An interesting look at a learning algorithm

Something else for Armagetron? Goody!
Post Reply
User avatar
delinquent
Match Winner
Posts: 731
Joined: Sat Jul 07, 2012 3:07 am

An interesting look at a learning algorithm

Post by delinquent »

with a bit of fiddle, I think this could prove pretty useful in servers that use bots to balance teams.

http://www.youtube.com/watch?v=xOCurBYI ... r_embedded

Source code here.

User avatar
Z-Man
God & Project Admin
Posts: 11439
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Re: An interesting look at a learning algorithm

Post by Z-Man »

delinquent wrote:with a bit of fiddle, I think this could prove pretty useful in servers that use bots to balance teams.
Probably not. It rather looks to be only viable for reproducible games, where the same input sequence produces the same result every time. Multiplayer games are definitely not in that category.

Still, cool stuff.

User avatar
delinquent
Match Winner
Posts: 731
Joined: Sat Jul 07, 2012 3:07 am

Re: An interesting look at a learning algorithm

Post by delinquent »

Actually, I was thinking it could be modified to recognise walls and tails, and maybe even zones. Instead of simulating a lightcycle, simulate the player behind it.

in reality, only three keys need be programmed, the idea behind the bot being that it makes random turns in the hope that it can find an opposing player, whilst avoiding destruction. I know that the basis for that already exists, but it might prove useful.

I'll have a look at some point, see what I can do.

Post Reply