Intelligent arrangement method for array antenna elements and clustering epigenetic optimization method

Abstract

The present invention relates to the technical field of artificial intelligence for antenna array deployment, in particular to an intelligent arrangement method for array antenna elements and a clustering epigenetic optimization method. The intelligent arrangement method for array antenna elements comprises: performing encoding for a J_K array, and initializing an epigenetic algorithm so as to generate J+K genes. The present invention uses a clustering algorithm in machine learning to divide a population into sub-populations having different characteristics. Adaptive learning is performed on a portion of individuals in each type, and an epigenetics-based inheritance operation is performed on the basis of the fitness probability of population individuals. Then, minor random mutation operations are performed to generate a new sub-population. Algorithm hyperparameters such as population size and mutation probability participate in the evolution of individual genes, and adaptively change on the basis of evolutionary states. The invention solves the problems of low convergence speed and poor obtained results in optimization processes of current L-shaped array antenna system arrangement algorithms.

Description

Intelligent arrangement method of array antenna elements and optimization method of clustering epigenetic Technical Field

[0001] The present invention relates to the field of artificial intelligence technology for antenna array layout, and in particular to an intelligent arrangement method for array antenna elements and an optimization method for clustering epigenetics. Background Art

[0002] In recent years, the industrial application of artificial intelligence has developed rapidly, including its application in the optimized design of array element arrangements for array antennas. Compared to uniform rectangular two-dimensional array antennas, L-shaped array antennas have advantages such as simple structure and good array layout. However, their direct beamforming pattern performance is poor. Due to the small number of array elements, their angular resolution and accuracy require intelligent optimization design. Therefore, the optimized layout of the L-shaped array is very important for beamforming and beam pattern availability. By intelligently optimizing the L-shaped array layout, the advantages of the L-shaped array's simple structure and small number of elements can be further enhanced, while also improving its disadvantages, that is, optimizing its beamforming pattern performance.

[0003] In the prior art, the related authorized patent "A New Planar Molecular Array Antenna Array Comprehensive Arraying Method Based on an Improved Genetic Algorithm (CN106099393B)" and the previously authorized patent of this application "A Method for Arranging Element of an L-Shaped Array Antenna Based on Acquired Genetics (CN107275801B)" have improved optimization methods and results, including beam pattern angular resolution and accuracy. However, these existing patented methods are still prone to shortcomings such as premature maturation, resulting in the resulting array antenna element arrangement results still not reaching the global optimum. This, in turn, prevents their beamforming and beam pattern optimization methods from achieving stable optimal results. To improve the overall and local optimization capabilities of intelligent optimization algorithms, most current solutions choose to combine two algorithms, such as a genetic algorithm and an annealing algorithm. Although optimizing with two or more algorithms can achieve relatively good results, these existing methods suffer from problems such as high computational complexity and slow optimization speed, and their global and local search capabilities need to be further improved. Therefore, intelligent arrangement methods and systems for array antennas need to be improved or enhanced.

[0004] Summary of the Invention

[0005] In view of the problems of the prior art, the present invention provides an array antenna element intelligent arrangement method and a cluster epigenetic optimization method.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] The present invention provides an intelligent arrangement method for array antenna elements, wherein the J_K array is an L-shaped array antenna having two adjacent boundary rows of array elements of J and K, respectively; the method comprises the following steps:

[0008] S1. Encode the J_K array and initialize the epigenetic algorithm: treat the J_K array as a chromosome. When forming individual genes, use J+K groups of randomly generated binary strings to represent the J_K array. The number of bits in the binary string is Na. Each binary string is called a gene on the chromosome. Each binary string represents the distance between the array element and the previous array element. Use the above method to generate J+K genes.

[0009] S2. Determine the hyperparameters to be optimized and generate L hyperparameter genes;

[0010] S3. Save the current J+K+L genes as the initial population of the genetic algorithm. For the convenience of representation, the total number of genes in the chromosome J+K+L is represented by d, and d=J+K+L;

[0011] S4, at this time each chromosome is recorded as The gene string is composed of Expressed as in represents the gene, j represents the sequence number of the gene in the chromosome; population Where k is the number of generations of population evolution; i is the number of chromosomes in the population; N G Indicates the population size;

[0012] S5, the initial population G k Make an adjustment; then calculate the population G k Each chromosome The fitness of F i ;

[0013] S6. Use clustering algorithm to cluster G k Divide into K different sub-populations, select the best individuals in each sub-population for adaptive learning and update fitness;

[0014] S7, perform epigenetic manipulation to obtain offspring G' k ;

[0015] S8. Randomly select M parents Calculate the fitness ratio of each parent

[0016] S9, M parents build the current gene pool

[0017] S10, according to the fitness ratio p of each parent individual, the roulette wheel selects the gene The larger the proportion of individuals, the greater the probability that their current gene will be selected;

[0018] S11, offspring Write the corresponding gene position

[0019] S12, j=j+1;

[0020] S13, repeat steps S8 to S12 until j=d;

[0021] S14. Producing offspring individuals

[0022] S15, repeat N G From S8 to S14, a temporary new population G' is obtained k ;

[0023] S16, according to the mutation probability p m Perform mutation operation to generate a new population G k+1 ;

[0024] S17. Calculate population G k+1 Update the algorithm hyperparameters for each chromosome hyperparameter gene and calculate the fitness of the updated population;

[0025] S18, repeating iterative steps S6 to S17 until a pre-set termination condition is met, and multiple optimal populations are obtained;

[0026] S19. According to different situations and requirements, one of the required optimal population genes is output and decoded into the array element arrangement of the L-shaped array antenna.

[0027] Wherein, in step S5, the adjustment includes converting the binary character string into a decimal number.

[0028] The adjustment method for adjusting the initial population in step S5 is:

[0029] First, convert each generation of J+K binary strings into decimal numbers. The decimal value after the binary string conversion corresponds to the array element spacing between the current array element and the previous array element. That is, after restoring the binary string, the array element spacing D is obtained.

[0030] When calculating the positions of the first J array elements, generate each element spacing D and count them, accumulating the value of the overall aperture. If the accumulated value of the element spacing D is about to exceed the maximum array aperture Da, the element spacing of the subsequent elements is forcibly adjusted to 1.

[0031] The adjustment method for the last K array elements is the same as that for the first J array elements.

[0032] Preferably, the maximum aperture Da of the array is 57-65.

[0033] Preferably, in step S18, the pre-set termination condition includes that the optimal individual fitness value does not change by more than 10 -6 and / or reaching a pre-set maximum number of simulations.

[0034] The present invention also provides an optimization method for clustering epigenetic inheritance, which comprises the following steps:

[0035] A1. Let d represent the total number of genes in a chromosome, J+K+L, so d=J+K+L.

[0036] A2. At this time, each chromosome is recorded as The gene string is composed of Expressed as in represents the gene, j represents the sequence number of the gene in the chromosome; population Where k is the number of generations of population evolution; i is the number of chromosomes in the population; N G Indicates the population size;

[0037] A3, the initial population G k Make an adjustment; then calculate the population G k Each chromosome The fitness of F i ;

[0038] A4. Use clustering algorithm to cluster G k Divide into K different sub-populations, select the best individuals in each sub-population for adaptive learning and update fitness;

[0039] A5. Perform epigenetic manipulation to obtain offspring G' k ;

[0040] A6. Randomly select M parents Calculate the fitness ratio of each parent

[0041] A7, M parents build the current gene pool

[0042] A8. According to the fitness ratio p of each parent individual, the roulette wheel selects the gene The larger the proportion of individuals, the greater the probability that their current gene will be selected;

[0043] A9, offspring Write the corresponding gene position

[0044] A10, j=j+1;

[0045] A11. Repeat steps A6 to A10 until j = d.

[0046] A12. Producing offspring individuals

[0047] A13, repeat N G A6~A12, get a temporary new population G' k ;

[0048] A14. According to the mutation probability p m Perform mutation operation to generate a new population G k+1 ;

[0049] A15. Calculate the population G k+1 Update the algorithm hyperparameters for each chromosome hyperparameter gene and calculate the fitness of the updated population;

[0050] A16. Repeat steps A4 to A15 until a pre-set termination condition is met, and multiple optimal populations are obtained.

[0051] Beneficial effects of the present invention:

[0052] The present invention first encodes the J_K array of the L-shaped array antenna, randomly generates an initial population, then calculates the fitness value of each individual in the array population to be optimized, and uses the clustering algorithm in machine learning to divide the population into sub-populations with different characteristics; some individuals in each category undergo adaptive learning, and perform epigenetic inheritance operations based on the probability of individual fitness in the population, and then perform a small amount of random mutation operations to generate new sub-populations; algorithm hyperparameters such as population size and mutation probability participate in the evolution of individual genes and adaptively change according to the evolutionary state; the above operations are repeated iteratively until a preset stop condition is met to obtain multiple optimal sub-populations; then the user outputs the optimal array element arrangement of the L-shaped array antenna according to different required application scenarios; the present invention application has outstanding intelligent arrangement characteristics and significant antenna performance improvements, solves the problems of slow convergence speed and poor results in the optimization process of the arrangement algorithm of the current L-shaped array antenna system, and is suitable for the array element arrangement setting of L-shaped and other array antennas. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] FIG1 is a flow chart of a method for intelligently arranging array elements of an array antenna according to the present invention.

[0054] FIG2 is a graph showing that the clustering epigenetic algorithm of the present invention converges to multiple excellent solutions.

[0055] FIG3 is a mixed coding diagram of algorithm parameters and hyperparameters for clustering epigenetics of the present invention.

[0056] Figure 4. Diagram of epigenetic manipulation.

[0057] FIG5 is a graph comparing the convergence speeds of the clustering epigenetic algorithm of the present invention and the acquired genetic algorithm of the prior art patent (CN107275801B). DETAILED DESCRIPTION

[0058] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the embodiments and the accompanying drawings. The contents mentioned in the embodiments are not intended to limit the present invention. The present invention will be described in detail below with reference to the accompanying drawings.

[0059] Example 1

[0060] A method for intelligently arranging array elements of an array antenna, wherein a J_K array is an L-shaped array antenna having two adjacent rows of array elements, each having J and K numbers of array elements, respectively; the method comprises the following steps:

[0061] S1. Encode the J_K array and initialize the epigenetic algorithm: treat the J_K array as a chromosome. When forming individual genes, use J+K groups of randomly generated binary strings to represent the J_K array. The number of bits in the binary string is Na. Each binary string is called a gene on the chromosome. Each binary string represents the distance between the array element and the previous array element. Use the above method to generate J+K genes.

[0062] S2. Determine the hyperparameters to be optimized and generate L hyperparameter genes;

[0063] S3. Save the current J+K+L genes as the initial population of the genetic algorithm. For the convenience of representation, the total number of genes in the chromosome J+K+L is represented by d, and d=J+K+L;

[0064] S4, at this time each chromosome is recorded as The gene string is composed of Expressed as in represents the gene, j represents the sequence number of the gene in the chromosome; population Where k is the number of generations of population evolution; i is the number of chromosomes in the population; N G Indicates the population size;

[0065] S5, the initial population G k Make an adjustment; then calculate the population G k Each chromosome The fitness of F i; The adjustments include converting binary strings into decimal numbers;

[0066] S6. Use clustering algorithm to cluster G k Divide into K different sub-populations, select the best individuals in each sub-population for adaptive learning and update fitness;

[0067] S7, perform epigenetic manipulation to obtain offspring G' k ;

[0068] S8. Randomly select M parents Calculate the fitness ratio of each parent

[0069] S9, M parents build the current gene pool

[0070] S10, according to the fitness ratio p of each parent individual, the roulette wheel selects the gene The larger the proportion of individuals, the greater the probability that their current gene will be selected;

[0071] S11, offspring Write the corresponding gene position

[0072] S12, j=j+1;

[0073] S13, repeat steps S8 to S12 until j=d;

[0074] S14. Producing offspring individuals

[0075] S15, repeat N G From S8 to S14, a temporary new population G' is obtained k ;

[0076] S16, according to the mutation probability p m Perform mutation operation to generate a new population G k+1 ;

[0077] S17. Calculate population G k+1 Update the algorithm hyperparameters for each chromosome hyperparameter gene and calculate the fitness of the updated population;

[0078] S18, repeating iterative steps S6 to S17 until the pre-set termination condition is met, and multiple optimal populations are obtained; wherein the pre-set termination condition includes that the fitness of the optimal individual does not change by more than 10 -6or / and reaching a preset maximum number of simulations, such as 40,000; the optimal individual satisfies a preset gain requirement, such as 20 dB, or an antenna aperture efficiency, such as 80%;

[0079] S19. Based on different situations and requirements, output one of the required optimal population genes and decode it into an element arrangement of an L-shaped array antenna; wherein the entire population is divided into different sub-populations, and the different situations and requirements include the antenna's main lobe gain, antenna opening utilization, or corresponding beam width, etc.

[0080] Specifically, through the above embodiment, the J_K array of the L-shaped array antenna is first encoded, and an initial population can be randomly generated. Then, the fitness value of each individual in the array population to be optimized is calculated, and the clustering algorithm in machine learning is used to divide the population into sub-populations with different characteristics; some individuals in each category undergo adaptive learning, and epigenetic inheritance operations are performed based on the probability of individual fitness in the population, and then a small amount of random mutation operations are performed to generate new sub-populations; algorithm hyperparameters such as population size and mutation probability participate in the evolution of individual genes and adaptively change according to the evolutionary state; the above operation is repeated iteratively until a preset stop condition is met to obtain multiple optimal sub-populations; then the user outputs the optimal element arrangement of the L-shaped array antenna according to different required application scenarios; the present invention application has outstanding intelligent arrangement characteristics and significant antenna performance improvements, solves the problems of slow convergence speed and poor results in the optimization process of the arrangement algorithm of the current L-shaped array antenna system, and is suitable for the element arrangement setting of L-shaped and other array antennas.

[0081] Example 2

[0082] In the second embodiment of the present application, the adjustment method for adjusting the initial population in step S5 is:

[0083] First, convert each generation of J+K binary strings into decimal numbers. The decimal value after the binary string conversion corresponds to the array element spacing between the current array element and the previous array element. That is, after restoring the binary string, the array element spacing D is obtained.

[0084] When calculating the positions of the first J array elements, generate each element spacing D and count them, accumulating the value of the overall aperture. If the accumulated value of the element spacing D is about to exceed the maximum array aperture Da, the element spacing of the subsequent elements is forcibly adjusted to 1.

[0085] The adjustment method for the last K array elements is the same as that for the first J array elements.

[0086] Example 3

[0087] In the third embodiment of the present application, the maximum aperture Da of the array is 57 to 65.

[0088] Example 4

[0089] In Example 4 of the present application, a clustering epigenetic optimization method is provided, which includes the following steps:

[0090] A1. Let d represent the total number of genes in a chromosome, J+K+L, so d=J+K+L.

[0091] A2. At this time, each chromosome is recorded as The gene string is composed of Expressed as in represents the gene, j represents the sequence number of the gene in the chromosome; population Where k is the number of generations of population evolution; i represents the number of chromosomes in the population; N G Indicates the population size;

[0092] A3, the initial population G k Make an adjustment; then calculate the population G k Each chromosome The fitness of F i ;

[0093] A4. Use clustering algorithm to cluster G k Divide into K different sub-populations, select the best individuals in each sub-population for adaptive learning and update fitness;

[0094] A5. Perform epigenetic manipulation to obtain offspring G' k ;

[0095] A6. Randomly select M parents Calculate the fitness ratio of each parent

[0096] A7, M parents build the current gene pool

[0097] A8. According to the fitness ratio p of each parent individual, the roulette wheel selects the gene The larger the proportion of individuals, the greater the probability that their current gene will be selected;

[0098] A9, offspring Write the corresponding gene position

[0099] A10, j=j+1;

[0100] A11. Repeat steps A6 to A10 until j = d.

[0101] A12. Producing offspring individuals

[0102] A13, repeat N G A6~A12, get a temporary new population G' k ;

[0103] A14. According to the mutation probability p m Perform mutation operation to generate a new population G k+1 ;

[0104] A15. Calculate the population G k+1 Update the algorithm hyperparameters for each chromosome hyperparameter gene and calculate the fitness of the updated population;

[0105] A16. Repeat steps A4 to A15 until a pre-set termination condition is met, and multiple optimal populations are obtained.

[0106] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention is disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes by using the technical content disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments according to the technology of the present invention are all within the scope of the technical solution of the present invention without departing from the content of the technical solution of the present invention.

Claims

1. A method for intelligently arranging array elements of an array antenna, wherein the J_K array is an array in which the number of array elements in two adjacent rows of an L-shaped array antenna is J and K respectively; characterized in that: The following steps are involved: S1. Encode the J_K array and initialize the epigenetic algorithm: take the J_K array as a chromosome. When forming the genes of an individual, use J+K groups of randomly generated binary strings to represent the J_K array. The number of bits of the binary string is Na. Each binary string is called a gene on the chromosome. The meaning of each binary string is the distance between the array element and the previous array element. Use the above method to generate J+K genes. S2. Determine the hyperparameters that need to be optimized and generate L hyperparameter genes; S3. Save the current J+K+L genes as the initial population of the genetic algorithm. For the convenience of representation, the total number of genes in the chromosome J+K+L is represented by d, and d=J+K+L; S4, at this time each chromosome is recorded as The gene string is composed of Expressed as in represents the gene, j represents the sequence number of the gene in the chromosome; population Where k is the number of generations of population evolution; i is the number of chromosomes in the population; N G Indicates the population size; S5, the initial population G k Make an adjustment; then calculate the population G k Each chromosome The fitness of F i ; S6. Use clustering algorithm to cluster G k Divide into K different sub-populations, select the best individuals in each sub-population for adaptive learning and update fitness; S7, perform epigenetic manipulation to obtain offspring G' k ; S8. Randomly select M parents Calculate the fitness ratio of each parent j＝1； S9, M parents build the current gene pool S10, according to the fitness proportion p of each parent individual, the roulette wheel selects the gene n∈[1,M], where the individual with a larger proportion has a greater probability of having its current gene selected; S11, offspring Write the corresponding gene position S12, j=j+1; S13, repeat steps S8 to S12 until j=d; S14. Produce offspring individuals S15, repeat N G S8 to S14, get a temporary new population G' k ; S16. According to the mutation probability p m Perform mutation operation to generate a new population G k+1 ; S17. Calculate the population G k+1 Update the algorithm hyperparameters for each chromosome hyperparameter gene and calculate the fitness of the updated population; S18, repeating iterative steps S6 to S17 until a preset termination condition is met, and multiple optimal populations are obtained; S19. According to different situations and requirements, one of the required optimal population genes is output and decoded into the array element arrangement of the L-shaped array antenna.

2. The method for intelligently arranging array elements of an array antenna according to claim 1, characterized in that: In the step S5, the adjustment includes converting the binary string into a decimal number.

3. The method for intelligently arranging array elements of an array antenna according to claim 1, characterized in that: The adjustment method described in step S5 for adjusting the initial population is: First, convert each generation of J+K binary strings into decimal numbers. The value of the decimal number after the binary string is converted corresponds to the array element spacing between the array element and the previous array element. That is, the array element spacing D is obtained after restoring the binary string; When calculating the positions of the first J array elements, each array element spacing D is generated and counted, and the value of the overall aperture is accumulated. If the accumulated value of the array element spacing D is about to exceed the maximum aperture Da of the array, the array element spacing of the following array elements is forcibly adjusted to 1; The adjustment method for the last K array elements is the same as that for the first J array elements.

4. The method for intelligently arranging array elements of an array antenna according to claim 3, characterized in that: The maximum aperture Da of the array is 57-65.

5. The method for intelligently arranging array elements of an array antenna according to claim 1, characterized in that: In step S18, the pre-set termination condition includes that the change in the fitness of the best individual does not exceed 10 -6 and / or a pre-set maximum number of simulations is reached.

6. A cluster epigenetic optimization method, comprising the following steps: A1. Let the total number of genes in the chromosome J+K+L be represented by d, and we have d=J+K+L; A2. At this time, each chromosome is recorded as The gene string is composed of Expressed as in represents the gene, j represents the sequence number of the gene in the chromosome; population Where k is the number of generations of population evolution; i is the number of chromosomes in the population; N G Indicates the population size; A3. Set the initial population G k Make an adjustment; then calculate the population G k Each chromosome The fitness of F i ; A4. Use clustering algorithm to cluster G k Divide into K different sub-populations, select the best individuals in each sub-population for adaptive learning and update fitness; A5. Perform epigenetic manipulation to obtain offspring G' k ; A6. Randomly select M parents Calculate the fitness of each parent The proportion of j＝1； A7, M parents build the current gene pool A8. According to the fitness proportion p of each parent individual, the roulette wheel selects the gene n∈[1,M], where the individual with a larger proportion has a greater probability of having its current gene selected; A9, offspring Write the corresponding gene position A10, j=j+1; A11, repeat steps A6 to A10 until j=d; A12. Producing offspring individuals A13, repeat N G A6 to A12, get a temporary new population G' k ; A14. According to the mutation probability p m Perform mutation operation to generate a new population G k+1 ; A15. Calculate the population G k+1 Update the algorithm hyperparameters for each chromosome hyperparameter gene and calculate the fitness of the updated population; A16. Repeat steps A4 to A15 until a preset termination condition is met, and multiple optimal populations are obtained.

Images