Optimal Design of Thinned Array Using a Hybrid Genetic Algorithm

Abstract

In this paper, a hybrid genetic algorithm (GA) is proposed for thinning a two-dimensional planar array by combining the conventional GA with moving least squares (MLS). This enhances the convergence rate and the global search performance. The MLS method is used to estimate local interpolation functions from non-uniform sample data (the population and the value of the objective function in the GA), and to find new and better populations from the interpolated functions. By incorporating these improved populations into the next generation, the MLS-GA achieves improved search performance of the global optimum and a faster convergence rate compared to conventional GA alone. Moreover, a nonlinear chirp function is used for an efficient thinning design. To verify the proposed MLS-GA, it is applied to a test function and the results are compared to that of the GA. The algorithm is then applied to thin an array with a rectangular grid and circular boundary. The design objectives are to minimize the peak side-lobe level and gain loss while satisfying a given thinning coefficient and to compare the results with the GA.