can be regarded as an extension of the phased array, while the phased array is a special case of the frequency-controlled array. It can be seen that frequency-controlled array radar not only has all the functional characteristics of phased array radar, but also has broad application potential in the fields of range-dependent beamforming, target detection, sidelobe suppression, electronic countermeasures, etc.

[0006] For the new array radar technology, there is a frequency offset much smaller than the center carrier between the transmitting array elements of the frequency-controlled array radar, which makes the phase superposition relationship of each carrier corresponding to different distances different, resulting in changes in peaks and troughs in the distance dimension, and then the main lobe and grating lobe of the distance dimension, or high side lobes in the distance dimension, thus producing a distance-dependent transmitting beam pattern, which overcomes the problem that traditional phased array radar cannot effectively control the distance pointing of the transmitting beam, but also brings the problem of grating lobe or high side lobe suppression in the distance dimension. Therefore, for the frequency-controlled array, it is necessary to consider suppressing the sidelobe suppression of the angle dimension and the distance dimension at the same time to effectively enhance the comprehensive performance of the frequency-controlled array beam pattern.

[0007] A large number of existing studies have shown that there are two major solutions to the distance-angle coupling properties of the frequency-controlled array beam pattern: one is to optimize the layout design of the array structure; the other is to perform nonlinear selection of the array frequency offset. Common nonlinear frequency offsets include logarithmic frequency offset, sinusoidal frequency offset, etc. So far, there has been no research on subarray-level distributed frequency-controlled arrays. Therefore, the present invention proposes for the first time to use an improved genetic algorithm that combines a pre-selection mechanism and taboo search to achieve subarray-level distributed frequency-controlled array sidelobe suppression, while achieving the purpose of effectively suppressing its distance- and angle-dimensional sidelobes. Summary of the invention

[0008] In view of this, the purpose of the present invention is to provide a subarray-level distributed frequency control array sidelobe suppression method based on an improved genetic algorithm, introduce a niche genetic method based on a pre-selection mechanism, introduce a competition mechanism between the parent and the offspring, and perform a replacement operation after comparing the fitness values ​​of the parent and the offspring. This method tends to replace individuals similar to itself, which can effectively maintain population diversity and greatly improve the optimization effect. Taboo search is introduced to accelerate the convergence speed, improve the optimization results of the subarray position and subarray frequency deviation of the subarray-level distributed frequency control array, and achieve better sidelobe suppression results from the distance dimension and the angle dimension. The present invention provides a good reference value for the engineering implementation of the subarray-level distributed frequency control array.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A subarray-level distributed frequency-controlled array sidelobe suppression method based on an improved genetic algorithm, which uses the improved genetic algorithm to simultaneously suppress the angle-dimensional and distance-dimensional grating lobes or high sidelobes of the distributed frequency-controlled array; the improved genetic algorithm is a niche genetic algorithm based on a pre-selection mechanism, which introduces a competition mechanism between the parent and the offspring, and only when the fitness value of the offspring individual is higher than that of the parent individual, the offspring individual can replace its parent individual accordingly, otherwise the parent individual is still retained in the next generation group, and the local and global optimal solutions are optimized in the feasible solution space; and taboo search is introduced to accelerate the convergence speed, avoid the premature phenomenon of the genetic algorithm and stagnation, thereby jumping out of the local optimal solution;

[0011] The method specifically comprises the following steps:

[0012] S1: Set parameters and initialize the population;

[0013] S2: Calculate the fitness value;

[0014] S3: perform selection operation;

[0015] S4: perform crossover operation;

[0016] S5: perform mutation operation;

[0017] S6: Generate offspring using a niche genetic algorithm based on a pre-selection mechanism to generate an approximate optimal solution;

[0018] S7: If the stagnation condition is met, the taboo search algorithm is executed with the approximate best optimization solution as the initial solution, otherwise the next iteration of the genetic algorithm is entered; specifically, it includes: using the microhabitat genetic algorithm after the pre-selection mechanism to generate the approximate best optimization solution as the initial solution of the taboo search, if the termination criterion of the taboo search algorithm is met, the optimal solution of the taboo search algorithm is output; otherwise, neighborhood optimization and taboo check are performed, and then the historical optimal value is updated, and then the taboo table is updated until the termination criterion of the taboo search algorithm is met, and the optimal solution of the taboo search algorithm is output.

[0019] S8: Compare the optimal solution obtained by the taboo search algorithm with the approximate optimal optimization solution, and retain the maximum value to enter the next iteration of the genetic algorithm; specifically, if the optimal solution of the taboo search algorithm is greater than the approximate optimal optimization solution, replace the subarray node position parameters and subarray frequency deviation parameters of the corresponding individuals of the two; if it is less than or equal to the approximate optimal optimization solution, retain the approximate optimal optimization solution and perform the next genetic algorithm iteration until the termination criterion of the genetic algorithm is met.

[0020] S9: End the optimization process of the algorithm.

[0021] If the stagnation condition is met, the taboo search algorithm is entered to output the optimal solution of the taboo search algorithm, which is compared with the approximate optimal optimization solution, and the larger value result is retained. Then, it is determined whether the termination criterion of the genetic algorithm is met. If so, the optimization process of the algorithm is terminated, and the optimization results of the subarray position and subarray frequency deviation of the subarray-level distributed array are output; otherwise, the next iteration of the genetic algorithm is continued, and steps S2 to S9 are repeated until the termination criterion of the genetic algorithm is met, the optimization process of the algorithm is terminated, and the optimization results of the subarray position and subarray frequency deviation of the subarray-level distributed array are output.

[0022] Furthermore, in step S1, the parameters set mainly include: the maximum number of iterations of the genetic algorithm GA MaxIter (Here the subscript MaxIter is used to represent the maximum number of iterations), the population size NP, the crossover probability P c (where P c The subscript is the first letter of the English word cross), the mutation probability P m (where P m The subscript is the first letter of the English mutation), the maximum number of iterations of the taboo search algorithm TS MaxIter (Here the subscript MaxIter is used to represent the maximum number of iterations); the neighborhood mapping function is a generalized 2-Opt, that is, a generalized 2-Optimization form, with a taboo table TabuList, a taboo length TabuSize, a continuous stagnation algebra G, and a difference ε that satisfies the stagnation condition, etc.

[0023] The optimization parameters are the idle aperture allocation spacing and subarray frequency offset value in the subarray node position. These parameters are initialized in the population by real-value mapping coding. These optimization parameters will directly affect the performance of the subarray-level distributed frequency-controlled array integrated beam pattern.

[0024] Optimization parameter process: Subarray node position C based on the minimum spacing constraint of the subarray mn and subarray frequency deviation Δf mn Real-value mapping encoding is performed, where m and n represent the row number and column number of the distributed frequency control array sub-array node, respectively, m = 0, ..., M-1, n = 0, ..., N-1, M and N represent the number of row nodes and column nodes of the distributed frequency control array sub-array, respectively, and the sub-array node position C mn Arrange them in descending order to form a chromosome individual containing the position information of the subarray nodes, and finally initialize the population randomly.

[0025] Further, in step S2, the fitness function is a function that represents the survival ability of an individual in the survival competition, and is an evaluation function for the quality of chromosomes. Calculating the fitness value specifically includes: according to the definition of the maximum sidelobe level, taking the sum of the maximum peak sidelobe level of the pitch dimension transmission beam pattern and the maximum peak sidelobe level of the distance dimension transmission beam pattern as the fitness function, and calculating the fitness value of each chromosome individual in the population according to the fitness function.

[0026] Further, in step S2, considering the static beam, the far-field beam pattern expression of the sub-array-level distributed frequency-controlled array after conventional beam weighting is:

[0027]

[0028] in, It is represented as a conventional beam weight vector, and the superscript H indicates the conjugate transpose of the matrix; It is represented as the steering vector corresponding to the sub-array-level distributed frequency control array; p and q represent the row number and column number of the array element in the distributed frequency control array sub-array, respectively, p=0,…,P-1, q=0,…,Q-1, P and Q represent the number of row nodes and column nodes of the array element in the distributed frequency control array sub-array, respectively, and j is an imaginary unit; dx m Indicates the subarray node spacing corresponding to the mth row number in the x-axis direction, dy n Indicates the subarray node spacing corresponding to the nth column number in the y-axis direction, which contains the position parameters to be optimized after layout constraint analysis: d x represents the spacing between array elements in the subarray in the x-axis direction, d y represents the spacing between the array elements in the subarray in the y-axis direction; θ0 represents the elevation angle between the distributed frequency control array and the observed target point, represents the azimuth angle between the distributed frequency control array and the observed target point in the zero plane, R0 represents the distance between the distributed frequency control array and the observed target point, θ represents the pitch angle between the distributed frequency control array and any point in space, represents the azimuth angle between the distributed frequency control array and the zero plane of space, R represents the distance between the distributed frequency control array and any point in space; f0 is the center carrier frequency, c is the speed of light, Δf mn Indicates the frequency deviation value of the subarray corresponding to row number m and column number n.

[0029] Further, in step S3, performing the selection operation specifically includes: according to the position of the subarray node (dx m ,dy n ) and subarray frequency deviation Δf mn The calculated fitness value, where dx m Indicates the subarray node spacing corresponding to the mth row number in the x-axis direction, dy nIndicates the subarray node spacing corresponding to the nth column number in the y-axis direction, Δf mn Indicates the frequency deviation value of the subarray corresponding to row number m and column number n; then the genes of the individuals in the population are arranged in descending order to ensure the correct optimization of the node positions of the subarrays, and the roulette strategy is combined to select the parent individuals. The larger the individual fitness value, the greater the chance of being selected;

[0030] At the same time, an elite preservation strategy is introduced, in which the individual with the best fitness value in the current population does not participate in genetic crossover and mutation operations, but is directly replaced with the individual with the worst fitness value in the offspring population produced by crossover and mutation operations. The advantage is that the best individual of a certain generation will not be destroyed by genetic crossover and mutation operations during the search process, which plays a role in protecting and promoting excellent individuals, and can ensure that the genetic algorithm converges to the optimal solution with probability.

[0031] In step S3, the selection operation combines roulette and elite retention strategies to directly retain excellent individuals, and then participate in the next genetic algorithm iteration, which plays a role in protecting excellent individuals.

[0032] Further, in step S4, performing the crossover operation specifically includes: performing a crossover operation on the individual after the selection operation is completed. Under the condition of satisfying the crossover probability, a multi-row and multi-column crossover method is used to perform a random multi-row and multi-column crossover on the positions of two consecutive parent individual sub-arrays, and similarly, a crossover operation is performed on the frequency deviations of two consecutive parent individual sub-arrays. This operation can greatly improve the global search capability of the genetic algorithm and help to pass on the chromosome fragments of excellent individuals to the offspring.

[0033] In step S4, the crossover operation is to perform multi-row and multi-column crossover on the chromosome segments in any two consecutive adjacent parents, so that the offspring individuals after the crossover can effectively inherit the excellent chromosome segments of the parents.

[0034] Further, in step S5, performing the mutation operation specifically includes: performing a mutation operation on the individual after the crossover operation is completed. Under the condition of satisfying the mutation probability, the individual is subjected to a mutation operation within the single-row multi-column real-valued aperture constraint range, and similarly, the individual is subjected to a mutation operation within the corresponding frequency deviation constraint range. This operation can enhance the local random search capability, thereby effectively accelerating the convergence to the optimal solution.

[0035] In step S5, the mutation operation is to perform a single-row and multiple-column mutation on the chromosome segments in any two consecutive adjacent parents so as to improve the local search capability.

[0036] Further, in step S6, the niche genetic algorithm based on the pre-selection mechanism specifically includes: in the pre-selection mechanism, only when the fitness value of the offspring exceeds that of its parent, the offspring can replace its parent and enter the next generation group. This method tends to replace individuals with similar coding structures to itself, so that the distribution characteristics of the group can be better maintained.

[0037] In step S6, a microhabitat genetic method based on a pre-selection mechanism is introduced to compare the parent and offspring generations using fitness values ​​as a measurement indicator, and then retain the one with a larger value, that is, "survival of the fittest", which is more effective in finding the global optimal solution.

[0038] Further, in step S7, the stagnation condition specifically includes: let the continuous algebra G(<GA MaxIter ), GA MaxIter Represents the maximum number of iterations of the genetic algorithm. Here, the subscript MaxIter is used to represent the maximum number of iterations. If both satisfy Where ε is a small positive number, the subscript max is used to represent the optimal solution of the fitness value, and fit represents the fitness function. is the optimal fitness value of the g-th generation in the genetic algorithm, is the optimal fitness value of the g-1th generation in the genetic algorithm; if the difference between the optimal fitness values ​​in consecutive G generations is less than ε, it indicates that the genetic algorithm is basically in a "stagnant" state. At this time, the approximate optimal optimization solution generated by the small habitat genetic algorithm after the pre-selection mechanism is used as the initial solution of the taboo search algorithm, making full use of the taboo search algorithm's strong "hill climbing" ability and high solution efficiency, so that the algorithm can quickly converge to the global optimum.

[0039] Furthermore, in step S8, the optimal solution of the taboo search is compared with the generated approximate optimal solution, and the larger value is retained. This has two advantages: one is to ensure that the taboo search does not destroy the excellent individuals generated by the pre-selection mechanism; the other is to ensure that the excellent individuals generated by the pre-selection mechanism avoid falling into the local optimal solution, thereby obtaining the global optimal solution.

[0040] Further, in step S9, the termination condition of the genetic algorithm is the maximum number of iterations of the genetic algorithm GA set MaxIter .

[0041] The beneficial effects of the present invention are as follows: the present invention aims at the problem that the subarray-level distributed frequency control array has grating lobes or high side lobes in both the angle dimension and the distance dimension, and simultaneously suppresses the grating lobes or high side lobes in the angle dimension and the distance dimension of the array by improving the genetic algorithm. In particular, the introduction of a microhabitat genetic method based on a pre-selection mechanism can effectively enhance the search for the global optimal solution while maintaining population diversity, and introduces a taboo search to accelerate the convergence speed, avoid the genetic algorithm from being premature and stagnant, and thus jump out of the local optimal solution. The method of the present invention can effectively improve the optimization results of the subarray position and subarray frequency deviation of the subarray-level distributed frequency control array, achieve better sidelobe suppression results from the distance dimension and the angle dimension, and provide a good reference value for the engineering realization of the subarray-level distributed frequency control array.

[0042] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0044] Figure 1 It is a schematic diagram of the overall implementation process of the improved genetic algorithm in Example 1;

[0045] Figure 2 This is a schematic diagram of a sub-array-level distributed frequency control array model in Embodiment 1;

[0046] Figure 3 A three-dimensional schematic diagram of linear frequency deviation values ​​corresponding to sub-array nodes in the first aspect of the second embodiment;

[0047] Figure 4 A two-dimensional schematic diagram of the positions of nodes of uniform sub-array spacing in the first aspect of the second embodiment;

[0048] Figure 5 It is a schematic diagram of the convergence curve of the traditional genetic algorithm subarray position optimization of the logarithmic frequency deviation in the second aspect of the second embodiment;

[0049] Figure 6 A three-dimensional schematic diagram of the logarithmic frequency deviation values ​​corresponding to the sub-array nodes in the second aspect of the second embodiment;

[0050] Figure 7 It is a two-dimensional schematic diagram of the traditional genetic algorithm sub-array position optimization in the second aspect of the second embodiment;

[0051] Figure 8It is a schematic diagram of a convergence curve for simultaneously optimizing the subarray position and frequency offset of the improved genetic algorithm in the third aspect of the second embodiment;

[0052] Fig. 9 A three-dimensional schematic diagram of the frequency deviation optimization value of the improved genetic algorithm in the third aspect of the second embodiment;

[0053] Fig.10 This is a two-dimensional schematic diagram of improving the sub-array position optimization of the genetic algorithm in the third aspect of the second embodiment. DETAILED DESCRIPTION

[0054] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0055] Embodiment 1:

[0056] like Figure 1 As shown, this embodiment provides a technical solution: combining the pre-selection mechanism and the improved genetic algorithm of taboo search to implement the subarray-level distributed frequency-controlled array sidelobe suppression method, which specifically includes the following steps:

[0057] S1: Set parameters and initialize the population.

[0058] Consider a distributed frequency-controlled array consisting of M×N subarrays, where each subarray is a P×Q uniform planar phased array, and the array element spacing in the x-axis and y-axis directions in each subarray is d x and d y Indicated by half wavelength, and the spacing between adjacent sub-arrays in the x-axis and y-axis directions is represented by D x and D y Indicates. Set the constraint array aperture to L x ×L y , all array elements are isotropic and located in the xoy plane in the three-dimensional coordinate system, such as Figure 2 shown.

[0059] Select the upper left corner of each subarray as the reference point and describe it as a subarray node. Then the coordinate position of each subarray node relative to the origin in the xoy plane is (dx m ,dy n ), where dx m Indicates the subarray node spacing corresponding to the mth row number in the x-axis direction, dy nIndicates the subarray node spacing corresponding to the nth column number in the y-axis direction, where m and n represent the row number and column number of the distributed frequency-controlled array subarray node, respectively, m = 0, ..., M-1, n = = 0, ..., N-1. In order to maximize the use of the array aperture of the rectangular plane array from the perspective of physical layout, the position coordinates of the subarray boundary nodes located at the four corners of the rectangular plane area are limited:

[0060]

[0061] Among them, dx0 represents the sub-array node spacing corresponding to the 0th row number in the x-axis direction, dy0 represents the sub-array node spacing corresponding to the 0th column number in the y-axis direction, and dx M-1 Indicates the subarray node spacing corresponding to the M-1th row number in the x-axis direction, dy N-1 Indicates the subarray node spacing corresponding to the N-1th column number on the y-axis, d x represents the spacing between array elements in the subarray in the x-axis direction, d y Indicates the spacing between array elements in the subarray in the y-axis direction, L x represents the array aperture in the x-axis direction, L y represents the array aperture in the y-axis direction, P represents the number of array elements in the sub-array in the x-axis direction, and Q represents the number of array elements in the sub-array in the y-axis direction.

[0062] At the same time, the subarray node positions in each direction should satisfy the subarray minimum spacing constraint, which can be expressed as:

[0063]

[0064] Among them, dx j Indicates the subarray node spacing corresponding to the j-th row number in the x-axis direction, dx i Indicates the subarray node spacing corresponding to the i-th row number in the x-axis direction, dy l Indicates the subarray node spacing corresponding to the lth column number in the y-axis direction, dy k represents the sub-array node spacing corresponding to the k-th column number in the y-axis direction, where j and i correspond to the row numbers of the distributed frequency control array sub-array nodes, and l and k correspond to the column numbers of the distributed frequency control array sub-array nodes.

[0065] In fact, the actual spacing of subarray nodes in all directions can be decomposed into two parts: the idle aperture allocation spacing and the minimum spacing of the subarray, which can be expressed as:

[0066]

[0067] Among them, dx m Indicates the subarray node spacing corresponding to the mth row number in the x-axis direction, dym Indicates the subarray node spacing corresponding to the nth column number in the y-axis direction, Sx m , D m They represent the idle aperture allocation spacing of the subarray corresponding to the mth row number in the x-axis direction and the minimum spacing constraint value of the subarray, Sx n , D n They respectively represent the idle aperture allocation spacing and the subarray minimum spacing constraint value of the nth column numbered subarray in the y-axis direction, where m and n represent the row number and column number of the distributed frequency control array subarray node, respectively. Since the position coordinates of the subarray boundary nodes at the four corners of the rectangular plane area are limited, the positions of the subarray boundary nodes at the four corners of the rectangular plane area are not considered here.

[0068] According to the above array aperture constraints, the idle aperture allocation spacing range in the x-axis direction and the y-axis direction can be obtained, which is expressed as:

[0069]

[0070] Among them, Sx m , D m They represent the idle aperture allocation spacing of the subarray corresponding to the mth row number in the x-axis direction and the minimum spacing constraint value of the subarray, Sx n , D n They represent the free aperture allocation spacing of the n-th column numbered subarray in the y-axis direction and the minimum spacing constraint value of the subarray, L x represents the array aperture in the x-axis direction, L y represents the array aperture in the y-axis direction, P represents the number of array elements in the subarray in the x-axis direction, Q represents the number of array elements in the subarray in the y-axis direction, where m and n represent the row number and column number of the subarray node of the distributed frequency control array, respectively. Since the position coordinates of the subarray boundary nodes at the four corners of the rectangular plane area are limited, the positions of the subarray boundary nodes at the four corners of the rectangular plane area are not considered here.

[0071] After the above analysis, the sub-array node spacing dx in the x-axis direction can be m Optimization is indirectly converted into spatial aperture allocation spacing Sx m Optimization, this encoding method can reduce the search space range from size L to x Mapped to size L x -D m -(P-1)×d x space; similarly, the sub-array node spacing dx in the y-axis direction n Optimization is indirectly converted into spatial aperture allocation spacing Sx n Optimize, expand the search space from size L y Mapped to size Ly -D n -(Q-1)×d y space, thereby effectively reducing the redundant search space.

[0072] In order to facilitate optimization, the node positions of the subarray are converted into a two-dimensional complex matrix C with a dimension of M×N for representation, that is, C mn =Sx m +jSy n Similarly, the subarray node position corresponds to the frequency deviation Δf mn Use a two-dimensional matrix ΔF with dimensions M×N mn , where m and n represent the row number and column number of the distributed frequency control array subarray node respectively, m=0,…,M-1, n=0,…,N-1, which includes the subarray node positions and subarray frequency deviations at the four corners of the rectangular plane after layout constraints, so as to be used for subsequent fitness numerical calculations.

[0073] The parameter setting mainly includes the maximum number of iterations of the genetic algorithm GA MaxIter , where the subscript MaxIter is used to represent the maximum number of iterations, the population size NP, and the crossover probability P c , where P c The subscript is the first letter of the English word cross, and the mutation probability P m , where P m The subscript is the first letter of the English mutation, the maximum number of taboo iterations TS MaxIter Here, the subscript MaxIter is used to represent the maximum number of iterations, the neighborhood mapping function is the generalized 2-Opt, that is, the generalized 2-Optimization form, the taboo list TabuList, the taboo length TabuSize, the continuous stagnation algebra G, the difference ε that satisfies the stagnation condition, etc., and the subarray node optimization position C is based on the subarray minimum spacing constraint mn and subarray optimized frequency offset Δf mn Real-value mapping encoding is performed, where m and n represent the row number and column number of the distributed frequency control array sub-array node, m = 0, ..., M-1, n = 0, ..., N-1, and the sub-array node position C mn Arrange them in descending order to form a chromosome individual containing the position information of the subarray nodes, and finally initialize the population randomly.

[0074] S2: Calculate the fitness value.

[0075] The fitness function is a function that represents the ability of an individual to survive in the survival competition, and is an evaluation function for the quality of chromosomes. According to the definition of the maximum sidelobe level, the sum of the maximum sidelobe level of the pitch dimension transmission beam diagram and the maximum sidelobe level of the distance dimension transmission beam diagram is used as the value of the fitness function fit, and the specific expression is as follows:

[0076]

[0077] Among them, Sx m Indicates the idle aperture allocation spacing of the subarray corresponding to the mth row number in the x-axis direction, Sx n represents the free aperture allocation spacing of the nth column numbered subarray in the y-axis direction, Δf mn Indicates the frequency deviation value of the subarray corresponding to row number m and column number n, where m and n represent the row number and column number of the distributed frequency control array subarray node respectively, and m=0,…,M-1, n=0,…,N-1. Ω θ Indicates the sidelobe interval corresponding to the pitch angle, Ω R Indicates the sidelobe interval corresponding to the distance dimension, It represents the expression of the sub-array level distributed frequency-controlled array beam pattern under far-field conditions, TB ML represents the main lobe peak, that is θ0 represents the pitch angle between the distributed frequency control array and the observed target point, represents the azimuth angle between the distributed frequency control array and the observed target point in the zero plane, R0 represents the distance between the distributed frequency control array and the observed target point, θ represents the pitch angle between the distributed frequency control array and any point in space, represents the azimuth between the distributed frequency control array and the spatial zero plane, R represents the distance between the distributed frequency control array and any point in space, and the subscript dB in the above formula indicates that the peak sidelobe level is measured in the form of power.

[0078] Based on the subarray minimum spacing constraint, the following fitness function optimization model is established for the subarray node position and subarray frequency deviation:

[0079]

[0080] Among them, Sx m Indicates the idle aperture allocation spacing of the subarray corresponding to the mth row number in the x-axis direction, Sx n Indicates the free aperture allocation spacing of the nth column numbered subarray in the y-axis direction, Sx i Indicates the idle aperture allocation spacing of the subarray corresponding to the i-th row number in the x-axis direction, Sx j Indicates the idle aperture allocation spacing of the subarray corresponding to the jth row number in the x-axis direction, where i and j correspond to the row numbers of the subarray nodes of the distributed frequency control array, Sx k Indicates the idle aperture allocation spacing of the subarray corresponding to the kth column number in the y-axis direction, Sx l Indicates the idle aperture allocation spacing of the subarray corresponding to the lth column number in the y-axis direction, where k and l correspond to the column numbers of the subarray nodes of the distributed frequency-controlled array, Δf mnIndicates the frequency deviation value of the subarray corresponding to row number m and column number n, Δf min Indicates the lower limit of frequency deviation, Δf max Indicates the upper frequency deviation limit.

[0081] The above formula shows that by optimizing the position of the subarray nodes corresponding to the idle aperture allocation spacing Sx m , Sx n and subarray frequency deviation Δf mn , where m and n represent the row number and column number of the distributed frequency control array sub-array node respectively, m=0,…,M-1, n=0,…,N-1, which can minimize the maximum sidelobe level in the area except the main lobe, that is, the maximum angle-dimensional sidelobe level and the maximum distance-dimensional sidelobe level are minimized at the same time.

[0082] The fitness value of each chromosome individual in the population is calculated according to the above fitness function.

[0083] Considering the static beam, the far-field beam pattern expression of the sub-array-level distributed frequency-controlled array after conventional beam weighting is:

[0084]

[0085] in, Expressed as a conventional beam weight vector, It represents the steering vector corresponding to the subarray-level distributed frequency control array. The superscript H represents the conjugate transpose of the matrix. m and n represent the row and column numbers of the nodes of the distributed frequency control array subarray, respectively. m=0,…,M-1, n=0,…,N-1, p and q represent the row and column numbers of the elements in the distributed frequency control array subarray, respectively. p=0,…,P-1, q=0,…,Q-1, and j is an imaginary unit. dx m Indicates the distance between nodes of the mth subarray in the x-axis direction, dy n Indicates the spacing of the nth subarray node in the y-axis direction, which includes the position parameters to be optimized after layout constraint analysis: the free aperture allocation spacing Sx in the x-axis direction m and the free aperture allocation spacing Sx in the y-axis direction n ;d x represents the spacing between array elements in the subarray in the x-axis direction, d y represents the distance between the array elements in the subarray in the y-axis direction; θ0 represents the pitch angle between the distributed frequency control array and the observed target point, represents the azimuth angle between the distributed frequency control array and the observed target point in the zero plane, R0 represents the distance between the distributed frequency control array and the observed target point, θ represents the pitch angle between the distributed frequency control array and any point in space, represents the azimuth angle between the distributed frequency control array and the zero plane of space, R represents the distance between the distributed frequency control array and any point in space; f0 is the center carrier frequency, c is the speed of light, Δf mn Indicates the frequency deviation value of the subarray corresponding to row number m and column number n.

[0086] S3: Execute the selection operation.

[0087] According to the subarray node position parameters (dx m ,dy n ) and subarray frequency offset parameter Δf mn Calculate the fitness function value, where dx m Indicates the subarray node spacing corresponding to the mth row number in the x-axis direction, dy n Indicates the subarray node spacing corresponding to the nth column number in the y-axis direction, Δf mn It represents the frequency deviation value of the subarray corresponding to row number m and column number n, where m and n represent the row number and column number of the distributed frequency control array subarray node respectively, m=0,…,M-1, n=0,…,N-1, and then the individual genes of the population are arranged in descending order, and the roulette strategy is combined to select the parent individuals. The larger the individual fitness value, the greater the chance of being selected.

[0088] At the same time, an elite preservation strategy is introduced, in which the individual with the best fitness value in the current population does not participate in genetic crossover and mutation operations, but is directly replaced with the individual with the worst fitness value in the offspring population produced by crossover and mutation operations. The advantage is that the best individual of a certain generation will not be destroyed by genetic crossover and mutation operations during the search process, which plays a role in protecting and promoting excellent individuals, and can ensure that the genetic algorithm converges to the optimal solution with probability.

[0089] S4: Perform crossover operation

[0090] After the selection operation is completed, the individuals are crossovered. Under the condition of satisfying the crossover probability, the multi-row and multi-column crossover method is used to crossover two consecutive parent individuals. The specific operation is as follows:

[0091] First, randomly generate M integers of 0 and 1, whose sum is guaranteed to be at least 2, and correspond the index position represented by 1 to the position index of the row to be crossed between the two consecutive parent individuals. Then, exchange all columns of the two consecutive parent individuals in the crossover row, and use the same method to perform the crossover operation on the frequency deviation parameter. This operation can greatly improve the global search ability of the genetic algorithm and help to pass on the chromosome fragments of excellent individuals to the offspring.

[0092] S5: Perform mutation operation.

[0093] After the crossover operation is completed, the individuals are mutated. Under the condition of satisfying the mutation probability, the individuals are mutated within the single-row multi-column real-valued aperture constraint range. Similarly, the individuals are mutated within the corresponding frequency deviation constraint range. This operation can enhance the local random search capability, thereby effectively accelerating the convergence to the optimal solution.

[0094] S6: Generate offspring using a niche genetic approach based on a pre-selection mechanism to generate a near-optimal optimization solution.

[0095] In the pre-selection mechanism, only when the offspring's fitness value exceeds that of its parent, can the offspring replace its parent and enter the next generation. This method tends to replace individuals with similar coding structures to itself, thus better maintaining the distribution characteristics of the population.

[0096] S7: If the stagnation criterion is met, the taboo search is performed with the approximate best optimization solution as the initial solution, otherwise the next iteration of the genetic algorithm is entered.

[0097] The approximate best optimization solution generated by the niche genetic algorithm after the pre-selection mechanism is used as the initial solution of the taboo search. If the termination criterion of the taboo search algorithm is met, the optimal solution of the taboo search algorithm is output; otherwise, neighborhood optimization and taboo check are performed, and then the historical optimal value is updated, and then the taboo table is updated until the termination criterion of the taboo search algorithm is met, and the optimal solution of the taboo search algorithm is output.

[0098] Set the following stagnation condition: Let the continuous algebra G(<GA MaxIter ), GA MaxIter Represents the maximum number of iterations of the genetic algorithm. Here, the subscript MaxIter is used to represent the maximum number of iterations. If both satisfy Where ε is a small positive number, the subscript max is used to represent the optimal solution of the fitness value, and fit represents the fitness function. is the optimal fitness value of the g-th generation in the genetic algorithm, is the optimal fitness value of the g-1th generation in the genetic algorithm. If the difference between the optimal fitness values ​​in consecutive G generations is less than ε, it indicates that the genetic algorithm is basically in a "stagnant" state. At this time, the approximate best optimization solution generated by the small habitat genetic method after the pre-selection mechanism is used as the initial solution of the taboo search algorithm, making full use of the taboo search algorithm's strong "hill climbing" ability and high solution efficiency, so that the algorithm can quickly converge to the global optimum.

[0099] S8: Compare the optimal solution of the taboo search with the approximate best optimization solution, and retain the maximum value to enter the next iteration of the genetic algorithm.

[0100] If the optimal solution of the taboo search algorithm is greater than the approximate optimal optimization solution, the subarray node position parameters and subarray frequency deviation parameters of the corresponding individuals are replaced; if it is less than or equal to the approximate optimal optimization solution, the approximate optimal optimization solution is retained and the next genetic algorithm iteration is performed until the genetic algorithm termination criterion is met.

[0101] S9: End the optimization process of the algorithm

[0102] If the stagnation condition is met, the taboo search algorithm is entered to output the optimal solution of the taboo search algorithm, which is compared with the approximate optimal optimization solution, and the larger value result is retained. Then, it is determined whether the termination criterion of the genetic algorithm is met. If so, the optimization process of the algorithm is terminated, and the optimization results of the subarray position and subarray frequency deviation of the subarray-level distributed array are output; otherwise, the next iteration of the genetic algorithm is continued, and steps S2 to S9 are repeated until the termination criterion of the genetic algorithm is met, the optimization process of the algorithm is terminated, and the optimization results of the subarray position and subarray frequency deviation of the subarray-level distributed array are output.

[0103] Embodiment 2:

[0104] In order to evaluate the performance of the present invention, this embodiment compares simulation experiments under different conditions to verify the optimization effect of the present invention.

[0105] Due to the grating lobes or high side lobes in the angle and distance dimensions of the sub-array-level distributed frequency-controlled array, the proposed method can be verified to suppress the grating lobes or high side lobes in the distance and angle dimensions at the same time. Comparative simulations are performed from the following three aspects:

[0106] First, given the sub-array-level distributed frequency-controlled array aperture range L x ×L y The aperture range is 25λ×25λ, and the wavelength is used as the unit to express the aperture range, which makes the subarray position optimization result more intuitive. The number of subarrays is set to M×N=6×6, and the subarray spacing in each direction is D x =D y =4λ, the spacing between array elements in each direction within the subarray is d x =d y =λ / 2, the azimuth is fixed at the zero plane, the target position is (θ0, R0) = (50km, 30°), the center carrier frequency is f0 = 3GHz, and the speed of light c = 3×10 8 m / s.

[0107] First, considering the beam pattern performance of the subarray-level distributed frequency-controlled array under the conditions of linear frequency deviation and uniform subarray spacing, the corresponding frequency deviation expression of each subarray is:

[0108] f mn =f0+Δf mn ,m=0,...,M-1,n=0,...,N-1

[0109] Where Δf mn =mΔf x +nΔf y It is expressed as the linear frequency deviation value of the subarray corresponding to row number m and column number n, f mn It is represented by the transmission frequency of the subarray corresponding to the row number m and the column number n, where m and n represent the row number and column number of the subarray node of the distributed frequency control array respectively.

[0110] From the properties of the frequency-controlled array, we know that: f0>>Δf x , f0>>Δf y For the convenience of analysis, let the frequency shift Δf in the x-axis direction x =3KHz, frequency offset Δf in the y-axis direction y =(M+1)Δf x , given as Figure 3-4 The simulation results are shown.

[0111] The second aspect: Considering the nonlinear frequency deviation and the traditional genetic algorithm to optimize the subarray position based on the layout constraints, the nonlinear frequency deviation here specifically refers to the logarithmic frequency deviation. The combination of the two acts on the subarray-level distributed frequency-controlled array beam pattern at the same time, then the corresponding frequency deviation expression of each subarray is:

[0112] f mn =f0+Δf mn ,m=0,...,M-1,n=0,...,N-1

[0113] Where Δf mn =log(m+1)Δf x +log(n+1)Δf y It is represented by the logarithmic frequency deviation value of the subarray corresponding to row number m and column number n, f mn It is represented by the transmission frequency of the subarray corresponding to the row number m and the column number n, where m and n represent the row number and column number of the subarray node of the distributed frequency control array respectively.

[0114] Traditional genetic algorithm setting parameters: population size NP = 50, maximum number of iterations GA max =100, where the subscript max is used to represent the maximum number of iterations, and the crossover probability P c =0.85, mutation probability P c =0.05. The selection operation adopts the roulette strategy, the crossover operation adopts the single-point crossover method, and the mutation operation adopts the single-point mutation method. Figure 5 to Figure 7 The simulation results are shown.

[0115] The third aspect: Consider using an improved genetic algorithm to simultaneously optimize the subarray position and subarray frequency offset based on layout constraints. The limited frequency offset range is: Δf min =-10KHz,Δf max =10KHz, where the subscript min is used to represent the lower limit of the frequency deviation, and the subscript max is used to represent the upper limit of the frequency deviation. Then the corresponding optimized frequency deviation expression of each subarray is:

[0116] f mn =f0+Δf mn ,m=0,...,M-1,n=0,...,N-1

[0117] Where Δf mn ∈random×[Δf min ,Δf max ] represents the random optimization value of the nonlinear frequency deviation of the subarray with row number m and column number n, f mn It is represented by the transmission frequency of the subarray corresponding to the row number m and the column number n, where m and n represent the row number and column number of the subarray node of the distributed frequency control array respectively.

[0118] Improved genetic algorithm setting parameters: population size NP = 50, maximum iteration number GA MaxIter =100, where MaxIter is used to represent the maximum number of iterations, and the crossover probability P c =0.85, mutation probability P m =0.05. Roulette and elite preservation strategies are used in selection operation, multi-row and multi-column crossover is used in crossover operation, and single-row and multi-column crossover is used in mutation operation. A niche genetic method based on pre-selection mechanism is introduced. By introducing a competition mechanism between the parent and offspring, the offspring with a fitness value greater than that of the parent is replaced, which effectively maintains the diversity of the population and improves the efficiency of finding the global optimal solution. Taboo search is introduced to break the stagnation of the genetic algorithm, thereby avoiding premature maturity.

[0119] Set the stagnation condition parameters: define the continuous generation number G = 6, the stagnation condition difference ε = 0.01, that is, the difference between the optimal fitness values ​​in the continuous G generations is less than ε, indicating that the genetic algorithm is basically stuck in a "stagnant" state. If the stagnation condition is met, the approximate optimal optimization solution generated by the niche genetic method based on the pre-selection mechanism is used as the initial solution to perform taboo search.

[0120] Set the taboo search parameters to: TS MaxIter=20, where the subscript MaxIter is used to characterize the maximum number of iterations of the taboo search algorithm. The neighborhood function is defined as a generalized 2-Optimization form, that is, two subarray node positions and corresponding subarray frequency deviations in the individual are randomly selected and exchanged in the form of single row and multiple columns and multiple rows and single column. The obtained new individuals constitute the neighborhood solution of this approximate optimal optimization solution. The neighborhood solution is substituted into the fitness function for calculation to generate a candidate solution. The taboo table TabuList is set to the subarray node row and column number, the taboo length TabuSize is 4, and the candidate solution is 40. After parallel comparison, 20 better candidate solutions are retained, and then the taboo search optimal solution is obtained by the taboo search algorithm and compared with the generated approximate optimal optimal solution. The larger value is retained to enter the next genetic iteration until the genetic termination condition is met and the optimization process ends.

[0121] Through the above operations, we can get Figure 8 to Figure 10 The simulation results are shown.

[0122] In summary, through the simulation of sidelobe suppression of the subarray-level distributed frequency control array in the above two groups of embodiments, it can be concluded that by introducing the niche genetic method and the taboo search method based on the pre-selection mechanism, the optimization results of the subarray position and the subarray frequency deviation of the subarray-level distributed array are obtained, and these results are simultaneously applied to the subarray-level distributed frequency control array, so that better sidelobe suppression results can be achieved in terms of distance dimension and angle dimension, thereby improving the performance of the distributed antenna, and having a certain reference value for the engineering design of the subarray-level distributed frequency control array.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A subarray-level distributed frequency-controlled array sidelobe suppression method based on an improved genetic algorithm, characterized in that: The method uses an improved genetic algorithm to simultaneously suppress the grating lobes or high side lobes of the angle dimension and the distance dimension of the distributed frequency-controlled array; the improved genetic algorithm is a niche genetic algorithm based on a pre-selection mechanism, and a competition mechanism is introduced between the parent generation and the offspring generation. Only when the fitness value of the offspring individual is higher than that of the parent individual, the offspring individual can replace its parent individual accordingly, otherwise the parent individual is still retained in the next generation group, and the optimization of local and global optimal solutions is realized in the feasible solution space; And introduce tabu search to speed up convergence; The method specifically comprises the following steps: S1: Set parameters and initialize the population; The parameters to be set include: Maximum number of iterations of the genetic algorithm GA MaxIter , population size NP, crossover probability P c , mutation probability P m , the maximum number of iterations of the taboo search algorithm TS MaxIter ; The neighborhood mapping function is a generalized 2-Opt, i.e., a generalized 2-Optimization form, with a taboo list TabuList, a taboo length TabuSize, a continuous stagnation algebra G, and a difference ε that satisfies the stagnation condition; Optimization parameters: Subarray node position C based on the minimum spacing constraint of the subarray mn and subarray frequency deviation Δf mn Real-value mapping encoding is performed, where m and n represent the row number and column number of the distributed frequency control array sub-array node, respectively, m = 0, ..., M-1, n = 0, ..., N-1, M and N represent the number of row nodes and column nodes of the distributed frequency control array sub-array, respectively, and the sub-array node position C mn Arrange in descending order to form a chromosome individual containing the position information of the subarray nodes, and finally initialize the population randomly; S2: Calculating the fitness value, specifically including: according to the definition of the maximum sidelobe level, taking the sum of the maximum peak sidelobe level of the transmission beam pattern in the elevation dimension and the maximum peak sidelobe level of the transmission beam pattern in the range dimension as the fitness function, and calculating the fitness value of each chromosome individual in the population according to the fitness function; Considering the static beam, the far-field beam pattern expression of the sub-array-level distributed frequency-controlled array after conventional beam weighting is: in, It is represented as a conventional beam weight vector, and the superscript H indicates the conjugate transpose of the matrix; It is represented as the steering vector corresponding to the sub-array-level distributed frequency control array; p and q represent the row number and column number of the array element in the distributed frequency control array sub-array, respectively, p=0,…,P-1, q=0,…,Q-1, P and Q represent the number of row nodes and column nodes of the array element in the distributed frequency control array sub-array, respectively, and j is an imaginary unit; dx m Indicates the subarray node spacing corresponding to the mth row number in the x-axis direction, dy n Indicates the subarray node spacing corresponding to the nth column number in the y-axis direction, which contains the position parameters to be optimized after layout constraint analysis: d x represents the spacing between array elements in the subarray in the x-axis direction, d y represents the spacing between the array elements in the subarray in the y-axis direction; θ0 represents the elevation angle between the distributed frequency control array and the observed target point, represents the azimuth angle between the distributed frequency control array and the observed target point in the zero plane, R0 represents the distance between the distributed frequency control array and the observed target point, θ represents the pitch angle between the distributed frequency control array and any point in space, represents the azimuth angle between the distributed frequency control array and the zero plane of space, R represents the distance between the distributed frequency control array and any point in space; f0 is the center carrier frequency, c is the speed of light, Δf mn Indicates the frequency deviation value of the subarray corresponding to row number m and column number n; S3: Execute the selection operation, specifically including: according to the position of the subarray node in the population individual (dx m ,dy n ) and subarray frequency deviation Δf mn The calculated fitness value, where dx m Indicates the subarray node spacing corresponding to the mth row number in the x-axis direction, dy n Indicates the subarray node spacing corresponding to the nth column number in the y-axis direction, Δf mn Indicates the frequency deviation value of the subarray corresponding to row number m and column number n; then the genes of the individuals in the population are arranged in descending order to ensure the correct optimization of the node positions of the subarrays, and the roulette strategy is combined to select the parent individuals. The larger the individual fitness value, the greater the chance of being selected; At the same time, an elite preservation strategy is introduced, where the individual with the best fitness value in the current population does not participate in genetic crossover and mutation operations, but is directly replaced with the individual with the worst fitness value in the offspring population generated by crossover and mutation operations; S4: perform crossover operation; S5: perform mutation operation; S6: Generate offspring using a niche genetic algorithm based on a pre-selection mechanism to generate an approximate optimal solution; The niche genetic algorithm based on the pre-selection mechanism specifically includes: in the pre-selection mechanism, only when the fitness value of the offspring exceeds that of its parent, the offspring can correspondingly replace its parent and enter the next generation population; S7: If the stagnation condition is met, the taboo search algorithm is executed with the approximate best optimization solution as the initial solution, otherwise the next iteration of the genetic algorithm is entered; The stagnation conditions specifically include: Let the continuous algebra G < GA MaxIter , G.A. MaxIter Represents the maximum number of iterations of the genetic algorithm; if all satisfy Where ε is a small positive number, the subscript max is used to represent the optimal solution of the fitness value, and fit represents the fitness function. is the optimal fitness value of the g-th generation in the genetic algorithm, is the optimal fitness value of the g-1th generation in the genetic algorithm; if the difference between the optimal fitness values ​​in consecutive G generations is less than ε, it indicates that the genetic algorithm is basically in a "stagnant" state. At this time, the approximate optimal optimization solution generated by the niche genetic algorithm after the pre-selection mechanism is used as the initial solution of the taboo search algorithm; Step S7 specifically includes: using the niche genetic algorithm after the pre-selection mechanism to generate an approximate optimal optimization solution as the initial solution of the taboo search, and if the termination criterion of the taboo search algorithm is met, then outputting the optimal solution of the taboo search algorithm; otherwise, performing neighborhood optimization and taboo checking, and then updating the historical optimal value, and then updating the taboo table, until the termination criterion of the taboo search algorithm is met, and then outputting the optimal solution of the taboo search algorithm; S8: Compare the optimal solution obtained by the taboo search algorithm with the approximate optimal optimization solution, retain the maximum value and enter the next iteration of the genetic algorithm, specifically including: if the optimal solution result of the taboo search algorithm is greater than the approximate optimal optimization solution, replace the subarray node position parameters and subarray frequency deviation parameters of the corresponding individuals of the two; if it is less than or equal to the approximate optimal optimization solution, retain the approximate optimal optimization solution and perform the next genetic algorithm iteration until the termination criterion of the genetic algorithm is met; S9: End the optimization process of the algorithm.

2. The subarray-level distributed frequency-controlled array sidelobe suppression method according to claim 1, characterized in that: In step S4, performing the crossover operation specifically includes: under the condition of satisfying the crossover probability, using the multi-row and multi-column crossover method to perform random multi-row and multi-column crossover on the positions of two consecutive parent individual sub-arrays, and similarly performing the crossover operation on the frequency deviations of two consecutive parent individual sub-arrays.

3. The subarray-level distributed frequency-controlled array sidelobe suppression method according to claim 1, characterized in that: In step S5, executing the mutation operation specifically includes: performing a mutation operation on the individual within a single-row multi-column real-valued aperture constraint range under a mutation probability condition, and similarly performing a mutation operation on the individual within a corresponding frequency deviation constraint range.

Images