7/28/06 - Updated my wiki. Changes: http://krillr.com/wiki/Special:Recentchanges
. Also, please use the Talk/Discussion?
pages for each article, and do not edit the articles themselves. Mediawiki has this capability builtin, which is wonderful. I think robowiki should migrate to it, its very nice and provides facilities to automagically import Usemod stuff.
I'm moving all of my stuff to my personal wiki, located [here].
This is the new page for AaronKrill. I am now known as Krillr. Woot.
I am currently developing a new bot called [[Krillr/Bots?/TheTardis|TheTardis?]], completely from scratch. Its going to use a new navigational technique called [[Krill/Concepts/MultiGrav?|MultiGrav?]], based on the AntiGrav? concept. Also, its going to use a defensive navigation technique called [[Krillr/Concepts/HumanShield?|HumanShield?]] (which utilizes some special features in [[Krill/Concepts/MultiGrav?|MultiGrav?]] ;-)) for melee. The initial variant is going to be quite large (upwards of 5K codesize in my estimation), but as I get reacquainted with java it'll get smaller.
I'm having problems with targetting for my bot. I really dont know anything about statistical targetting, virtual guns, or anything else other than linear, circular, and headon. If someone wants to spend some time with me on an IM service walking me through the basic principles, I'd much appreciate it. Just pop me an email @ firstname.lastname@example.org if you're interested in helping out.
- Bot News
- July 24th, 2006 - I've uploaded DoctorWho, a bot based on my new KrillBot? framework and implementing DookiSaber? and DookiSailer? by voidious. This is my learning bot as I reacquaint myself with robocoding, and as i learn how to do awesome targetting/movement (by dissecting Dookious). Thanks voidious for your help so far.
- July 25th, 2006 - Added a page for the future robot, TheTardis
- Comments (Please thread these appropriately)
- I don't always have time on IM (see ContactInfo) to help walk you through it, but if you catch me on AIM or Google Talk, I would try and help you. And I'd gladly help answer specific questions you have here on the wiki, as would others. I would check out the GuessFactorTargeting/Tutorial that Kawigi wrote, or GFTargetingBot by PEZ. It's a very common and very advanced form of targeting, and it's not really too complicated on the conceptual level, either. Good luck. -- Voidious
- Thanks for the tips. I'm changing the name of my bot to Tardis since Dalek is already taken ;-). I'm currently working on completely rewriting the base code, I was having navigation problems that I believe rooted from the fact the new code was a hacked up version of an old version of ChumbaMini?. I've decided to focus on movement for now, testing how long it takes for bots like ChumbaWumba and Dookious to get a good pattern lock. I am hoping to be done with this tonight, then I can move on to targetting. -- Krillr
- Some suggestions if you plan on writing a 2000+ bot:
- Write framework code first. For example, classes that store the data from scanned robot events, classes that do robocode trig (projecting a point, finding angle between two points, etc). Most of these CodeSnippets can be just copy-pasted from the wiki, but how you want to organize things into packages, classes, and methods is up to you.
- Write gun code second. Test it using non-firing bots, like the ones in the MovementChallenge. I'd advise NOT testing against MovementChallenge2K6 bots, since things that make you hit Shadow (for example) more often usually don't improve your aim against most other bots.
- Add some RandomMovement to complete your first version. Hooray! A working bot! Don't spend too much time on your RandomMovement because RandomMovement won't get you very high in the ratings. So by now you should have a robot that does all the basics: it keeps the radar locked on an enemy, it stores some data on the enemy, it moves and it shoots. It should be rated somewhere around 1700, depending on how good your gun and movement are. Duelist, DuelistMicro, and DuelistMini are all RandomMovement + GuessFactorTargeting bots, and are all between 1735 and 1760.
- Once you have all that done, THEN start working on AdaptiveMovement. Leave it until last because it's really, really hard to get it bug-free. Use RobocodeGL or the debugging graphics stuff on this page to help you get your AdaptiveMovement debugged. Bug-free adaptive movement + average GF gun = free ticket into The2000Club.
- I've already done number one quite successfully. I'll be posted code and information on it tomorrow probably, after I modify it to use pluggable prediction modules. The second one I planned already. Random movement I've already got covered, but I plan on replacing it with a form of antigrav. I'm using Robocode 1.1.1's builtin debugging drawing for debug graphics already. Thanks for the suggestions though, I appreciate it. Im studying the guess factor guns from many different bots, the only reason why I'm implementing changes based on Voidious' is because the coding method is near identical to my own. Stay tuned over the next week or two to see what I can get done :-p -- Krillr