Robo Home | Changes | Preferences | AllPages

To be built by 12th March. One of my lecturers has given me a project to build a robocode robot with good targeting. :p I've chosen to make it into a team for bonus marks.

Any hints/suggestions on good Teams strategy? It's one thing I've seen very little discussion about.

-- Jamougha

Nice project! How about the best gun in existance? =) -- PEZ

I guess all the obvious stuff. Concentrate on one enemy at a time, don't fire if your bullet will intercept a teammate good melee movement, etc.

PEZ unfortunately GuessFactorTargeting does not yet qualify as an academically recognized machine learning techniwue. :-( Perhaps after I publish my Phd things will be different. ;-) I think I can get away with calling something like Tron's gun K Nearest Neighbour Search In Time-Series Pattern Matched Vector Spaces and referring him to a few JPL pages though, so I will probably go down that route.

Concentrating on one enemy at a time - good idea - I can even make it sound good in my paper (multi-agent distributed cooperation with stygmaturgy, hmmm ;-) ). -- Jamougha

You go for a team without having the melee base? this is great then you really nead some good strategies. How about:

-- rozu

Interesting ideas... I'll have to look into all this, I didn't even realize teams had leaders. :-) I like the idea of using something like 1v1 movement though, how about something like SandBoxLump?'s hyper-aggressive movement but with 2/3 bots teaming up on one opposing bot? That could be amusing... -- Jamougha

Oh, hell! This isn't due on the 12th it's due at 11am tomorrow... OK, rozu I'm stealing HawkOnFire's movement for the demo if you don't mind. :-) I'll write something completely different when I release the team.

OK, so now I have 8 hours to write a melee-capable PM gun and some data-sharing code. Anyone know of any other open-source team bots??? :-) -- Jamougha

wow,whatever class you are taking sounds really cool. i wish my classes involved having robots due by a certain time,or doing them at all. stupid highschool--andrew

Troodon and Nimrod come to mind. -- Kawigi

Thanks! Found Troodon and I will now have a look at Nimrod. This doesn't look as hard as I thought, I should have a team-capable clone of Tron's gun ready to roll in a couple of hours. :-) I love open-source... -- Jamougha

Maybe some of your wave-surfing code could be used for friendly fire avoidance? I've been thinking that a minimum-risk sort of movement with the focus of assigning risk to your own bullets and maybe other predicted bullets would be good for teams (rather than focusing on not being targeted like Coriantumr). -- Kawigi

Makes sense - I guess that if one team-member isn't being targeted then another is. Avoiding your own bullets is pretty easy at least. :-) Avoiding head-on at the same time would be easy, avoiding complex targeting would be harder than in 1v1 but probably doable. Looks like there's a lot that can be done in teams, but not much I can do in 6 hours, eh. I'll probably come back to it when I'm happier with the general rumble table. ;-) -- Jamougha

LOL, have fun Jam. If you're getting all academic on Robocode, could you do some research for me on discrete 1D physical random walks? I know there has to be a paper out there somewhere about flat random walks with momentum, but I can't seem to find it. :) -- nano

Robo Home | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited December 23, 2004 3:36 EST by Jonathan (diff)