Efficient Simulation-Based Global Antenna Optimization Using Characteristic Point Method and Nature-Inspired Metaheuristics

Antenna structures are designed nowadays to fulfill rigorous demands, including multiband operation, where the center frequencies need to be precisely allocated at the assumed targets while improving other features, such as impedance matching. Achieving this requires simultaneous optimization of antenna geometry parameters. When considering multimodal problems or if a reasonable initial design is not at hand, one needs to rely on global search. Yet, a reliable rendition of the system outputs necessitates the employment of electromagnetic (EM) analysis, associated with considerable CPU costs. Global optimization under such circumstances is extremely challenging. This especially applies to nature-inspired algorithms recognized for exceptionally low computational efficacy. Whereas surrogate-assisted approach is of limited use due to difficulties related to the construction of reliable behavioral antenna models. Here, we suggest an innovative methodology for efficient global optimization (EGO) of multiband antennas, where the surrogate is repeatedly built and refined using custom-defined response features. The infill criteria are based on minimizing surrogate-evaluated objective function, whereas the underlying optimization engine is the particle swarm optimization algorithm (PSO). Comprehensive benchmarking demonstrates superiority of the presented approach over surrogate-assisted methods handling antenna frequency responses, as well as direct nature-inspired optimization.