an evolution by any other name…

Engineers are using evolutionary algorithms to design new ion engines for spacecraft.

After you read enough papers and their accompanying popular science write-ups, you start to see a pattern of distracting additions to the subject matter either because the writer just discovered a concept and thinks it's a really, really cool idea that must be painstakingly detailed and emphasized again and again, or to make a big and hopefully interesting story out of something nifty but actually pretty mundane to non-experts. And it seems that the recent article in New Scientist about a new ion engine design is a front and center case of the latter, trying to attribute Charles Darwin's work on natural selection to a space flight engineer's attempt to design a more efficient and longer lasting ion thruster. Make no mistake, Cody Farnell's work is very nicely done and is certainly going to be of interest to other engineers, but that said, the emphasis on evolutionary algorithms and the connection to Darwin was that of the write-up's author rather than that of the paper itself.

Basically, an evolutionary algorithm is a way to speed up trial and error simulations and grading each attempt as to how successful it is. The better a certain model performs, the long it's kept around and the better it'll get assuming the standards by which its success is judged don't change. How does this apply to ion engines? Since they use a grid through which ionized xenon gas streams, generating thrust, eventually the grids wear out due to collisions with stray ions and the engine stops working. Evolutionary algorithms can play with every shape the grid can take until they reach an optimum design. That's just what Farnell had a computer program do and came up with an engine twice as durable, meaning that it could achieve twice the velocity because an ion engine accelerates by continuous thrust. The longer the impulse, the faster it will eventually go.

And that's the real story here: ion engines that have the potential to be faster, stronger and more durable, fit for missions deep into the outer solar system. The algorithms used in the process are a mathematical invention bearing a very passing and surface resemblance with the evolutionary science of biology. And with all my due respect for Darwin and his contributions to modern science, the evolutionary algorithm as we know it today is not a direct result of his work. Rather, it's a hybrid of different fields with similar goals, build on a long running set of ideas about probability and statistical analysis. Take note science writers. Not everything with the word evolution in its name is necessarily related to Darwin's work…