Wire-antenna designs using genetic algorithms

There is a large class of electromagnetic radiators designated as wire antennas. As a rule, an inductive process is used to design these antennas. Either an integral equation is formulated or a simulator is used that gives the current distributions on the wires of the antenna, from which the electromagnetic properties of the antenna can then be determined. Once the antenna properties are known, the parameters are optimized, using guides such as intuition, experience, simplified equations, or empirical studies. However, using an electromagnetics simulator in conjunction with a genetic algorithm (GA), it is possible to design an antenna using a completely deductive approach: the desired electromagnetic properties of the antenna are specified, and the wire configuration that most closely produces these results is then synthesized by the algorithm. In this paper, we describe four antennas designed using GAs. The first is a monopole, loaded with a modified folded dipole that was designed to radiate uniform power over the hemisphere at a frequency of 1.6 GHz. We keep the same general shape of the loaded monopole, and use the algorithm to optimize its wire lengths. The second antenna consists of seven wires, the locations and lengths of which are determined by the GA alone, that radiates waves with right-hand-circular polarization at elevation angles above 10 degrees, also at 1.6 GHz. The last two antennas are modified Yagis. One is designed for a broad frequency band and very low sidelobes at a center frequency of 235 MHz. The other is designed for high gain at a single frequency of 432 MHz. It will be obvious that these antennas, with their unusual shapes, could not have been designed using an inductive approach. We have built and tested these antennas. The loaded monopole radiated uniform power over nearly the whole hemisphere, and the ''crooked-wire'' genetic antenna had nominal circular polarization at angles 10 degrees above the horizon. Although these antennas were optimized at a single frequency, they both turned out to be broadband. The broadband/low-sidelobe Yagi had side and backlobes that were greater than 25 dB down over most of the band. The high-gain genetic Yagi, having 17 elements and a boom length of 4.88 lambda, had a gain that was 0.4 dB higher than that of a conventional 18-element Yagi that had a boom length of 5.16 lambda. The measurements agreed very well with the computational results. We believe that this new process may revolutionize the design of wire antennas.