Select your localized edition:

Close ×

More Ways to Connect

Discover one of our 28 local entrepreneurial communities »

Be the first to know as we launch in new countries and markets around the globe.

Interested in bringing MIT Technology Review to your local market?

MIT Technology ReviewMIT Technology Review - logo

 

Unsupported browser: Your browser does not meet modern web standards. See how it scores »

To become a professional antenna designer, you can follow one of two paths: you can enroll in college- and graduate-level courses on electromagnetism, immerse yourself in the empirical study of antenna shapes, and apprentice yourself to an established technician willing to impart the closely guarded secrets of the discipline.

Or you can do what Jason Lohn did: let evolution do the work.

Physicists know a lot about Maxwell’s equations and the other principles governing wireless communications. But antenna design is still pretty much a dark art, says Lohn, a computer scientist working at NASA Ames Research Center outside Mountain View, CA. “The field is so squirrelly. All your learning is through trial and error, the school of hard knocks.”

So why not automate trial and error? Antenna design, Lohn believes, is one of many engineering problems that could best be solved by evolutionary algorithms, an emerging class of software that produces lots of different designs, rejecting the less fit in order to select the most functional. The resulting designs often seem a little inhuman – inelegant and uncanny.

Evolutionary algorithms, also known as genetic algorithms or GAs, take their cue from biological evolution, which can turn a crawling reptile into a soaring bird without any kind of forward-looking blueprint. In sexual reproduction, the shuffling of each parent’s genes – combined with random genetic mutation – creates organisms with new characteristics, and the less fit organisms tend not to pass on their genes to succeeding generations. Evolutionary algorithms work much the same way, but inside a computer. When Lohn creates a new antenna, for example, he starts off with a population of randomly generated designs and grades their relative performance. Designs that come close to preset goals win the right ­to intermingle their properties with those of other successful candidates. Designs that disappoint go the way of the archaeopteryx: oblivion.

Breeding antennas takes time, of course. Most designs are downright awful, and it takes a large number of computing cycles to find decent performers. Still, when you’ve got a computer that can generate and test 1,000 generations an hour, interesting ideas do emerge*. Lohn, a PhD who hasn’t taken a course on electromagnetism since his undergraduate years, expects to have at least one of his team’s antenna designs go into space this year as part of NASA’s Space Technology 5 mission, which will test a trio of miniature satellites. His favorite computer-designed antenna: a corkscrew contraption small enough to fit in a wine glass, yet able to send a wide-beam radio wave from space to Earth. It resembles nothing any sane radio engineer would build on her own.

“Evolutionary algorithms are a great tool for exploring the dark corners of design space,” Lohn says. “You show [your designs] to people with 25 years’ experience in the industry and they say, ‘Wow, does that really work?’” The slightly spooky answer is that yes, they really do, as Lohn established after months of testing. “If we’re lucky, we could have as many as six antenna designs going into space” in 2005, Lohn says.

Not every problem will succumb to the evolutionary approach. But those that will share a common characteristic: they all sit beyond what mathematician John von Neumann dubbed the “complexity barrier,” the dividing line between problems that can be solved using traditional, reductionist methods and those that require a more intuitive, throw-it-up-and-see-what-sticks approach. Until recently, crossing this barrier was an expensive proposition. But today’s computers are fast enough to sift through millions of offbeat designs in hope of finding one that works. Couple that with modern designers’ growing skill in applying evolutionary algorithms, says David Goldberg, director of the Illinois Genetic Algorithms Laboratory at the University of Illinois at Urbana-Champaign, and you get what engineers lovingly call “scalability”: the ability to tackle both miniature and massive design challenges.

“Just as the steam engine created mechanical leverage to do larger tasks, genetic algorithms are starting to give individuals a kind of intellectual leverage that will reshape work,” Goldberg says. “By automating some of the heavy lifting of thought, we free ourselves to operate at a higher, more creative level.” Such freedom comes at a price, of course. It requires that engineers recognize the impossibility of peering into each and every “dark corner” and put their trust in yet another layer of mechanical assistance. But more and more of them are taking that leap.

Pages

0 comments about this story. Start the discussion »

Tagged: Biomedicine

Reprints and Permissions | Send feedback to the editor

From the Archives

Close

Introducing MIT Technology Review Insider.

Already a Magazine subscriber?

You're automatically an Insider. It's easy to activate or upgrade your account.

Activate Your Account

Become an Insider

It's the new way to subscribe. Get even more of the tech news, research, and discoveries you crave.

Sign Up

Learn More

Find out why MIT Technology Review Insider is for you and explore your options.

Show Me