CN-121980963-A - BPSO-MLP algorithm-based multi-objective optimization method for pixel antenna

Abstract

The invention relates to a pixel antenna multi-target optimization method based on a BPSO-MLP algorithm, which is applied to the technical field of wireless communication antennas. The method comprises the steps of carrying out global search in a high-dimensional discrete coding space of a pixel antenna through an improved binary particle swarm optimization algorithm, introducing dynamic temperature parameters to adjust the shape of a Sigmoid mapping function, adopting a probability-driven variation mechanism to effectively balance exploration and development capabilities, establishing a performance prediction model through a designed multi-layer perceptron neural network to realize rapid and accurate prediction of key performance indexes such as bandwidth, gain, radiation pattern and the like of a candidate antenna structure, submitting a high-potential solution screened by prediction to full-wave electromagnetic simulation through an intelligent optimization mechanism, feeding simulation results back to the optimization algorithm, and dynamically updating the prediction model. The invention effectively solves the problems of high-dimensional discrete search, high simulation cost and multi-performance index coordination in multi-target optimization of the pixel antenna, and provides an efficient solution for the design of the reconfigurable antenna of a new generation communication system.

Inventors

• ZHANG NAIBAI
• YU KAILE
• REN WEIZHENG
• HUANG JIANMING
• CUI YANSONG
• ZHANG YIRAN
• WANG XI
• REN ZHIXING

Assignees

• 中国电子科技集团公司第五十四研究所
• 北京邮电大学
• 芯界创微(苏州)科技有限公司

Dates

Publication Date

20260505

Application Date

20260330

Claims (4)

1. 1. The pixel antenna multi-target optimization method based on the BPSO-MLP algorithm is characterized by comprising the following steps of: Step 1, constructing a BPSO-MLP collaborative optimization framework, wherein the framework comprises a global searcher BPSO and a prescreener MLP; Modeling a pixel antenna to be optimized as an MxN digital coding super surface, wherein each pixel unit corresponds to one radio frequency switch, and coding a pixel antenna structure by using a binary character string with the length of L, wherein '0' or '1' of each bit in the character string corresponds to an 'open' or 'close' state of each radio frequency switch, wherein L=MxN, M and N are the row number and the column number of the pixel unit; Step 3, optimizing the pixel antennas in a binary coding space by using a binary particle swarm optimization algorithm by using a global searcher BPSO to obtain a plurality of candidate pixel antennas, and predicting the performance of the candidate pixel antennas after BPSO optimization by using a multi-layer perceptron neural network by using MLP to obtain predicted values of bandwidth, gain and main lobe position of a radiation pattern; And step 4, judging whether the predicted value meets the performance requirement, if so, performing full-wave electromagnetic simulation evaluation, calculating an adaptability function, selecting an optimal candidate pixel antenna, returning to the step 3, and if not, directly returning to the step 3 until the set iteration times or the convergence of the adaptability function are reached.
2. 2. The multi-objective optimization method of a pixel antenna based on a BPSO-MLP algorithm according to claim 1, wherein the multi-layer perceptron neural network adopted by the prescreener MLP comprises an input layer, two hidden layers and an output layer; The number of neurons of the input layer is L, the number of neurons of the first hidden layer is 2L, the number of neurons of the second hidden layer is L, the number of neurons of the output layer is 3, the predicted values respectively correspond to the bandwidth, the gain and the main lobe position of the radiation pattern, and the hidden layer uses a leakage ReLU activation function and adopts an Adam self-adaptive moment estimation algorithm as an optimizer for network training.
3. 3. The multi-objective optimization method for the pixel antenna based on the BPSO-MLP algorithm as claimed in claim 1, wherein the step 3 is characterized in that the pixel antenna is optimized in a binary coding space by using a binary particle swarm optimization algorithm by using a global searcher BPSO, and the specific process is as follows: Step 301, initializing binary particle swarm optimization algorithm parameters, including an iteration number upper limit, a particle number, a learning factor and a value range of inertia weight, wherein the position of a particle is a coding result of a pixel antenna structure, and the speed is a moving direction and a step length of the particle in a search space; Step 302, calculating inertial weights Learning factor And ; ; ; In the formula, Respectively the maximum value and the minimum value of the inertia weight, As an upper limit on the number of iterations, The current iteration number; And Respectively the initial values of the learning factors; And The final value set by the learning factor when the iteration number reaches the upper limit of the set iteration number; step 303, utilizing inertial weights Learning factor And Updating the speed of a binary particle swarm optimization algorithm; ; In the formula, In order to update the speed of the vehicle, For the current speed of the vehicle, And Is a random number within the interval 0, 1, For the best fitness position found from the search start to the current iteration itself, The optimal adaptability position is found for all particles in the whole particle swarm, The current position coordinate of the ith particle in the search space is the current iteration time t; step 304, performing location update: Dynamic temperature parameters The attenuation formula of (2) is: ; In the formula, In order to control the parameters for the decay rate, Setting a temperature parameter value when the cycle number reaches the set maximum cycle number; Probability of location update The formula is: ; Calculating the probability of taking '1' of each bit by combining with Sigmoid function of dynamic temperature parameter, and determining new binary position of particle by random sampling : ; Where rand () represents a random number between 0 and 1; Then introducing adaptive variation probability based on particle velocity : ; In the formula, Is a variation probability parameter; According to adaptive mutation probability The binary position of the particles is adjusted.
4. 4. The multi-objective optimization method for a pixel antenna based on a BPSO-MLP algorithm according to claim 1, wherein the fitness function in step 4 is: ; Wherein, the For the maximum gain to be achieved, To meet the number of frequency points corresponding to the maximum continuous bandwidth with return loss S 11 < -10dB, For the angle index corresponding to the maximum gain, Indicating the direction interval.

Description

BPSO-MLP algorithm-based multi-objective optimization method for pixel antenna Technical Field The invention relates to a pixel antenna multi-target optimization method based on binary particle swarm optimization and a multi-layer perceptron (BPSO-MLP) algorithm, which is mainly applied to the technical field of wireless communication antennas and belongs to the field of radio frequency front-end devices. Background With the rapid development of fifth generation mobile communication, low-orbit satellite internet and integrated sensing and communication systems, unprecedented demands are made on the antenna for multiple functions, reconfigurability, broadband and high efficiency. Although the traditional microstrip antenna can realize beam scanning, a complicated feed network, a phase shifter and a power amplifier are needed, so that the system has large volume, high power consumption and high cost. In addition, the response speed of the mechanical steering antenna is low, and the requirement of modern communication on real-time beam switching is difficult to meet. As an emerging reconfigurable antenna technology, a pixel antenna can achieve flexible reconfiguration of frequency, pattern and polarization by controlling the on-off state of the surface radiating elements. The method has the core advantage that the beamforming can be realized by controlling the surface current distribution only through digital coding without a complex phase shifter network. However, the optimization design of the pixel antenna faces serious challenges: There is a highly non-linear mapping between the performance of the pixel antenna and the switching state of its surface elements. For a pixel antenna with N switching elements, the possible switching state combinations amounts to 2 N. When N is large, the traditional full-permutation simulation method is not feasible in calculation, and the problem of 'combined explosion' exists. For example, S.Koziel, J.tan et al, at IEEE Transactions on ANTENNAS AND, published on the possibility of using optimization algorithms such as genetic algorithms, simulated annealing algorithms and conventional particle swarm algorithms to handle such high-dimensional discrete optimization problems, but have slow convergence rates, are prone to local optima, and require a large number of time-consuming electromagnetic simulations, severely limiting design efficiency. Most of the existing pixel antenna designs only support frequency reconstruction or one-dimensional directional diagram scanning, and the collaborative optimization of multi-target performances such as broadband, two-dimensional large-angle scanning, high gain and the like is difficult to realize at the same time. For example, although phased arrays based on certain pattern reconfigurable antennas can realize a scanning range of + -60 DEG, the phased arrays are limited to one-dimensional scanning, and certain two-dimensional scanning arrays have the problems of low gain and narrow bandwidth. Disclosure of Invention The invention aims to overcome the defects of low calculation efficiency, easy sinking into local optimum, insufficient multi-target coordination capacity and the like in the existing pixel antenna optimization technology, and provides a pixel antenna multi-target optimization method based on a BPSO-MLP algorithm. The method aims to remarkably reduce the times of full-wave electromagnetic simulation, greatly improve the optimization efficiency on the premise of ensuring the optimization performance, and realize synchronous optimization of bandwidth, gain and radiation pattern. The invention adopts the following technical scheme: A pixel antenna multi-target optimization method based on a BPSO-MLP algorithm comprises the following steps: Step 1, constructing a BPSO-MLP collaborative optimization framework, wherein the framework comprises a global searcher BPSO and a prescreener MLP; Modeling a pixel antenna to be optimized as an MxN digital coding super surface, wherein each pixel unit corresponds to one radio frequency switch, and coding a pixel antenna structure by using a binary character string with the length of L, wherein '0' or '1' of each bit in the character string corresponds to an 'open' or 'close' state of each radio frequency switch, wherein L=MxN, M and N are the row number and the column number of the pixel unit; Step 3, optimizing the pixel antennas in a binary coding space by using a binary particle swarm optimization algorithm by using a global searcher BPSO to obtain a plurality of candidate pixel antennas, and predicting the performance of the candidate pixel antennas after BPSO optimization by using a multi-layer perceptron neural network by using MLP to obtain predicted values of bandwidth, gain and main lobe position of a radiation pattern; And step 4, judging whether the predicted value meets the performance requirement, if so, performing full-wave electromagnetic simulation evaluation, calculating an adap