Fast Globalized Gradient-Based Optimization of Multi-Band Antennas By Means Smart Jacobian Updates and Response Features

Abstract

EM-driven optimization has become one of the most fundamental tools of modern antenna design. Complexity of contemporary antenna structures does not allow the theoretical insight to go beyond a conceptual stage. Fulfilling stringent performance requirements calls for a precise adjustment of the antenna dimensions in multi-parameter spaces. Solving such tasks is expensive when using conventional algorithms. Furthermore, global optimization is needed in many cases, which increases the level of design process difficulty. As a matter of fact, utilization of state-of-the-art population-based metaheuristics is prohibitive when the antenna is evaluated by means of full-wave EM analysis. This paper proposes a novel technique for globalized design optimization of multi-band antennas. Our approach adopts the accelerated trust-region gradient search with Jacobian change monitoring as well as response feature technology. The latter permits reliable allocation of the antenna resonances even if the initial design is poor and traditional local optimization techniques fail to find a satisfactory design. The algorithm is demonstrated using a dual-band uniplanar dipole antenna. The numerical results indicate that optimization can be performed at the cost typical for local search routines while retaining global search capabilities.