Talk:GITS/Targeting/Source

From Robowiki
Jump to navigation Jump to search

Wiki Targeting Implementation

Ok, I found the Wiki Targeting method, which is kinda what I want to do. How the devil to I implement it? HACKhalo2 21:05, 18 May 2009 (UTC)

There are several ideas there. Do you mean the dynamic segmentation? The data saving / pre-loading? If it's the saving / pre-loading, I use WT concepts in the gun data saving in Dookious, so you could take a look there. (Check dgun/MainGunHighBuffer.java for main routine, utils/StatBuffer.java (superclass) has some of the helper methods). As for the dynamic segmentation stuff, well, not a lot of people have tried that. Toad is the only bot I know of that has, and I do believe it is open source, but I haven't looked at it myself. You'd be in fairly uncharted waters. --Voidious 14:37, 19 May 2009 (UTC)

Thank you Voidious. I've been looking at Dookious because it implements the Dynamic Segmentation/Wiki Targeting. To be truthful, it confuses the devil out of me. I did successfully create a class for the GhostGun though, so it's outside the main robot (which will keep the code tidy :) ). Anyways, I'll look at the files you said and get some ideas. With reminds me, I need to update my Bot's page to the Bot Template. --HACKhalo2 16:45, 19 May 2009 (UTC)

To be clear, Dookious doesn't do any dynamic segmentation. It does use the idea of Super Nodes in its method of data saving. This lets it save relevant data for 80-90% of the enemy's movements in a very small file, usually under 1kb. What is confusing you exactly? How to dynamically resegment your stats? Segmentation in general? Dooki's code? If you haven't done much before in terms of segmentation, GuessFactor Targeting, Visit Count Stats, etc, I'd suggest starting with the GuessFactor Targeting Tutorial or GFTargetingBot. Dynamic segmentation is a really difficult problem, so you definitely want to take it one step at a time. Let me know if you have questions about Dookious' code. --Voidious 17:36, 19 May 2009 (UTC)

There are no threads on this page yet.