A Sparse Array Beam Synthesis Method Based on an Improved Genetic Algorithm
By improving the genetic algorithm and combining it with the differential evolution algorithm, the selection, crossover and mutation probabilities are adaptively adjusted, which solves the problem of weak local search capability of the genetic algorithm in sparse array beam synthesis and realizes the design of sparse array with low sidelobes and high gain.
Patent Information
- Application Number
- CN202310689730.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-12
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2043-06-12
AI Technical Summary
Genetic algorithms have weak local search capabilities in sparse array beam synthesis, are prone to getting trapped in local optima, converge slowly, and are difficult to achieve low sidelobes without amplitude and phase weighting while ensuring gain.
An improved genetic algorithm is combined with a differential evolution algorithm. The population is generated and mixed through differential evolution, and the selection, crossover and mutation probabilities are adaptively adjusted to improve global search capability and convergence speed, and avoid getting trapped in local optima.
This method enables low sidelobes in sparse arrays without amplitude and phase weighting, improving the radiation performance of sparse arrays and meeting the design requirements for gain and sidelobe level.
Smart Images

Figure CN116738840B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of array antenna beam synthesis technology, and specifically refers to a sparse array beam synthesis method based on an improved genetic algorithm. Background Technology
[0002] Array antennas, characterized by low sidelobes, strong directivity, and ease of beamforming, are widely used in radar and communication systems. Uniform array antennas, due to their ease of theoretical calculation and simple assembly, have also gained widespread application. However, in large arrays, uniform arrays have high heat density, and to avoid grating lobes, the element spacing is typically less than λ / 2, resulting in strong mutual coupling effects. Compared to uniform arrays, sparse arrays offer advantages such as lower mutual coupling effects, low sidelobes without amplitude-phase weighting, suppression of grating lobes through aperiodic arrangement, and lower antenna system cost. This makes sparse arrays promising for applications in military, astronomy, communications, remote sensing, and deep space exploration.
[0003] To obtain a radiation pattern that meets specific requirements, beam synthesis optimization of the array antenna is a key technical step in sparse array design. Common beam synthesis methods can be divided into analytical methods, numerical methods, and intelligent optimization algorithms. Due to their efficiency and convenience, intelligent optimization algorithms have become the commonly used beam synthesis method.
[0004] Genetic algorithms are a classic global optimization algorithm. Based on biological evolution, they simulate the natural world's "survival of the fittest" and "selection of the weak" principles. Genetic algorithms encode the parameters of the problem to be solved as chromosomes and then use selection, crossover, and mutation operators to iteratively exchange chromosome information within the population. In each iteration, the fitness value of individuals in the population is calculated. Following natural selection, individuals with better fitness are retained for the next generation, while those with poorer fitness are eliminated. This iterative process allows the population to continuously search for higher fitness. However, genetic algorithms have weak local search capabilities, are prone to getting trapped in local optima, and converge slowly. Summary of the Invention
[0005] In view of this, the present invention proposes a sparse array beam synthesis method based on an improved genetic algorithm. This method improves the traditional genetic algorithm and enables the sparse array to achieve low sidelobes without amplitude and phase weighting while ensuring gain.
[0006] The technical solution adopted in this invention is as follows:
[0007] A sparse array beam synthesis method based on an improved genetic algorithm includes the following steps:
[0008] (1) Determine the fitness function based on the required array pattern specifications;
[0009] (2) Create an initial population P1(0) and set optimization parameters, including the maximum number of iterations D, the iteration counter d, the population size p, and the selection probability P. i Initial crossover probability P c0 Minimum crossover probability P c_min Initial mutation probability P m0 Minimum mutation probability P m_min ;
[0010] (3) Generate a population P2(d) of the same size as the current population using the differential evolution algorithm, and mix it with the current population;
[0011] (4) Calculate the fitness of the mixed population and sort them by fitness, retaining the top half of the population P(d) with better fitness;
[0012] (5) Apply the selection operator to the population, based on the selection probability P i Perform a selection operation;
[0013] (6) Apply the crossover operator to the population selected in step (5), and follow the two-point crossover strategy with an adaptive crossover probability P. c The exchange of some chromosomes between parent individuals produces new offspring individuals;
[0014] (7) Apply the mutation operator to the population after the crossover operation in step (6) with an adaptive mutation probability P. m To alter several gene loci in an individual;
[0015] After the selection, crossover, and mutation operators are applied, a new generation population P(d+1) is obtained;
[0016] (8) Calculate the fitness of the new generation population and determine whether the fitness of the best individual meets the termination condition. If the termination condition is met, the iteration ends and the optimization result is output. If the termination condition is not met, return to step (3) and continue the loop iteration until the termination condition is met or d > D.
[0017] Furthermore, the fitness function in step (1) is:
[0018] f = |SLL i_0 -SLL i_max |
[0019] Where f is the fitness of individual i, SLL i_0 For the target sidelobe level of the sparse array, SLL i_maxThe maximum sidelobe level is sparse.
[0020] Furthermore, step (5) is specifically performed as follows: based on the selection probability P i Select N individuals with good fitness from the population, then randomly select n individuals from among them, and finally select the best individual from these n individuals as the parent for the next crossover operation.
[0021] Furthermore, the adaptive crossover probability P in step (6) c for:
[0022]
[0023] Among them, P c0 Let P be the initial crossover probability. c_min For the minimum crossover probability, f avg f is the average fitness value of the entire population. min This represents the minimum fitness value for the entire population.
[0024] Furthermore, the adaptive mutation probability P in step (7) m for:
[0025]
[0026] Among them, P m0 Let P be the initial mutation probability. m_min Let f be the minimum mutation probability, d be the current iteration number, D be the maximum iteration number, and f be the maximum mutation probability. avg f is the average fitness value of the entire population. min This represents the minimum fitness value for the entire population.
[0027] The beneficial effects of this invention are as follows:
[0028] 1. This invention improves the genetic algorithm and combines it with the differential evolution algorithm to propose an improved genetic algorithm.
[0029] 2. In order to improve the global search capability of the algorithm, this invention uses the differential evolution algorithm to generate a population of the same size as the original population before each iteration. The two populations are then mixed, reordered and selected before the three operators of the genetic algorithm are applied.
[0030] 3. To avoid getting trapped in local optima, this invention improves the selection operator, which greatly increases the probability of selecting excellent individuals while retaining the possibility of selecting poor individuals, thereby improving the convergence speed of the algorithm and maintaining the diversity of the population.
[0031] 4. In order to ensure the effectiveness and efficiency of optimization, this invention can adaptively adjust the crossover probability and mutation probability at different optimization stages to achieve good optimization results. Attached Figure Description
[0032] Figure 1 This is a flowchart of the sparse array beam synthesis method based on the improved genetic algorithm in this invention;
[0033] Figure 2 This is a diagram showing the optimized sparse array cell distribution in Embodiment 1 of the present invention;
[0034] Figure 3 This is the optimized convergence performance curve in Embodiment 1 of the present invention;
[0035] Figure 4 This is the optimized array pattern in Embodiment 1 of the present invention;
[0036] Figure 5 This is a diagram showing the optimized sparse array cell distribution in Embodiment 2 of the present invention;
[0037] Figure 6 This is the optimized convergence performance curve in Embodiment 2 of the present invention;
[0038] Figure 7 This is the optimized array pattern in Embodiment 2 of the present invention. Detailed Implementation
[0039] To more clearly demonstrate the purpose, technical solution, advantages, etc. of this invention, the invention will be further described in detail below with reference to the accompanying drawings and examples.
[0040] A sparse array beam synthesis method based on an improved genetic algorithm includes the following steps:
[0041] Step 1: Based on the required array pattern specifications, determine the fitness function of the algorithm. The fitness function is set as Equation (1), and the fitness of an individual i can be expressed as:
[0042] f = |SLL i_0 -SLL i_max | (1)
[0043] Among them, SLL i_0 For the target sidelobe level of the sparse array, SLL i_max The maximum sidelobe level is sparse.
[0044] Step 2: Set optimization parameters and create an initial population. Create an initial population P1(0), set the number of iterations D, the iteration counter d, the population size p, and the selection probability P. i Crossover probability P c Probability of mutation P m Parameters such as these.
[0045] Step 3: Use the differential evolution algorithm to generate a population P2(d) of the same size as the current population, and mix it with the current population.
[0046] Step 4: Calculate the fitness of the mixed population and sort them by fitness. Retain the top half of the population P(d) with better fitness to improve the global search capability of the algorithm.
[0047] Step 5: Apply the selection operator to the population and perform the selection operation according to the improved selection operator method. First, select N individuals with good fitness from the population. Then, randomly select n individuals from these N individuals. Finally, select the best individual from these n individuals as the parent for the next crossover operation. This improvement significantly increases the probability of selecting excellent individuals while retaining the possibility of selecting weaker individuals, thus improving the convergence speed of the algorithm and maintaining the diversity of the population.
[0048] Step 6: Apply the crossover operator to the population and perform the crossover operation according to the improved crossover operator. Following the "two-point crossover" strategy, use an adaptive crossover probability P. c Exchanging partial chromosomes between parent individuals produces new offspring; the adaptive crossover probability P c It can be expressed as equation (2):
[0049]
[0050] Among them, P c0 Let P be the initial crossover probability. c_min For the minimum crossover probability, f avg f is the average fitness function value of the population. min This is the minimum fitness function value of the population.
[0051] Step 7: Apply the mutation operator to the population and perform the mutation operation according to the improved mutation operator. Use an adaptive mutation probability P. m Changing several gene loci in an individual. The adaptive mutation probability P. m It can be expressed as equation (3):
[0052]
[0053] Among them, P m0 Let P be the initial mutation probability. m_min Let f be the minimum mutation probability, d be the current iteration number, D be the total number of iterations, and f be the total mutation probability. avg f is the average fitness function value of the population. min This is the minimum fitness function value of the population.
[0054] The genetic algorithm is used sequentially to select, crossover, and mutate the next generation population P(d+1).
[0055] Step 8: Calculate the fitness value of the new generation population and determine whether the fitness value of the optimal individual satisfies the algorithm's termination condition. If the termination condition is met, the iteration ends and the optimization result is output. If the termination condition is not met, return to Step 3 and continue the iteration loop until the termination condition is met, or d > D.
[0056] Step 9: Verify whether the current optimization result meets the index requirements. If not, return to Step 1 and modify the fitness function and the parameters required for each optimization.
[0057] Here are some specific examples:
[0058] Example 1:
[0059] The sparse array operates at a frequency of 15MHz, with a full array size of 70×70 before sparsening, a specified sparsity of 80%, a scan angle of 60°, and a cell spacing of 10m. The sparse planar array is optimized by adjusting the cell arrangement to reduce the sidelobe level, with a target gain of 32dB and a target sidelobe level of -22dB. To ensure the sparse array maintains its maximum aperture size, the four corner positions of the grid plane are not sparsed.
[0060] like Figure 1 As shown, the steps of sparse array beam synthesis are as follows:
[0061] Step 1: Based on the required array pattern specifications, determine the fitness function of the algorithm. The fitness function is set as Equation (4), and the fitness of an individual i can be expressed as:
[0062] f = |SLL i_0 -SLL i_max | (4)
[0063] Among them, SLL i_0 For the target sidelobe level of the sparse array, SLL i_max The maximum sidelobe level is sparse.
[0064] Step 2: Set optimization parameters and create an initial population. Create an initial population P1(0) using binary encoding, where "0" indicates a sparse cell and "1" indicates a retained cell. Since a uniform array has 4900 cells, and the optimization objective requires that the cells at the four corners of the array not be sparse, the individual's gene string length is set to 4896. Set the iteration count D = 500, the iteration counter d = 0, the population size p = 50, and the selection probability P... i Crossover probability P c Probability of mutation P m Parameters such as these.
[0065] Step 3: Use the differential evolution algorithm to generate a population P2(d) of the same size as the current population, and mix it with the current population.
[0066] Step 4: Calculate the fitness of the mixed population and sort them by fitness. Retain the top half of the population P(d) with better fitness to improve the global search capability of the algorithm.
[0067] Step 5: Apply the selection operator to the population and perform the selection operation according to the improved selection operator method. First, select 10 individuals with good fitness from the population. Then, randomly select 4 individuals from these 10. Finally, select the best individual from these 4 individuals as the parent for the next crossover operation. This improvement significantly increases the probability of selecting excellent individuals while retaining the possibility of selecting weaker individuals, thus improving the convergence speed of the algorithm and maintaining the diversity of the population.
[0068] Step 6: Apply the crossover operator to the population and perform the crossover operation according to the improved crossover operator. Following the "two-point crossover" strategy, use an adaptive crossover probability P. c Exchanging partial chromosomes between parent individuals produces new offspring; the adaptive crossover probability P c It can be expressed as equation (5):
[0069]
[0070] Among them, P c0 Let P be the initial crossover probability. c_min For the minimum crossover probability, f avg f is the average fitness function value of the population. min This is the minimum fitness function value of the population.
[0071] Step 7: Apply the mutation operator to the population and perform the mutation operation according to the improved mutation operator. Use an adaptive mutation probability P. m Changing several gene loci in an individual. The adaptive mutation probability P. m It can be expressed as equation (6):
[0072]
[0073] Among them, P m0 Let P be the initial mutation probability. m_min Let f be the minimum mutation probability, d be the current iteration number, D be the total number of iterations, and f be the total mutation probability. avg f is the average fitness function value of the population. min This is the minimum fitness function value of the population.
[0074] The genetic algorithm is used sequentially to select, crossover, and mutate the next generation population P(d+1).
[0075] Step 8: Calculate the fitness value of the new generation population and determine whether the fitness value of the optimal individual satisfies the algorithm's termination condition. If the termination condition is met, the iteration ends and the optimization result is output. If the termination condition is not met, return to Step 3 and continue the iteration loop until the termination condition is met, or d > D.
[0076] Step 9: Verify whether the current optimization result meets the index requirements. If not, return to Step 1 and modify the fitness function and the parameters required for each optimization.
[0077] After 500 iterations of optimization using an improved genetic algorithm, the sparse optimized cell distribution diagram is obtained as follows: Figure 2 As shown, the number of array elements after sparsening is 980. The convergence performance curve is as follows. Figure 3 As shown in the figure, the horizontal axis represents the number of iterations, and the vertical axis represents the fitness function value. The figure shows that after 210 iterations, the fitness function value tends to stabilize. This indicates that the hybrid optimization algorithm has a relatively fast convergence speed, and the optimization results are relatively stable with small variance after multiple iterations. The optimized sparse array pattern is shown below. Figure 4 As shown in the figure, after sparse optimization using the improved genetic algorithm, the array achieves a gain of 32.56 dB and a sidelobe level of -22.66 dB at a 60° scan angle, both meeting the design requirements and demonstrating good radiation performance. Therefore, beam synthesis optimization using the algorithm of this invention can achieve a sparse array that meets design requirements.
[0078] Example 2:
[0079] The sparse array operates at a frequency of 2.3 MHz, with a full array size of 10 × 20 before sparsening, a specified sparsity of 46%, a scanning direction along the normal direction, and an element spacing of 65.2 mm. The sparse planar array is optimized by adjusting the element arrangement to reduce the sidelobe level, with a target gain of 26 dB and a target sidelobe level of -21 dB. To ensure the sparse array maintains its maximum aperture size, the four corner positions of the grid plane are not sparsed.
[0080] like Figure 1 As shown, the steps of sparse array beam synthesis are as follows:
[0081] Step 1: Based on the required array pattern specifications, determine the fitness function of the algorithm. The fitness function is set as Equation (7), and the fitness of an individual i can be expressed as:
[0082] f = |SLL i_0 -SLL i_max | (7)
[0083] Among them, SLL i_0For the target sidelobe level of the sparse array, SLL i_max The maximum sidelobe level is sparse.
[0084] Step 2: Set optimization parameters and create an initial population. Create an initial population P1(0) using binary encoding, where "0" indicates a sparse cell and "1" indicates a retained cell. Since a uniform array has 200 cells, and the optimization objective requires that the cells at the four corner positions of the array not be sparse, the individual's gene string length is set to 196. Set the iteration count D = 500, the iteration counter d = 0, the population size p = 50, and the selection probability P... i Crossover probability P c Probability of mutation P m Parameters such as these.
[0085] Step 3: Use the differential evolution algorithm to generate a population P2(d) of the same size as the current population, and mix it with the current population.
[0086] Step 4: Calculate the fitness of the mixed population and sort them by fitness. Retain the top half of the population P(d) with better fitness to improve the global search capability of the algorithm.
[0087] Step 5: Apply the selection operator to the population and perform the selection operation according to the improved selection operator method. First, select 10 individuals with good fitness from the population. Then, randomly select 4 individuals from these 10. Finally, select the best individual from these 4 individuals as the parent for the next crossover operation. This improvement significantly increases the probability of selecting excellent individuals while retaining the possibility of selecting weaker individuals, thus improving the convergence speed of the algorithm and maintaining the diversity of the population.
[0088] Step 6: Apply the crossover operator to the population and perform the crossover operation according to the improved crossover operator. Following the "two-point crossover" strategy, use an adaptive crossover probability P. c Exchanging partial chromosomes between parent individuals produces new offspring; the adaptive crossover probability P c It can be expressed as equation (8):
[0089]
[0090] Among them, P c0 Let P be the initial crossover probability. c_min For the minimum crossover probability, f avg f is the average fitness function value of the population. min This is the minimum fitness function value of the population.
[0091] Step 7: Apply the mutation operator to the population and perform the mutation operation according to the improved mutation operator. Use an adaptive mutation probability P.m Changing several gene loci in an individual. The adaptive mutation probability P. m It can be expressed as equation (9):
[0092]
[0093] Among them, P m0 Let P be the initial mutation probability. m_min Let f be the minimum mutation probability, d be the current iteration number, D be the total number of iterations, and f be the total mutation probability. avg f is the average fitness function value of the population. min This is the minimum fitness function value of the population.
[0094] The genetic algorithm is used sequentially to select, crossover, and mutate the next generation population P(d+1).
[0095] Step 8: Calculate the fitness value of the new generation population and determine whether the fitness value of the optimal individual satisfies the algorithm's termination condition. If the termination condition is met, the iteration ends and the optimization result is output. If the termination condition is not met, return to Step 3 and continue the iteration loop until the termination condition is met, or d > D.
[0096] Step 9: Verify whether the current optimization result meets the index requirements. If not, return to Step 1 and modify the fitness function and the parameters required for each optimization.
[0097] After 500 iterations of optimization using an improved genetic algorithm, the sparse optimized cell distribution diagram is obtained as follows: Figure 5 As shown, the number of array elements after sparsification is 108. The convergence performance curve is as follows. Figure 6 As shown in the figure, the horizontal axis represents the number of iterations, and the vertical axis represents the fitness function value. The figure shows that after 220 iterations, the fitness function value tends to stabilize. This indicates that the hybrid optimization algorithm has a relatively fast convergence speed, and the optimization results are relatively stable with small variance after multiple iterations. The optimized sparse array pattern is shown below. Figure 7 As shown in the figure, after sparse optimization using the improved genetic algorithm, the array's gain is 26.13 dB and the sidelobe level is -21.31 dB when the scanning direction is normal, both meeting the design requirements and exhibiting good radiation performance. Therefore, using the algorithm of this invention for beam synthesis optimization can achieve a sparse array that meets the design requirements.
[0098] In summary, this invention improves upon the genetic algorithm and combines it with the differential evolution algorithm. To avoid the algorithm getting trapped in local optima and improve its convergence speed, improvements are made to the three operators (selection, crossover, and mutation) of the genetic algorithm. To enhance the algorithm's global search capability, the improved genetic algorithm is combined with the differential evolution algorithm, thus refining the population update strategy. Simulation examples validate the algorithm, demonstrating good optimization results when applied to sparse array beam synthesis, meeting the array performance design requirements.
Claims
1. A sparse array beam synthesis method based on improved genetic algorithm, characterized in that, Includes the following steps: (1) Determine the fitness function based on the required array pattern specifications: where f is the fitness of individual i, is the target sidelobe level of the sparse matrix, is the maximum sidelobe level of the sparse matrix; (2) Create initial population P1(0), set optimization parameters, including maximum iteration number D, iteration number counter d, population size p, selection probability , initial crossover probability , minimum crossover probability , initial mutation probability , minimum mutation probability ; (3) Generate a population P2(d) of the same size as the current population using the differential evolution algorithm, and mix it with the current population; (4) Calculate the fitness of the mixed population and sort them by fitness, retaining the top half of the population P(d) with better fitness. (5) Apply the selection operator to the population and determine the selection probability. Perform a selection operation; (6) Apply the crossover operator to the population selected in step (5) and, according to the two-point crossover strategy, use an adaptive crossover probability. New offspring are produced by exchanging partial chromosomes between parent individuals; where an adaptive crossover probability is used. for: in, The initial crossover probability, To minimize the crossover probability, The average fitness value of the entire population. This is the minimum fitness value for the entire population; (7) Apply the mutation operator to the population after the crossover operation in step (6) with an adaptive mutation probability. To alter several gene loci in an individual; After the selection, crossover, and mutation operators are applied, a new generation population P(d+1) is obtained. (8) Calculate the fitness of the new generation population and determine whether the fitness of the best individual meets the termination condition. If the termination condition is met, the iteration ends and the optimization result is output. If the termination condition is not met, return to step (3) and continue the loop iteration until the termination condition is met or d > D.
2. The sparse array beam synthesis method based on an improved genetic algorithm according to claim 1, characterized in that, The specific method of step (5) is as follows: based on the selection probability Select N individuals with good fitness from the population, then randomly select n individuals from among them, and finally select the best individual from these n individuals as the parent for the next crossover operation.
3. The sparse array beam synthesis method based on an improved genetic algorithm according to claim 1, characterized in that, The adaptive mutation probability in step (7) for: in, The initial mutation probability, Let d be the minimum mutation probability, d be the current iteration number, and D be the maximum iteration number. The average fitness value of the entire population. This represents the minimum fitness value for the entire population.
Citation Information
Patent Citations
Array sparse optimization method based on adaptive genetic algorithm
CN111160556A
Antenna array sparse distribution optimization method, device and equipment and readable storage medium
CN114861446A
