Rocket Scientists Use Darwinian Software to Evolve Better Ion Engine Designs

I love it. Scientists DESIGN a program in order to DESIGN better rockets.

Um hello, natural selection is not evolution... it is selection. The program might randomly select values, through a process of random selection, but this is not evolution. A new, better, and different outcome would have to occur... say, the program would have to randomly select a better combination of values for a teleporting machine, but it cannot, because it cannot select what does not exist.

You are wrong. Evolution is the slow changing of one something into another through small mutations over time. The machine creates new values (mutations) then runs them through a test or simulation. The ones that are beneficial and that help the engine last longer then get selected and move to the next level. Your example would be like having a human give birth to an elephant, which cannot happen.

that photo looks just like the photo they use on wikipedia for the ion thruster article.

lol at the visual given from Deenreka's comment

they used similiar processes involved in evolution to make a solid improvement...you are all nitpicking

It must be good because it emits a cool blue flare that makes it look stylish and advanced.

Ooo, a practical use of natural selection. It was always a great theory, but now it is being demonstrated in a practical manner. What more do we need before evolution is considered an absolute fact?

And hopefully this will accelerate the space program. :)

This is pretty cool.... but soooo much more money needs to be diverted into launching a 20 megawatt nuke reactor into space to power a VASIMR plasma/ion drive. Mars in 2-5 weeks during the most favorable launch windows, and still possible out of the window. Picture it, hell even I could do mars in a month. NASA's current design for ion thrusters is, in my opinion, completely illogical. Anyway, interesting use for natural selection, may even be the first stepping stones for computers that can design better computers, starting the exponential growth curve that leads towards singularity.

Why dont they use evolutionary algorithims more often?

Why are we calling this natural selection? This is trial and error. Natural selection is about the ability to reproduce, whereas this is just tweaking to see what works a little better. Or maybe I misunderstood and this description was just a euphemism: "A high voltage between a pair of gridded electrodes accelerates positively charged ions toward the negative electrode. The ions that pass through the field generate thrust."

and I thought I was cool because I was doing FEA analysis on industrial pieces...

Shame on me !!! hihihhhih

@paulcrosoft:

you are both right and wrong, yes this is a type of trial and error. But...

Natural selection behaves in a manner similar to trial and error.

Here's why:

over the course of evolution, some new mutations emerge, if they work, they get passed on and succeed. if they dont work, then the gene dissapears as the animals die off.

in a similar manner, trial and error works by trying things to see if they work. if it works, you keep using it, if it fails, don't use it.

hopefully i've explained this adequately.

Trial and error = 1 generation, selecting the best of that generation.

Natural selection = making a change to one or more variables, selecting the best of that generation and then changing OTHER variables in the second generation or round of testing.

In time, the units or objects used in testing look very little like the originals as they have been modified incrememntally over time.

If i change grid pattern AND grid distances, x is the result.

If I combine results from previous step AND higher ion density, Y is the result

If I combine y with variable 1, Z is the result

This is natural selection at its finest folks.....NOT Trial and error or some other less-intelligent design model....

@logical_atheist: Natural Selection is a testable and verifiable theory, proven to be a mechanism that selects from traits that naturally vary within species. The supposed Evolution over long periods of time from Natural Selection is a speculated, untestable theory that has no real-world application. All we've ever seen is animals LOSING traits, as in the example of animals living in caves. Evolution can only be proven if an animal can somehow GAIN a new trait. I'm sure Natural selection creates new species, given enough time of genetic isolation. It cannot, however, create a new Genus, and certainly not a new Family.

There's no way Natural Selection could have made the gecko's foot. Research the physics of the hairs, and the mechanics of its joints, and you'll see what I mean. If you think it's possible, you'd better explain it in detail.

Most of that was a setup for the euphemism joke, but seriously this is no different than what my toddler does. You change a few things and try again. It isn't natural selection, because there is no selection naturally made by the results toward dominance (in nature, and New Jersey, dominance is the ability to reproduce). Instead you have a creator stepping in and saying what values should and should not be encouraged. Nature hasn't selected what makes something dominant in this experiment. It is LIKE natural selection in that they both use the same repeat trial and error process, but if this is natural selection then Cody Farnwell is God, which makes Darwin cry.

Shoot me down if I'm missing something, but this is all semantics. Let's not lose sight of how awesome this is.

it is arguing semantics...i believe the article intended to mean they are useing techniqes similair to natural selection...onihikage needs to stick to facts, not beliefs on a science website

Just curious as to how you generate a random number when a computer needs a specific set of values to work with and a specific number of variables. What's random?