A Hybrid Optimized Algorithm Based on Ego and Taguchi's Method for Solving Expensive Evaluation Problems of Antenna Design

Abstract

In this paper, we propose a hybrid optimization approach that combines the Efficient Global Optimization (EGO) algorithm with Taguchi's method. This hybrid optimized algorithm is suited for problems with expensive cost functions. As a Bayesian analysis optimization algorithm, EGO algorithm begins with fitting the Kriging model with n sample points and finds the (n + 1)th point where the expected improvement is maximized to update the model. We employ Taguchi's method in EGO to obtain the (n + 1)th point in this paper. A numerical simulation demonstrates that our algorithm has advantage over the original EGO. Finally, we apply this hybrid optimized algorithm to optimize an ultra-wide band (UWB) transverse electromagnetic (TEM) horn antenna and a linear antenna array. Compared to Taguchi's method and the Integer Coded Differential Evolution Strategy, our algorithm converges to the global optimal value more efficiently.

A Modified Taguchi's Optimization Algorithm for Beamforming Applications

“…The algorithm is initialized by selecting a suitable fitness function fit and a proper orthogonal array OA (E, P , L, t), where E is the number of runs (or experiments), P is the number of parameters (or variables) to be optimized, L is the number of levels, and finally t is the strength. An orthogonal array with 3 levels and strength 2, i.e., OA (E, P , 3, 2), has been found to be efficient for most problems [21][22][23][24][25]. This type of OA is used below.…”

“…The present work introduces an effective ABF technique based on the recently announced Taguchi's Optimization (TagO) method [21][22][23][24][25]. To the best of the author's knowledge, the TagO method has never been applied before in antenna array beamforming problems.…”

Hybrid-Surrogate-Model-Based Efficient Global Optimization for High-Dimensional Antenna Design

“…The surrogate model employed in the EGO algorithm [1,[22][23][24][25][26][27]] is the Kriging model, which can be written as…”

“…However, it is difficult for the conventional EGO to avoid falling into local optima when the dimensions of optimization increase [24,25]. In order to overcome this difficulty, some improved EGO algorithms have been proposed for high-dimensional problems, e.g., the Taguchi'smethod-based EGO [26] and the GA-based EGO [27]. However, the convergence rates of these methods remain to be improved.…”

“…In comparison with the population-based optimization techniques, such as the genetic algorithm (GA) [2][3][4][5], particle swarm optimization (PSO) [6][7][8][9][10][11][12][13][14] and differential evolution strategy (DES) [15][16][17][18][19][20][21], the EGO requires fewer function evaluations to obtain the global optimum [22][23][24][25][26][27]. However, it is difficult for the conventional EGO to avoid falling into local optima when the dimensions of optimization increase [24,25].…”