Unified particle swarm optimization with random ternary variables and its application to antenna array synthesis

Abstract

Antenna array synthesis sometimes involves both real and integer parameters as a mixed integer optimization problem. In this paper, a modified particle swam optimization (PSO) algorithm is proposed in order to deal with real and integer variables in a unified manner. Two major modifications are made compared to the classical PSO algorithm. First a unified vector having continuous values between 0 and 1 is defined, and at each iteration this vector (or part of it) is mapped to real variables and/or rounded to integer variables, which makes the updating process unified for any type of parameters. Second, random ternary variables are introduced to compensate quantization errors caused by the rounding-off operations, which could accelerate the speed of convergence and lead to improved topology exploration capability. In order to demonstrate the effectiveness of the proposed method, three previous examples about antenna array synthesis are revisited, and better results than those in the existing literatures are obtained in all these examples.

Keywords:

• unified particle swam optimization
• ternary variable
• antenna array synthesis