A method, terminal and readable storage medium for obtaining multiple solutions of filter synthesis based on multi-objective assisted optimization

Abstract

The present invention relates to a method, a terminal and a readable storage medium for obtaining multiple solutions of filter synthesis based on multi-objective assisted optimization. The method includes: defining a topological transformation matrix of a filter according to a coupling matrix of a conventional topology of the filter and a coupling matrix of a target topology; defining a multi-objective function according to the coupling matrix of the target topology and the topological transformation matrix; solving a decision vector of the topological transformation matrix according to the multi-objective function to obtain a plurality of candidate topological transformation matrices, and outputting the coupling matrix of the target topology and the plurality of candidate topological transformation matrices. By retaining multiple candidate solutions during the optimization process, the present invention avoids only focusing on the optimal solution as in single-objective optimization, comprehensively considers the values of all objective functions and their positions on the solution boundary, explores more potential solutions, and can generate multiple solutions that meet the objectives simultaneously in the same running process, thereby improving the design flexibility.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method, a terminal, and a readable storage medium for obtaining multiple solutions of filter synthesis based on multi-objective assisted optimization. Background Art

[0002] In microwave and radio frequency systems, filters and multiplexers composed of multiple filters are crucial for communication systems. Such devices allow specific frequencies to pass through while blocking other frequencies. With the rapid development of the fifth-generation (5G) communication system and low Earth orbit satellite Internet, the demand for efficient use of spectrum resources has intensified, making the synthesis and design of high-performance filters and multiplexers increasingly challenging. Due to the complex design requirements of high-order filters and multiplexers, it is necessary to use solutions that are most suitable for physical layout and manufacturing processes.

[0003] In the prior art, the analysis and synthesis methods of conventional topologies of filters have been mature. Based on the classical filter coupling matrix rotation method, the corresponding implementation methods of conventional structures, such as folded topologies and arrow topologies, can be easily obtained. However, for an arbitrary topology, it is difficult to determine the corresponding rotation order. In addition, for high-order multiplexers, such as all-resonator multiplexers, only optimization-based synthesis methods can be used to obtain the coupling matrix. Traditionally, high-order filter synthesis can also be performed by numerical analysis methods. Although this method similar to exhaustive search can find all filter synthesis solutions, it takes an extremely long time and is difficult to use in actual industrial production.

[0004] Therefore, various optimization algorithms have been proposed in the prior art for filter synthesis, but these methods can only achieve single-solution filter synthesis. Even if the solution is successfully obtained through optimization methods, only one solution can be searched, ignoring a large number of other solutions, and this solution may not be the optimal solution for implementing the corresponding filter design. Therefore, the methods of applying existing optimization schemes to filter synthesis have the following problems. One is that they may fall into local optimal solutions. The second is that even if the global optimal solution is achieved, it may be difficult to use in actual industrial production. Eventually, the solutions obtained by existing optimization methods may not meet the needs of technicians, lacking design flexibility. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, a terminal, and a readable storage medium for obtaining multiple solutions of filter synthesis based on multi-objective assisted optimization, aiming to solve the problem that the existing filter synthesis methods only obtain single solutions and lack design flexibility.

[0006] The technical solutions adopted by the present invention to solve the technical problems are as follows:

[0007] The present invention provides a method for obtaining multiple solutions of filter synthesis based on multi-objective assisted optimization. The method for obtaining multiple solutions of filter synthesis based on multi-objective assisted optimization includes:

[0008] Define a topological transformation matrix of the filter according to the coupling matrix of the conventional topology of the filter and the coupling matrix of the target topology.

[0009] Define a multi-objective function according to the coupling matrix of the target topology and the topological transformation matrix.

[0010] Solve the decision vector of the topological transformation matrix according to the multi-objective function to obtain multiple candidate topological transformation matrices, and output the coupling matrix of the target topology and the multiple candidate topological transformation matrices.

[0011] Further, the step of solving the decision vector of the topological transformation matrix according to the multi-objective function to obtain multiple candidate topological transformation matrices specifically includes:

[0012] Divide the decision vector of the topological transformation matrix according to the multi-objective function, and divide each element of the decision vector into multiple sub-components.

[0013] Iteratively optimize each sub-component to obtain multiple decision vectors.

[0014] Obtain multiple candidate topological transformation matrices according to each decision vector and output the coupling matrix of the target topology.

[0015] Further, the step of dividing the decision vector of the topological transformation matrix according to the multi-objective assisted optimization function and dividing each element of the decision vector into multiple sub-components specifically includes:

[0016] Identify the relevance between each element according to the multi-objective function.

[0017] Divide the relevant elements into the same sub-component.

[0018] Further, the step of identifying the relevance between each element according to the multi-objective function specifically includes:

[0019] Calculate the collaborative difference, the first element difference and the second element difference between two elements according to the multi-objective function.

[0020] Judge whether the collaborative difference, the first element difference and the second element difference meet the association condition.

[0021] If the collaborative difference, the first element difference, and the second element difference satisfy the association condition, it is determined that the two elements are associated; if the collaborative difference, the first element difference, and the second element difference do not satisfy the association condition, it is determined that the two elements are not associated.

[0022] Further, the iterative optimization of each sub-component to obtain a plurality of the decision vectors specifically includes:

[0023] Generating an initial population according to the solution space of the decision vector;

[0024] Iteratively performing differential evolution on each sub-component in turn until the termination condition is satisfied;

[0025] Obtaining the Pareto optimal solution set of the initial population to obtain a plurality of the decision vectors.

[0026] Further, the iterative differential evolution of each sub-component in turn until the termination condition is satisfied, and each time differential evolution is performed, it includes:

[0027] For each sub-individual, randomly selecting a random optimal individual from the Pareto optimal solution set, and randomly selecting two different sub-individuals from the population;

[0028] Performing mutation according to the random optimal individual and the different sub-individuals to obtain a mutation variable;

[0029] Randomly selecting a crossover variable from the mutation variable and the variable of the corresponding sub-component;

[0030] Replacing the variable of the corresponding sub-component with the crossover variable to obtain a crossover individual of the sub-individual;

[0031] Selecting between the crossover individual and the sub-individual to update the sub-individual;

[0032] Updating the Pareto optimal solution set of the population according to each selected sub-individual.

[0033] Further, after the iterative optimization of each sub-component to obtain a plurality of the decision vectors, it further includes:

[0034] Performing local search optimization on each decision vector.

[0035] Further, the local search optimization is specifically optimized with the lowest cumulative value of each objective function in the multi-objective function as the goal.

[0036] In addition, to achieve the above object, the present invention further provides a terminal, which includes: a memory, a processor, and a multi-objective auxiliary optimization-based filter synthesis multi-solution acquisition program stored on the memory and executable on the processor. When the multi-objective auxiliary optimization-based filter synthesis multi-solution acquisition program is executed by the processor, it controls the terminal to implement the steps of the above-mentioned multi-objective-based filter synthesis multi-solution acquisition method.

[0037] In addition, to achieve the above object, the present invention further provides a readable storage medium storing a multi-objective auxiliary optimization-based filter synthesis multi-solution acquisition program. When the multi-objective auxiliary optimization-based filter synthesis multi-solution acquisition program is executed by a processor, it implements the steps of the above-mentioned multi-objective auxiliary optimization-based filter synthesis multi-solution acquisition method.

[0038] The present invention adopts the above technical solutions and has the following effects:

[0039] By retaining multiple candidate solutions during the optimization process of filter synthesis, the present invention avoids focusing only on the optimal solution as in single-objective optimization. It not only does not rely on a single objective value but also comprehensively considers the values of all objective functions and their positions on the solution boundary, thereby encouraging the algorithm to explore more potential solutions in the solution space of filter synthesis. It can generate multiple solutions that meet the objectives simultaneously in the same running process, improving the flexibility of filter design. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flowchart of the steps of a multi-objective auxiliary optimization-based filter synthesis multi-solution acquisition method in a preferred embodiment of the present invention;

[0041] Figure 2 is a schematic framework diagram of a multi-objective auxiliary optimization-based filter synthesis multi-solution acquisition method in a preferred embodiment of the present invention;

[0042] Figure 3 is a comparison schematic diagram of the topological structure and response of a filter in a preferred embodiment of the present invention;

[0043] Figure 4 is a schematic diagram of the topological structure transformation of a filter in a preferred embodiment of the present invention;

[0044] Figure 5 is a value-taking schematic diagram of a multi-solution problem in a preferred embodiment of the present invention;

[0045] Figure 6 is a schematic diagram of the topological structure of a multiplexer in a preferred embodiment of the present invention;

[0046] Figure 7 is a response schematic diagram of a multiplexer in a preferred embodiment of the present invention;

[0047] Figure 8 It is a schematic diagram of the convergence position of the multi-objective function in a preferred embodiment of the present invention;

[0048] Figure 9 It is a schematic diagram for comparing the topological responses before and after transformation in a preferred embodiment of the present invention.

[0049] Figure 10 Schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. Detailed implementation manners

[0050] To make the objectives, technical solutions and advantages of the present invention clearer and more definite, the present invention will be further described in detail below with reference to the accompanying drawings and by way of examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0051] Embodiment 1

[0052] Please refer to Figure 1 and Figure 2 , Embodiment 1 of the present application is a method for obtaining multiple solutions of filter synthesis based on multi-objective assisted optimization, which includes the steps of:

[0053] S1. Define a topological transformation matrix of the filter according to the coupling matrix of the conventional topology of the filter and the coupling matrix of the target topology.

[0054] In this embodiment, the filtering device includes a single filter and also a multiplexer composed of multiple filters. First, a single filter will be taken as an example for illustration.

[0055] Please refer to Figure 3 shown. In order to achieve the response curve of (b) in Figure 3 with the topological structure of (a) in Figure 3 , it is necessary to couple between the input and output of the filter and adjust the size of a resonator between the input and output. Among them, the S parameter represents the loss or gain parameter.

[0056] Specifically, in the actual design of the filter, in order to achieve the response curve of (b) in Figure 3 , in order to better express the corresponding relationship between structures, it can be represented by the topological graph of Figure 4 . Based on the filter synthesis method proposed in the prior art, it is easy to determine the mathematical model of (a) in Figure 4 . Among them, both S and L represent filter ports. However, the actual physical structure corresponding to this mathematical model is very difficult to achieve in the industrial production of filters. For example, the L port needs to feed power to resonator 1 and resonator 8 at the same time, and resonator 8 also needs to be connected to resonator 1, 2, and 7 at the same time. In the actual design process, the fewer the connection paths, the easier it is to achieve, such asFigure 4 as shown in (b) of Figure 4 to design a filter corresponding to the topology of (b) in Figure 3 i.e., to achieve the response curve of (b) in Figure 3 with the topological structure of (a) in Figure 4 one needs to find the corresponding coupling matrix. At this time, starting from the known Figure 4 conventional topology of (a) in

[0057] The coupling matrix of the conventional topology is denoted by which is obtained by the filter synthesis method of the prior art. The coupling matrix of the target topology is denoted by To convert the conventional topology to the target topology while keeping the response of the filter unchanged, it can be achieved by applying a transformation matrix denoted as This process follows the principle of orthogonal transformation, and the mathematical representation of this transformation is as follows:

[0058] ;

[0059] where the superscript denotes transpose.

[0060] Therefore, finding a suitable topology transformation matrix is crucial for filter design. In this embodiment, the topology transformation matrix is expressed as:

[0061] ;

[0062] For the unknown variables in the topology transformation matrix all are denoted by and these unknown variables form a decision vector where denotes the first element of the decision vector denotes the second element of the decision vector denotes the -th element of the decision vector denotes the -th element of the decision vector denotes the -th element of the decision vector denotes the number of unknowns in a row of the matrix.

[0063] Thus, in order to achieve an effective topology transformation, the problem is converted into finding a decision vector​​​​​ , to ensure that the coupling matrix of the transformed target topology meets the specified requirements. The goal is to find a suitable vector such that the coupling matrix of the resulting target topology aligns with the required filter topology.

[0064] S2. Define a multi-objective function according to the coupling matrix of the target topology and the topology transformation matrix.

[0065] Traditionally, a single-objective function is set and solved to obtain a single solution. In this embodiment, a multi-objective function is used for assistance to find multiple solutions. In filter and multiplexer synthesis, there may be multiple solutions that all meet the actual design criteria of the filter. Therefore, different parameters can be selected to achieve the same target response. Then there is a problem: which parameters are the easiest to implement in industrial production? Designers need to find the optimal solution among multiple solutions.

[0066] Since individual populations are regarded as independent entities, parallel computing and population partitioning can accelerate evolution. Specifically, in this embodiment, population partitioning uses a density-based clustering method to identify high-density regions in the population and divide the solution space into several clusters (sub-populations). Regions with higher density are divided into a sub-population, and regions with lower density are regarded as noise points and usually not assigned to any sub-population. After clustering, each sub-population is optimized independently. Noise points are allowed to be reclassified in multiple iterations to avoid falling into local optima.

[0067] Please refer to Figure 5 , where multiple solutions are similar to the peaks of the solution space, and the synthesis algorithm hopes to obtain multiple feasible solutions in one run. Then, the most suitable solution for processing is selected from these feasible solutions for the actual processing design of the filter.

[0068] Since the purpose of global search is to obtain as many solutions as possible, the objective function is set to be multi-objective to adapt to the multi-solution problem.

[0069] ;

[0070] Among them, represents the defined feasible solution domain, represents the vector composed of objective functions, represents the first objective function, represents the second objective function, represents the th objective function, represents the number of objective functions.

[0071] Converting a single-objective optimization problem into a multi-objective optimization problem lies in decomposing the objective function to form multiple mutually related optimization objectives. By converting it into multi-objective optimization, the mechanism of the multi-objective optimization algorithm is used to naturally maintain population diversity and avoid premature convergence to a single solution. This method not only enhances the exploration of the search space but also enables the retention of multiple different topological solutions in the solution, thus providing more design options.

[0072] Specifically, in this embodiment, the multi-objective function of a single filter is specifically as follows:

[0073] ;

[0074] Among them, represents the objective value of the multi-objective function, represents the index set of the coupling values that should be 0 in is the identity matrix, represents the th column of the matrix, represents the th row of the matrix, that is, the th column and th row element. The reconstruction process can be expressed as a system of non-linear equations, and solving gives the vector .

[0075] That is, taking a single filter as an example, it can be decomposed into two objective functions, which are respectively and .

[0076] To achieve the topological transformation of the filter, both of its two objective values must be 0, so that it can be completely equivalent in mathematical sense. Therefore, to achieve the solution of multiple solutions for filter synthesis, in this embodiment, the original single objective function is first converted into two independent objective functions by separating the objective function. For these two objective functions, realizes the topological transformation, and enables the response function to remain unchanged during the topological transformation. It can be considered that these two objectives influence each other but are not contradictory.

[0077] In this way, when decomposed according to the objective function, it becomes two independent optimization objectives. Each objective function corresponds to a different optimization path, and this decomposition may generate multiple solutions during the optimization process. Compared with single-objective optimization algorithms, which only correspond to one optimization path and thus converge to one optimal solution in the end, for multi-objective optimization, each objective guides a different optimization path. Due to the interaction of multiple paths, multiple solutions that simultaneously satisfy each objective function are finally generated.

[0078] In addition, the method of the present invention can not only be used for the use of a single filter, but also be applied to the synthesis of other filtering devices, such as the multi-solution search of a multiplexer. Specifically, please refer to Figure 6 , Figure 6 wherein P 1 , P 2 , P 3 , P 4 and P 5 all represent the ports of the multiplexer. When the multiplexer implements frequency division multiplexing, it uses multiple filtering channels to select different signal frequency bands and combines multiple signals through multiplexing technology. This is a common implementation method of the multiplexer. In particular, multiple filters are used to select and isolate different frequency bands of signals, and each signal occupies a specific frequency band. These filters ensure that each signal does not interfere with each other when combined.

[0079] As Figure 7 shown, Figure 7 is the response curve that the Figure 6 multiplexer needs to achieve. For an unconventional all-resonator multiplexer, only an optimization algorithm can be used for synthesis. Usually, the optimization model is constructed based on the S-parameter specification, which allows accurate determination of the coupling values that meet the required characteristics.

[0080] In this embodiment, for the multiplexer synthesized by multiple filters, its objective function is set according to the multiplexer response. As Figure 7 shown, in the frequency band , the return loss is less than the first return loss standard , in the frequency band , the return loss is less than the second return loss standard , in the frequency band , the return loss is less than the third return loss standard , in the frequency band , the return loss is less than the fourth return loss standard , and so on, where and Denote the frequency points collected within the passband or stopband from port 1 to port 2. Denote the frequency points collected within the stopband from port 1 to port 3. Denote the frequency points collected within the stopband from port 1 to port 4. Therefore, the multi-objective function can be set as follows:

[0081] ;

[0082] Wherein, Denote the objective value of the multi-objective function. Denote the frequency points collected within the passband corresponding to port P from port 1. Denote the frequency points collected within the stopband from port 1 to port P.

[0083] Please refer to Figure 8 After solving the multi-objective function, different from the single-objective function that finally converges to a single point, the optimal solution of the multi-objective function will finally form a line-like shape, that is, multiple solutions can be obtained.

[0084] S3. Solve the decision vector of the topological transformation matrix according to the multi-objective function to obtain multiple candidate topological transformation matrices, and output the coupling matrix of the target topology and the multiple candidate topological transformation matrices.

[0085] Specifically, in this embodiment, a multi-objective assisted co-evolution algorithm is used to evolve all variables, and this process will continue until a predefined stop criterion is reached. When solving, an initial value is first generated. In this embodiment, the initial value is generated by a randomization method, and a large number of candidate solutions are randomly generated within the specified variable range to form an initial population.

[0086] This method ensures the diversity of the population and helps to improve the global search ability. In actual operation, according to the variable upper and lower bounds of the problem, uniform distribution random sampling can be used to generate a certain number of population individuals. Each individual contains the initial values of all variables. The randomly generated population provides a good starting point for the subsequent optimization process and helps the algorithm avoid falling into local optimal solutions.

[0087] The co-evolution algorithm groups according to variable relationships, decomposing a complex large-scale problem into smaller sub-components. To achieve variable decomposition, differential grouping is an effective method. This method needs to examine the interdependencies and interactions between decision variables, and then effectively group the decision variables and iteratively perform mutation, crossover, and selection operations until the preset termination condition is met.

[0088] Specifically, cooperative collaboration is an effective method for solving large-scale optimization problems. This effectiveness is attributed to the decomposition of large components into multiple sub-components. Once the decomposition of cooperative collaboration is achieved, the entire problem can be solved through separate optimization. Due to its modular nature, the cooperative collaboration method is a framework suitable for large-scale optimization. However, the main difficulty in applying collaboration lies in decomposing decision variables into sub-components. In this embodiment, specifically, it is carried out by adopting the method of differential grouping.

[0089] The cumulative value defined as the objective function is:

[0090] ;

[0091] Among them, represents the cumulative value of the objective function, represents the th objective function.

[0092] The th element of the decision vector and the th element of the decision vector can be identified by comparing the difference calculated by the following equation:

[0093] ;

[0094] ;

[0095] ;

[0096] Among them, represents the collaborative difference, represents the first element difference, represents the second element difference, represents the th element of the decision vector's perturbation value, represents the th element of the decision vector's perturbation value. Given these differences, if the following conditions are met, the two variables are correlated:

[0097] ;

[0098] Among them is the parameter specifying the allowable error, that is, setting the error threshold.

[0099] To decompose all vectors into sub-components, the method first identifies the associations of all decision variables, which sequentially detects the associations between each variable and all other variables. Specifically, it first examines the interactions between the first decision variable and all other variables in the set. If an interaction is detected between the first variable and any other variable, the algorithm removes the interacting variables from the entire set of decision variables and then places that variable into a sub-component. This process continues until all variables that interact with the first variable are identified, forming the first sub-component. Then, the same process is repeated for the remaining decision variables. This loop iterates until all decision variables have been evaluated and assigned to appropriate sub-components. After partitioning, the elements of the decision variables are partitioned into several sub-components.

[0100] After partitioning the variables, differential evolution can be performed according to the grouping. The core of the differential evolution (DE) method is mutation, which involves adding a weighted difference vector between two individuals to a third individual. After mutation, these mutated individuals will undergo a discrete crossover and selection process together with the individuals of the previous generation. The mutation, crossover, and selection operations are iteratively executed continuously for iterative optimization until a pre-established termination condition is met.

[0101] Specifically, in this embodiment, during iteration, differential evolution is performed after grouping. For example, in the first iteration, the first sub-component is evolved, and in the second iteration, the second sub-component is evolved. Each time evolution is performed, the mutation step is first carried out:

[0102] ;

[0103] where and represent the variables of different individuals randomly selected from the contemporary population in the th sub-component, is the sub-component for which differential evolution is performed in this round of optimization, is the variable of the randomly optimal individual in the contemporary population in the th sub-component, is the scaling factor, is the rd sub-individual (i.e., the target individual) after mutation. It should be noted that since a multi-objective function is adopted in this embodiment, there will be multiple optimal solutions. Therefore, one optimal solution is selected from the multiple optimal solutions as the randomly optimal individual for mutation.

[0104] The mutated individual is crossed with the target individual to generate a new solution:

[0105] ;

[0106] Among them, is a random function, indicating to generate a random value between 0 and 1, is the crossover rate. is the th crossover variable of the th sub-component of the th crossover individual, is the variable of the th sub-individual in the th sub-component. Replace the th sub-individual 's th sub-component variable to obtain the crossover individual of the th sub-individual .

[0107] After that, selection is performed through the objective function :

[0108] ;

[0109] According to the value of the objective function, it is determined whether the crossover individual or the original individual will be left as the next generation.

[0110] Among them, 's condition is calculated based on the Pareto front method. If the condition is met, it indicates that performs better than on all objective functions. If it is not better than on one objective function, the condition is not met.

[0111] After each evolutionary iteration, the Pareto optimal solution (the Pareto optimal solution is a solution where no other solution performs better on all objectives) is derived based on the Pareto front, and these steps are repeated until the pre-established termination condition is met. At this time, the Pareto optimal solution is processed as a potential global optimal solution in the next step, and the Pareto optimal front (PF) is defined as the set containing the objective value vectors in the Pareto optimal set (PS).

[0112] After completing the global search and collecting multiple Pareto optimal solutions in the Pareto optimal set, local search is also performed for each optimal solution to further optimize the collected multiple solutions.

[0113] Specifically, the Levenberg-Marquardt algorithm is an iterative optimization method (hereinafter referred to as the LM algorithm) used to solve nonlinear least squares problems. The LM algorithm combines the advantages of the Gauss-Newton method (fast convergence near the minimum) and the robustness of the steepest descent method (the ability to handle non-convex objective functions and poor initial guesses). After global search, each global potential solution (the Pareto-optimal solution of the previous step) is used as the initial guess for the LM algorithm, and then the LM algorithm performs local search around this point to find a more accurate solution.

[0114] At this time, the cumulative value of the objective function is adopted Search with the lowest as the goal. In the filter, the cumulative value of the objective function at this time is:

[0115] ;

[0116] In other words, the objective at this time is modified to the sum of all objective functions. This change ensures that all specified constraints are satisfied. After that, all non-repeated solution sets are combined to represent multiple solutions that satisfy all objectives.

[0117] At this time, all the solved decision vectors are substituted into the topological transformation matrix to obtain multiple different topological transformation matrices .

[0118] ;

[0119] This topological transformation matrix can be transformed to , that is, the synthesis of the target topology is realized. As Figure 9 shown, the responses corresponding to the two topologies are exactly the same.

[0120] ;

[0121] Compared with the synthesis of the filter, the synthesis of the multiplexer only has differences in the objective function, and the other steps are exactly the same. Therefore, for the multiplexer, the multi-objective assisted co-evolutionary differential evolution algorithm can also identify multiple solutions. When a solution that meets the design specifications is found, the coupling matrix will accurately reflect the required characteristics, thus becoming a valuable tool for guiding the actual physical structure design of the multiplexer.

[0122] The co-evolutionary differential evolution algorithm decomposes complex large-scale problems into manageable sub-components, and each sub-component is optimized independently, effectively avoiding the problem of the curse of dimensionality and improving the optimization efficiency of the algorithm in high-dimensional problems. The integrated parallel computing method enables multiple solutions to be optimized simultaneously, greatly accelerating the overall convergence speed of the algorithm. Especially in the design of complex, high-order filters and multiplexers, it can significantly shorten the calculation time.

[0123] After obtaining each candidate topological transformation matrix, each candidate topological transformation matrix is output as a result.

[0124] By retaining multiple candidate solutions during the optimization process, the present invention avoids focusing only on the optimal solution as in single-objective optimization. It not only does not rely on a single objective value but also comprehensively considers the values of all objective functions and their positions on the solution boundary, thereby encouraging the algorithm to explore more potential solutions in the solution space. It can generate multiple solutions that meet the objectives simultaneously during the same running process, enhancing the design flexibility.

[0125] Embodiment 2

[0126] Please refer to Figure 10 , based on the above method, the present invention also provides a terminal, which includes a processor 10, a memory 20, and a display 30. However, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively.

[0127] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as the hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the terminal. Further, the memory 20 may also include both the internal storage unit and the external storage device of the terminal. The memory 20 is used to store application software installed on the terminal and various types of data, such as the program code installed on the terminal. The memory 20 may also be used to temporarily store data that has been output or will be output. In one embodiment, a multi-solution acquisition program 40 for filter synthesis based on multi-objective assisted optimization is stored on the memory 20, and this multi-solution acquisition program 40 for filter synthesis based on multi-objective assisted optimization can be executed by the processor 10, thereby implementing the terminal in the present application.

[0128] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chips, which are used to run the program code stored in the memory 20 or process data, such as executing the relevant programs of the multi-objective assisted optimization-based filter synthesis multi-solution acquisition method, etc.

[0129] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light Emitting Diode) toucher, etc. The display 30 is used to display the information in the terminal and to display a visual user interface.

[0130] In one embodiment, when the processor 10 executes the multi-objective assisted optimization-based filter synthesis multi-solution acquisition program 40 in the memory 20, the steps of the multi-objective assisted optimization-based filter synthesis multi-solution acquisition method described above are implemented.

[0131] Embodiment III

[0132] This embodiment provides a storage medium. The readable storage medium stores a multi-objective assisted optimization-based filter synthesis multi-solution acquisition program. When the multi-objective assisted optimization-based filter synthesis multi-solution acquisition program is executed by a processor, the steps of the multi-objective assisted optimization-based filter synthesis multi-solution acquisition method described above are implemented.

[0133] In summary, the present invention retains multiple candidate solutions during the evolution process, avoiding focusing only on the optimal solution as in single-objective optimization. It not only does not rely on a single objective value, but also comprehensively considers the values of all objective functions and their positions on the solution boundary, thereby encouraging the algorithm to explore more potential solutions in the solution space. It can generate multiple solutions that meet the objectives simultaneously in the same running process, improving the design flexibility.

[0134] It should be noted that in this article, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or terminal. Without more limitations, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, article or terminal including the element.

[0135] Of course, those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware (such as a processor, a controller, etc.) through a computer program. The program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the above method embodiments. The storage medium can be a memory, a magnetic disk, an optical disc, etc.

[0136] It should be understood that the application of the present invention is not limited to the above examples. For those of ordinary skill in the art, improvements or transformations can be made according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. A filter comprehensive multi-solution acquisition method based on multi-objective assisted optimization, characterized in that: The filter comprehensive multi-solution acquisition method based on multi-objective auxiliary optimization includes: A topology transformation matrix of the filter is defined according to a coupling matrix of a conventional topology of the filter and a coupling matrix of a target topology; Defining a multi-objective function according to the coupling matrix of the target topology and the topology transformation matrix; Solving the decision vector of the topology transformation matrix according to the multi-objective function to obtain multiple candidate topology transformation matrices, and outputting the coupling matrix of the target topology and the multiple candidate topology transformation matrices; The filter is a multiplexer, each objective function of the multiple objective functions corresponds to a channel one by one, and the return loss of the objective function in the frequency band of the corresponding channel is less than the set return loss of the corresponding frequency band; Solving the decision vector of the topology transformation matrix according to the multi-objective function to obtain multiple candidate topology transformation matrices specifically includes: Dividing the decision vector of the topological transformation matrix according to the multi-objective function, and dividing each element of the decision vector into a plurality of subcomponents; Iteratively optimizing each subcomponent to obtain a plurality of decision vectors; Obtain multiple candidate topology transformation matrices according to each of the decision vectors and output a coupling matrix of the target topology; The iterative optimization of each subcomponent to obtain the plurality of decision vectors specifically includes: Generating an initial population according to the solvable space of the decision vector; Iterate and perform differential evolution on each subcomponent in turn until the termination condition is met; Obtaining a Pareto optimal solution set of the initial population to obtain a plurality of decision vectors; The differential evolution is iterated on each subcomponent in turn until the termination condition is met, and each differential evolution includes: For each sub-individual, a random optimal individual is randomly selected from the Pareto optimal solution set, and two different sub-individuals are randomly selected from the population; Perform mutation according to the random optimal individual and the different sub-individuals to obtain a mutation variable; Randomly select a crossover variable from the mutation variable and the variable of the corresponding subcomponent; Replace the variables of the corresponding subcomponents with the crossed variables to obtain the crossed individuals of the sub-individuals; Select the crossover individuals and sub-individuals to update the sub-individuals; Update the Pareto optimal solution set of the population according to each selected sub-individual.

2. The filter comprehensive multi-solution acquisition method based on multi-objective auxiliary optimization according to claim 1 is characterized in that: The step of dividing the decision vector of the topological transformation matrix according to the multi-objective function and dividing each element of the decision vector into a plurality of subcomponents specifically includes: Identify the correlation between the elements according to the multi-objective function; Group related elements into the same subcomponent.

3. The filter comprehensive multi-solution acquisition method based on multi-objective auxiliary optimization according to claim 2 is characterized in that: The identifying the correlation between the elements according to the multi-objective function specifically includes: Calculate the synergy difference between the two elements, the first element difference, and the second element difference according to the multi-objective function; Determining whether the synergy difference, the first element difference, and the second element difference satisfy an association condition; If the collaborative difference, the first element difference and the second element difference meet the association condition, it is determined that the two elements are associated; if the collaborative difference, the first element difference and the second element difference do not meet the association condition, it is determined that the two elements are not associated.

4. The filter comprehensive multi-solution acquisition method based on multi-objective auxiliary optimization according to claim 1 is characterized in that: The iterative optimization of each subcomponent to obtain a plurality of decision vectors further includes: Perform local search optimization on each decision vector.

5. The filter comprehensive multi-solution acquisition method based on multi-objective auxiliary optimization according to claim 4 is characterized in that: The local search optimization is specifically optimized with the goal of minimizing the cumulative value of each objective function in the multi-objective function.

6. A terminal, characterized in that: The terminal includes: a memory, a processor, and a filter comprehensive multi-solution acquisition program based on multi-objective assisted optimization stored in the memory and executable on the processor. When the filter comprehensive multi-solution acquisition program based on multi-objective assisted optimization is executed by the processor, the terminal is controlled to implement the steps of the filter comprehensive multi-solution acquisition method based on multi-objective assisted optimization as described in any one of claims 1 to 5.

7. A readable storage medium, characterized in that: The readable storage medium stores a filter comprehensive multi-solution acquisition program based on multi-objective assisted optimization, and when the filter comprehensive multi-solution acquisition program based on multi-objective assisted optimization is executed by a processor, the steps of the filter comprehensive multi-solution acquisition method based on multi-objective assisted optimization as described in any one of claims 1-5 are implemented.

Images