[Home]AgentSmith

Robo Home | Changes | Preferences | AllPages

AgentSmith

by Wolfman

What's special about it?

This bot has a pluggable learning interface. Just slap in a new learning method and AgentSmith will use it without any changes. Currently AgentSmith uses GeneticAlgorithms to learn. It will also be my first bot to support WaveSurfing and also im trying to implement a new targeting technique which I call SmartFactorTargeting.

Great, I want to try it. Where can I download it?

Nowhere yet, its still in development.

How competitive is it?

Dont know yet, still in development.

How does it move?

It will use WaveSurfing when finished. Probably my own variant.

How does it fire?

It will use SmartFactorTargeting.

How does it dodge bullets?

See WaveSurfing.

How does the melee strategy differ from one-on-one strategy?

It will be a 1v1 bot only to begin with.

How does it select a target to attack/avoid in melee?

N/A

What does it save between rounds and matches?

That depends upon the learning method that is currently plugged in. With Genetic algorithms it saves the current population of chromosomes.

Where did you get the name?

From the movie Matrix - AgentSmith is a bot that learns, but can only learn within the constraints of his surroundings.

Can I use your code?

Not sure yet, I will see how good or bad it does in the rumble before I decide.

What's next for your robot?

Finishing it for release.

Does it have any WhiteWhales?

Not yet, but there are bound to be a few ...

What other robot(s) is it based on?

None.


Page History

09/06/04 - Small changes, added SmartFactorTargeting.
27/10/04 - Added some general information on the bot.
26/10/04 - Moved all the old AgentSmith page to /Old.

Comments, questions, feedback:


If you're using evolutionary algorithms to find new guessfactors as a way of foiling surfers, then right on. Have you had much success yet? I found these things learn too slowly. -- Kuuran

No, im not going to use genetic algorithms to do that. Im planning on using them to tweak all the parameters that get hard coded in my bot during development. Then, when I release the bot I can use the best solutions for those parameters. Then for the release version I can plug in a faster learning module, perhaps based on reenforcement learning into the bot that should help with things like learning against specific bots. Thats the plan at least, still got a long way to go before I finish it! --Wolfman

Oh, alright. That's really cool, too, actually. If you search the wiki there was someone (I honestly can't recall who did this) who used evolutionary algorithms to write his bots from the ground up some time back. He posted lots of information and code either here or on one of the forums, which you might be interested in for some ideas/perspective. Good luck. :-) -- Kuuran

I think that was Geep, but PEZ has done something like that to tweak movement and stuff, possibly much like you are. -- Kawigi

Are you sure someone else didn't do it as well? I swear I recall someone posting some open source framework stuff. I may be mistaken. -- Kuuran

Ah, yes, searching around, there's Jacob, who did the paper at MIT about GeneticAlgorithms? in Robocode. -- Kawigi


Robo Home | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited June 11, 2005 7:52 EST by Kawigi (diff)
Search: