A good bot to test genetic algorithms

[Simone Robutti]

Greetings,

For an entry level uni course on machine learning, i want to develop a GA to estimate the best parameters for a Robocode bot. It doesn't necessarily need to be effective, it could be a toy bot but complex enough to show off the GA i will develop on it. 

I'm here to ask what's the best approach? Are there open source bots suitable for this? Or should i develop it myself? What algorhithms should i use for each task?

Thanks

[Patrick Cupka]

Hmm... There are lots of open source bots you could use, including
some of the strongest bots out there. But I'd actually suggest not
using a super strong bot. The odds of improving upon the already
highly tuned variables in such bots is pretty slim - it would be more
pronounced if you start with a simpler or custom system and tune
things that have a large impact. Simpler bots will also run much
faster, which is crucial for running lots of GA generations.

My most focused GA experiment was the gun for RetroGirl:
http://robowiki.net/wiki/RetroGirl/Gun ... My goal was to create as
strong a gun as possible with pre-defined rules for how to aim (ie,
not learning anything). I did all my GA completely outside of
Robocode, actually - classifying data sets collected with Robocode
using a tool I wrote called WaveSim
(http://robowiki.net/wiki/WaveSim), then porting it back to a real
bot.

Maybe you could modify some of the sample bots? Like factor out
variables and start with settings that have them behave exactly as the
default sample bots, then use GA to try to evolve a bot that crushes
the sample bots. There really are a lot of options, so it's kind of
hard to give you a concrete suggestion on which way to go. But I'd be
curious to hear how it turns out. :-)

> --
>
> ---
> You received this message because you are subscribed to the Google Groups
> "robocode-developers" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to robocode-develo...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>



--
Patrick Cupka, pcu...@gmail.com
--

[Robin Harper]

This

Yehonatan Shichel, Eran Ziserman and Moshe Sipper, 2005,‘GP- Robocode: Using Genetic Programming to Evolve Robocode Players’, Proceedings of the 8th European Conference on Genetic Programming, vol. 3447, pp. 143–154. 

was an interesting papers using GP, in effect, to evolve parameters for a  robocode bot.

If you also want to evolve the program there have been one or two more since.