"Essentially, we start with something which might not be the best, then let the evolutionary process evolve to an optimal solution," said Rahmat-Samii, an expert in satellite and personal communications antennas and electromagnetic theory, who also was the elected 1995 President of IEEE Antennas and Propagation Society.

In keeping with the evolutionary model, Rahmat-Samii assigns a "gene" to each part of the antenna system he is designing. These genes are then collected to form a "chromosome" creating all the parts needed to be optimized. From these chromosomes, he creates individuals. Then he allows them to mate and reproduce.

"Just as nature has done, the algorithm selects the best species."

What is "profoundly appealing" about the process, he said, is that "sometimes you come up with new design concepts you never would have thought of before."

The process is complete when it has produced "something that meets your design objectives." This can take anywhere from five minutes to ten hours.

"Sometimes, something good evolves from bad genes," Rahmat-Samii said. "So from time to time, we inject a few flawed elements."

Unlike nature, however, Rahmat-Samii controls the switch which can stop the process. He determines when the program has produced a product that meets his definition of fitness.

"In nature, there is no definition of fitness," Rahmat-Samii said, "so you have to define satisfaction criteria."

Rahmat-Samii's research will appear in the book, Electromagnetic Optimization by Genetic Algorithms, which he is co-editing with Eric Michielssen and is to be published this year by John Wiley & Sons Inc.

Why apply a nonlinear method to the design of a mathematically precise device? According to Rahmat-Samii, this is the most efficient method of experimenting with alternative concepts and potential solutions. "Testing everything would take forever," he said, and random selection is unproductive.

Typically, genetic algorithm optimizations require three phases: Initiation, reproduction and generation replacement.

In the initiation phase, Rahmat-Samii creates a population of individuals, usually at random. This is called the "Current Generation." He then assigns a fitness value to each individual.

The parents "breed," thereby producing children that are the result of cross-over and mutation. He repeats the process until enough children have been generated to fill the next generation.

The procedure is altered slightly in some implementations. Selection is used to fill the new generation, and then crossover and mutation are applied through random pairings.

In its simplest scheme, known as a generational genetic algorithm, the new generation is the same size and completely replaces the previous generation. In more complicated implementations, the new generation can be of a different size than the previous generation or there can be overlap between the new generation and the older generation. Methods employing overlapping populations are called steady-state genetic algorithms. Additionally, both of these methods can be combined.

During the generational-replacement phase, the new generation replaces the current generation and fitness values are assigned to each of the new individuals. The termination criterion is then evaluated and, if it has not been met, the reproduction process is repeated.

Rahmat-Samii and his students have produced some compelling results using genetic algorithms. "These are designs that are not intuitive at all," he said. These startling new designs, in turn can be used to seed other genetic algorithms.

Genetic algorithms have been used to solve numerous real-world problems in engineering and a wide variety of electromagnetic and antenna problems. These include the design of broadband microwave absorbers; synthesis of antenna arrays and wire antennas of various forms; the design of frequency-selective surfaces; radar target cross section reduction for stealth applications; shaped reflector antennas for satellite communications; as well as dual band antennas for modern personal communications. (A dual band antenna based on this evolutionary methodology has been built and measured as shown in the photo below.)

Rahmat-Samii has been an invited plenary speaker at many international symposia and has presented tutorial talks on this subject worldwide.