Radar multi-target optimization design method and system considering parasitic effect constraint

Abstract

The application discloses a radar multi-target optimization design method and system considering parasitic effect constraints, relates to radar systems and radar optimization design technology, and the application constructs a proxy model of main effects and parasitic effects through Latin hypercube sampling, and implements adaptive sampling based on local precision; then, the NSGAII algorithm is improved, dimension difference is eliminated through target normalization, and constraint boundary focusing and elite reservation are executed in the later evolution stage, so that a morbid front is accurately captured; subsequently, Monte Carlo simulation is introduced for robustness verification, and hard constraints of solutions with high risk of exceeding the standard are automatically tightened and re-optimized; finally, a gradient fine-tuning mechanism with an attenuation index is adopted at the boundary, so that the step length is nonlinearly attenuated with the decrease of the boundary distance, and out-of-bound is avoided. The application shortens the radar design period from several weeks to several hours, ensures that the design point has a very high yield under the manufacturing tolerance, and solves the quantitative trade-off problem of radar performance and power consumption under strong constraints.

Description

Technical Field

[0001] This application relates to the fields of radar systems and radar optimization design technology, and in particular to a radar multi-target optimization design method and system that considers parasitic effect constraints. Background Technology

[0002] Radar design needs to maximize detection performance while satisfying parasitic constraints such as power consumption, heat dissipation, and cost. Traditional sequential design optimizes performance first and then calculates constraints, which can easily lead to wasted margins or iterative rework. Although multidisciplinary design optimization (MDO) has been introduced, most methods treat all parasitic effects as hard constraints and only output single-point solutions, failing to answer decision-making questions such as "how much will a 10% increase in power consumption improve the detection probability?"

[0003] Even if some parasitic effects are set as targets, the strong nonlinear coupling of radar parameters will still cause the Pareto front to exhibit a narrow, ill-conditioned concave shape. When dealing with this ill-conditioned front, the standard NSGA-II suffers from problems such as target dimension differences, insufficient boundary resolution, and the discarding of elite solutions, resulting in a front coverage of only 65% ​​after 200 generations. Furthermore, high-fidelity simulation is computationally expensive and requires surrogate models, but the accuracy of surrogate models near the Pareto front lacks specific guarantees.

[0004] If all parasitic effects are taken as objectives, the optimization problem will become too dimensional, further exacerbating the difficulty of convergence. Summary of the Invention

[0005] This application provides a radar multi-target optimization design method and system that considers parasitic effect constraints. It uses NSGA-II with an adapted ill-conditioned frontier to generate a complete Pareto frontier, and combines gradient fine-tuning of boundary range attenuation and Monte Carlo robust verification to obtain a radar design that strictly satisfies the constraints and is robust.

[0006] This application provides a radar multi-target optimization design method considering parasitic effect constraints, including: The main performance index of the radar system is determined as the first optimization objective. At least one key parasitic effect index is selected from multiple parasitic effects as the second optimization objective. The remaining parasitic effects are set as hard constraints of inequalities. Select design parameters that affect any optimization objective and constraint, construct a design space, and determine the value range of each parameter. The design parameters include continuous variables and discrete integer variables. Maximize the main effect and minimize the key parasitic effect index to establish a multi-objective optimization problem based on the first optimization objective and the second optimization objective, and satisfy all the inequality hard constraints; An improved multi-objective evolutionary algorithm is used to solve the multi-objective optimization problem. The objective values ​​of the main effects and key parasitic effects are normalized, and the constraint processing follows the feasibility-first principle. In the later stage of the improved multi-objective evolutionary algorithm, the search range is narrowed to the vicinity of the constraint boundary to improve the frontier resolution, and elite solutions near the boundary of each generation are retained. Integer constraints are maintained for discrete integer variables during the evolution process. The Pareto front is output. The final design point is selected from the Pareto front based on engineering requirements.

[0007] This application provides a radar multi-target optimization design system considering parasitic effect constraints, including a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the steps of the radar multi-target optimization design method considering parasitic effect constraints as described above.

[0008] This application embodiment constructs and verifies a surrogate model through a binary partitioning strategy, generates a complete Pareto front using NSGA-II adapted to ill-conditioned fronts, and obtains a robust radar design that strictly satisfies the constraints by combining gradient fine-tuning of boundary distance decay and Monte Carlo robust verification.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart illustrating the radar multi-target optimization design method considering parasitic effect constraints according to an embodiment of this application. Figure 2 This is a schematic diagram of the causal analysis of design parameters, main effects, and parasitic effects in an embodiment of this application. Figure 3 This is a schematic diagram of the proxy model construction and adaptive verification process in an embodiment of this application; Figure 4 This is a schematic diagram of the NSGA-II constraint dominance principle and Pareto front in an embodiment of this application; Figure 5 This is a schematic diagram of the Pareto front in an embodiment of this application; Figure 6This is a schematic diagram of the Monte Carlo robust verification box line in an embodiment of this application; Figure 7 This is a schematic diagram comparing the convergence of the standard NSGA-II and the adapted NSGA-II on the ill-conditioned frontier of the radar, as shown in the embodiments of this application. Detailed Implementation

[0011] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0012] This application provides a radar multi-target optimization design method considering parasitic effect constraints, such as... Figure 1 As shown, it includes the following steps: In step S1, the binary division of the target and constraints is optimized. Based on the tactical indicators of the radar system, the target system is established, the main performance indicators of the radar system are determined as the first optimization target, at least one key parasitic effect indicator is selected from multiple parasitic effects as the second optimization target, and the remaining parasitic effects are set as inequality hard constraints. Figure 2 This diagram illustrates a causal analysis of design parameters, main effects, and parasitic effects. In subsequent embodiments of this application, the detection probability is selected. As the primary performance indicator, the overall power consumption was selected. As a key parasitic effect indicator, this embodiment of the application uses it as an example; the selection methods for other indicators will not be elaborated. Other indicators affecting manufacturability and reliability, such as junction temperature... Production cost C is set as a hard constraint of inequality, for example... ≤125℃, C≤50,000.

[0013] In step S2, design parameters affecting any optimization objective and constraint in S1 are selected to form a design space and determine the value range of each parameter. These design parameters include continuous variables and discrete integer variables. In a specific example, the design variable set x = { ,Ae,τ, , }.in, Where Ae is the radar transmit power, Ae is the antenna aperture area, and τ is the pulse width. Noise figure For the number of T / R components, Ae, τ, and NF are continuous variables. For discrete integer variables, upper and lower limits are set for each parameter to form a multidimensional design space Ω.

[0014] An initial sample point set is generated within the parameter space using Latin hypercube sampling (LHS). The radar performance model is then used to calculate the main effect index and parasitic effect index corresponding to each sample point; specifically, the index values ​​of the corresponding main effect index and parasitic effect index are calculated. Specifically, 250 sample points can be generated within Ω using LHS sampling. For each sample point, radar link-level simulation software is used to calculate... Thermodynamic simulation software calculation The cost estimation model is called to calculate C.

[0015] In step S3, a dual optimization objective is established based on the main effect index and the specified key parasitic effect index. The remaining parasitic effect indices are used as hard constraints to construct a multi-objective optimization problem, satisfying all the aforementioned inequality hard constraints. This embodiment of the application aims to maximize the main effect. and minimize key parasitic effects Taking two optimization objectives as an example, and considering other parasitic effects as inequality constraints, other optimization examples will not be elaborated upon. The problem is formulated as: maximizing... Minimize (x) (x), satisfying (x)≤ C(x)≤ and other constraints.

[0016] In step S4, an improved non-dominated sorting genetic algorithm is used to solve the multi-objective optimization problem. During the solution process, the search range of the mutation operator in the non-dominated sorting genetic algorithm is limited to a preset neighborhood from the current Pareto front solution set to each hard constraint boundary. The improvements made to the NSGA-II algorithm in this application include: Constraint handling: The feasibility priority principle is adopted. That is, in the population competition, any feasible solution that satisfies the hard constraints has a higher priority than an infeasible solution; two feasible solutions are ranked according to Pareto dominance; two infeasible solutions are ranked according to the absolute value of the total number of constraints violated.

[0017] Objective normalization: linearly mapping the objective function value to the interval [0,1], for example, ... (x) (x) is mapped to the [0,1] interval to eliminate dimensional differences and ensure that the algorithm can correctly identify the shape of the leading edge.

[0018] In step S5, the Pareto front solution set containing the trade-off between the main effect index and the key parasitic effect index is output.

[0019] This application embodiment constructs and verifies a surrogate model through a binary partitioning strategy, generates a complete Pareto front using NSGA-II adapted to ill-conditioned fronts, and obtains a robust radar design that strictly satisfies the constraints by combining gradient fine-tuning of boundary distance decay and Monte Carlo robust verification.

[0020] In some embodiments, after constructing the design space, the method further includes building surrogate models corresponding to the main effect index and the parasitic effect index, respectively, for example, using Gaussian process regression (GPR) to fit them respectively. , , C, to construct a surrogate model to replace the physical model for massive evaluation. In some embodiments, the method further includes validating the local prediction accuracy of the surrogate model using a test set: Based on the approximate Pareto front solution set output by the current surrogate model, determine the region of interest in the parameter space, and calculate the local root mean square error of the surrogate model within the region of interest; for example, calculate the local RMSE within ±5% of the parameter variation in the parameter neighborhood of the solution set.

[0021] If the local root mean square error exceeds a preset threshold, adaptive additional sampling is performed within the region of interest, and the surrogate model is updated using the calculation results of the additional sample points until the local prediction accuracy within the region of interest meets the requirements. Figure 3 As shown, accuracy verification can be performed by randomly selecting 50 sample points as the test set and calculating the coefficient of determination. And root mean square error (RMSE). The acceptance criterion is set as: main effect. of ≥0.95 and RMSE≤0.02; parasitic effect waiting The local RMSE is ≥0.90 and the relative root mean square error (RRMSE) is ≤10%. If the local RMSE exceeds 1.5 times the global RMSE, 20 additional sampling points are added to that region, the high-fidelity model is called again, and the surrogate model is updated until the local accuracy meets the standard.

[0022] In some embodiments, the preset neighborhood is defined as ±5% to ±10% of the variation of each design parameter; and, in the later stages of evolution, for the boundary solutions in the current Pareto front solution set, individuals with a specified proportion of those solutions are retained as elite individuals and directly enter the next generation population.

[0023] This application's improvement to the NSGA-II algorithm also includes constraint boundary focusing: a boundary focusing mechanism is activated when the evolutionary process enters the last 30% of generations. The search range of the polynomial mutation operator is limited to ±10% of the distance from the current Pareto front solution set to each constraint boundary. For example, if the temperature of a solution is only 5℃ away from the upper limit of 125℃, the temperature fluctuation range of the new solution generated by mutation is limited to ±0.5℃. This greatly improves the resolution of the front boundary.

[0024] This application also establishes an elite retention strategy: after each generation of computation, the top 5% of solutions located at the boundary in the Pareto front are selected as elite solutions, such as the maximum... point, minimum The points are directly copied to the next generation to prevent the loss of these extremely high-quality solutions due to congestion calculation errors.

[0025] In some embodiments, the discrete parameter includes the number of T / R components; In solving the multi-objective optimization problem, the continuous variables generated after the crossover and mutation operation are processed by rounding to the nearest integer, and the radar performance model is re-applied for gain rollback verification at the rounded design points. In a specific example, for... After cross-mutation, a real value is generated. This value is first rounded to the nearest integer, and then the radar beamforming model is called to perform gain rollback verification. This corrects the array gain error caused by the change in the number of channels, ensuring... The accuracy of the prediction.

[0026] The design results output a Pareto front plot, such as... Figure 5 As shown in the figure, this graph visually illustrates the maximum detection probability achievable under different power consumption budgets. In some embodiments, such as Figure 6 As shown, it also includes a Monte Carlo robustness verification step: For candidate design points in the Pareto front solution set, Monte Carlo simulation is performed using radar manufacturing tolerance parameters; for each candidate solution, parameter tolerances are set, such as... ∼N(μ, ), Ae∼U(μ-2%μ,μ+2%μ), and perform 5000 random sampling simulations.

[0027] Statistically determine the probability that each parasitic effect index exceeds the hard constraint boundary under tolerance perturbation; if the junction temperature of a certain solution in the statistical results... If the probability of exceeding 125℃ is greater than 1%, then the solution is deemed insufficiently robust.

[0028] If the probability of any parasitic effect index exceeding the set tolerance exceeds the limit, the corresponding hard constraint boundary is reduced, and an improved non-dominated sorting genetic algorithm is used for iteration. If no satisfactory solution is found, the original hard constraints are automatically tightened, such as... Reduce by 5%, then return to step S5 for re-optimization.

[0029] In some embodiments, reducing the corresponding hard constraint boundary includes: reducing the junction temperature limit, or the cost budget by a set percentage.

[0030] For selected points that have passed verification and are located on the leading edge boundary, local fine-tuning is performed using the analytical gradient of the surrogate model. To prevent out-of-bounds errors due to surrogate model errors at the boundary, a boundary distance decay mechanism is introduced. In some embodiments, a gradient fine-tuning step is also included: Select a target design point located at the Pareto front boundary and calculate the numerical gradient of the surrogate model at any target design point; The fine-tuning step size is adjusted according to the boundary distance decay mechanism and the normalized distance of any target design point from each hard constraint boundary, wherein the fine-tuning step size decays non-linearly based on the decrease of the normalized distance. The radar design parameters are adjusted along the main performance indicators and within the constrained feasible region to achieve local optimization.

[0031] This application introduces a boundary distance decay mechanism during fine-tuning: the fine-tuning step size is dynamically decayed based on the normalized distance of the design point from each hard constraint boundary, and the parameters are adjusted along the main effect improvement direction and into the constraint feasible region, so as to avoid fine-tuning divergence caused by the decrease in accuracy of the surrogate model at the boundary.

[0032] The adaptive sampling strategy in this application specifically enhances accuracy in local regions near the Pareto front, ensuring high confidence in the optimization within the areas of greatest concern for decision-making. Compared to directly using a high-fidelity model for evolutionary optimization, this method can complete the Pareto front solution within hours on a typical workstation. The boundary distance decay mechanism effectively prevents fine-tuning divergence caused by insufficient boundary accuracy of the surrogate model.

[0033] Based on the method of this application, this application proposes a case study of the optimization design of detection probability, power consumption, and heat dissipation constraints for an airborne multi-function radar: In this example, the radar operates in search mode, and the RCS of a typical target needs to be maximized to 3m. 2 Detection probability of Swerling I ( = (Distance 80km). Power supply and heat dissipation limitations: Average power consumption of the entire system. ≤ 2000W, emitter junction temperature ≤125°C, ambient temperature 55°C, thermal resistance 0.03°C / W, cost C ≤ 150,000.

[0034] Design parameters: Transmit power ∈[500, 2000]W, antenna aperture area Ae∈[0.5,1.5]m 2 Pulse width τ∈[0.5,5]μs, noise figure NF∈[2,5] dB, number of T / R components ∈{100,..., 400}.

[0035] The surrogate model is constructed and adaptively validated. In this example, 200 points are sampled using a Latin hypercube. A second-order polynomial is used for fitting. Parasitic effects on linear / quadratic model fitting. Results for the initial 40-point test set: Model R 2 =0.97, RMSE=0.015; Model R 2 =0.99, RRMSE=1.5%; Cost model R 2 =0.96, RRMSE=3.2%. The local RMSEs of all three parameters near the Pareto front (e.g., design parameter points corresponding to a power consumption target of 1800-2000W) did not exceed 1.5 times the global RMSE, indicating that adaptive sampling was not triggered. The surrogate model accuracy is acceptable.

[0036] Comparative verification of binary partitioning strategies: Under the same constraints, three optimization strategies are compared. Strategy A (All Constraints): Treat power consumption, junction temperature, and cost as hard constraints, and only optimize... The result is a single-point solution ( =1350W, =0.91, =1870W), unable to provide power consumption-performance trade-off information.

[0037] Strategy B (All Objectives): Will , , All three variables, C and D, are used as optimization objectives, resulting in a 4-objective optimization. Using NSGA-III for the solution, the population still has not converged after 500 generations, the Pareto front fragments, and a large number of pseudo-optimal solutions are present.

[0038] Strategy C (the binary partitioning of this invention): ... As the second optimization objective C is used as a hard constraint. A complete Pareto front is generated within 200 generations.

[0039] Convergence comparison between the adapted NSGA-II and the standard NSGA-II: On the same surrogate model, both the standard NSGA-II and the NSGA-II adapted in this application for radar ill-conditioned frontier, with target normalization, constraint boundary focusing, and elite preservation strategies, were run. Figure 7As shown, the standard algorithm achieves only 65% ​​Pareto front coverage after 200 generations, with a significant lack of fronts in the 1700-1900W power consumption range. The adapted algorithm achieves 95% coverage within 150 generations, with complete and evenly distributed fronts. The constraint boundary focusing strategy effectively improves the front resolution near the boundary in the last 60 generations. The elitist retention strategy cumulatively protects approximately 12% of high-quality boundary solutions, preventing them from being discarded due to population diversity maintenance.

[0040] The solution is optimized using an adapted NSGA-II, with a population of 100 and 300 generations of evolution. After continuous-domain crossover mutation and rounding to the nearest integer, the Pareto front is obtained: point A ( =1200W, Ae=1.2m 2 τ=2μs, NF=2.5dB =250) → =0.92, =1850W, C=148000. Based on the engineering requirement of power consumption ≤2000W, point A is selected from the Pareto front as the final design point.

[0041] Robustness verification and gradient fine-tuning, applying tolerance to point A: ±5%, A_e±2%, τ±10%, NF±0.5dB. Results from 10,000 Monte Carlo simulations. The mean is 0.91, and the 5th percentile is 0.86; the mean power consumption is 1850W, and the 99th percentile is 1980W; the mean cost slightly exceeds the budget. Gradient fine-tuning uses a boundary distance decay mechanism, calculating the normalized distance from the current point to the cost constraint boundary, and then decaying the fine-tuning step size accordingly. Ae is reduced to 1.15m. 2 , With a power level of 1150W, the cost decreased to 147,000, and no divergence occurred during the fine-tuning process. Ultimately... It dropped to 0.905, but still meets the requirements.

[0042] The method in this application elevates key parasitic effects such as power consumption to optimization objectives, and the directly generated Pareto front answers the core question of "how much performance can be gained by increasing power consumption?" Comparative experiments show that treating all factors as constraints (strategy A) only yields single-point solutions, and treating all factors as objectives (strategy B) causes the four-objective optimization to fail to converge within 500 iterations. In contrast, the binary partitioning strategy (strategy C) in this application can obtain the complete front within 200 iterations.

[0043] This application also proposes a radar multi-target optimization design system considering parasitic effect constraints, including a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the steps of the radar multi-target optimization design method considering parasitic effect constraints as described above.

[0044] It should be noted that, in the embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0045] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0046] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0047] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims. All of these forms are within the protection scope of this application.

Claims

1. A radar multi-target optimization design method considering parasitic effect constraints, characterized in that, include: The main performance index of the radar system is determined as the first optimization objective. At least one key parasitic effect index is selected from multiple parasitic effects as the second optimization objective. The remaining parasitic effects are set as hard constraints of inequalities. Select design parameters that affect any optimization objective and constraint, construct a design space, and determine the value range of each parameter. The design parameters include continuous variables and discrete integer variables. Maximize the main effect performance index and minimize the key parasitic effect index to establish a multi-objective optimization problem covering the first optimization objective and the second optimization objective, and satisfy all the inequality hard constraints; An improved multi-objective evolutionary algorithm is used to solve the multi-objective optimization problem. The objective values ​​of the main effects and key parasitic effects are normalized. The constraint processing follows the principle of feasibility first. In the later stage of the improved multi-objective evolutionary algorithm, the search range is narrowed to the vicinity of the constraint boundary to improve the frontier resolution, and the elite solutions near the boundary of each generation are retained. Maintain integer constraints for discrete integer variables during evolution; output Pareto front; The final design point is selected from the Pareto front based on engineering requirements.

2. The radar multi-target optimization design method considering parasitic effect constraints as described in claim 1, characterized in that, After constructing the design space, the process also includes building surrogate models corresponding to the main effect index and the parasitic effect index, respectively, and verifying the accuracy of the surrogate models: Based on the approximate Pareto front solution set output by the current surrogate model, determine the region of interest in the parameter space; Calculate the local validation metrics of the proxy model within the region of interest, wherein the local validation metrics include at least the coefficient of determination R. 2 And root mean square error RMSE or relative root mean square error RRMSE; If the local verification index exceeds the preset threshold, adaptive additional sampling is performed within the region of interest, and the surrogate model is updated using the calculation results of the additional sample points until the local prediction accuracy within the region of interest meets the requirements.

3. The radar multi-target optimization design method considering parasitic effect constraints as described in claim 1, characterized in that, The preset neighborhood is defined as ±5% to ±10% of the variation in each design parameter; and... In the later stages of evolution, for the boundary solutions in the current Pareto front solution set, a specified proportion of individuals are retained as elite individuals and directly enter the next generation of the population.

4. The radar multi-target optimization design method considering parasitic effect constraints as described in claim 3, characterized in that, The discrete parameters include the number of T / R components; In solving the multi-objective optimization problem, the continuous variables generated after the crossover and mutation operation are processed by rounding to the nearest integer, and the radar performance model is called again to perform gain rollback verification on the rounded design point.

5. The radar multi-target optimization design method considering parasitic effect constraints as described in claim 1, characterized in that, The output of the Pareto front is followed by a Monte Carlo robustness verification step: Monte Carlo simulations are performed on candidate design points in the Pareto front solution set, incorporating radar manufacturing tolerance parameters. The probability of each parasitic effect index exceeding the hard constraint boundary under tolerance perturbation is statistically analyzed. If the probability of any parasitic effect index exceeding the set tolerance exceeds the limit, the corresponding hard constraint boundary is reduced, and an improved non-dominated sorting genetic algorithm is used for iteration.

6. The radar multi-target optimization design method considering parasitic effect constraints as described in claim 5, characterized in that, Reducing the corresponding hard constraint boundaries includes: lowering the junction temperature limit, or reducing the cost budget by a certain percentage.

7. The radar multi-target optimization design method considering parasitic effect constraints as described in claim 1, characterized in that, It also includes gradient fine-tuning steps: Select a target design point located at the Pareto front boundary and calculate the numerical gradient of the surrogate model at any target design point; The fine-tuning step size is adjusted according to the boundary distance decay mechanism and the normalized distance of any target design point from each hard constraint boundary, wherein the fine-tuning step size decays non-linearly based on the decrease of the normalized distance. The radar design parameters are adjusted along the main performance indicators and within the constrained feasible region to achieve local optimization.

8. A radar multi-target optimization design system considering parasitic effect constraints, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, which, when executed by the processor, implements the steps of the radar multi-target optimization design method considering parasitic effect constraints as described in any one of claims 1 to 7.

Images