User talk:Loser: Difference between revisions

User talk:Loser (view source)

Revision as of 13:11, 21 September 2008

2,575 bytes added , 21 September 2008

...

Loser

271

edits

@@ Line 234: / Line 234: @@
 ::Loser, before you get busy with school, please release the files for the sorta explosive barrel (ONIE, material, impact, etc): http://loser.oni2.net/Videos/Sorta_exploding_barrel.wmv  We can now edit the ONIE in .xml!
 ::I want to add to the request from geyser and Iritscen, the files related to jumping Mukade
+----
+OK, first I apologize for not released files, but I deleted them after video was taped. I tried to reproduce it, but it would require more time than I have now. So I have to tell you verbally what I done:
+*I tried to make it so AI2s can jump on obstacles. To achieve it, '''pathfinding grid''' was altered in selected BNVs. Those BNVs are (can be seen with "chr_show_bnv=1"):
+:*'''50''', that is main "jump area" where obstacles are
+:*'''52''', neighbour of 50
+:*'''54''', BNV where datapad lies
+:*'''46''', neighbour of 54
+*Originally, under those obstacles there are red "impassable" tiles. I changed it so there were orange "danger" tiles around obstacles and dark blue "border 4" inside of them.
+:*Also, experiments with [[AKBA]]s were done for BNV 54, tried to change GQs in [[AKAA]]s so AI2s can use alternate route which leads directly on the closest-to the-edge crate near savepoint 2 door. However, since AI2s then were a bit "dumb" (they kept falling down), pathfinding grids of BNV 54, 52 and 46 were modified as well, added double line of orange tiles along edge. Then it was almost OK.
+*About AI2 jumping abilities -
+:*Range where AI2 can switch to MELE (and to "vector of movement" instead of pathfinding) is still unknown for me. So I did not touch it.
+:*It looks that something is broken in AI2 jumping logic. When AI2 jumps, it calculates distance from where it should jump. This formula takes into account basic tap jump, all velocities etc. (well, I think, I cannot read it from engine) AND JETPACK TIMER AS WELL. At least it seems for me. But then, when executed, AI2 cannot do more than tap jump. So it ends in AI2 doing jumps too early. When jetpack timer was decreased AI2 started to jump more and more that it lands correctly so it can hurt enemy with airborne attack. In all experiments, victim (player) stood still. Dunno what is taken into account if victim moves.
+:* MELE was simplified, so it only contained ''throw'' and ''run jump forward punch''. Jump was strongly uplifted. Basically, AI2 jumped all the time ^_^.
+:*Max run in distance for run jump forward punch was riddiculously large (00|00|7A|44, something around 1000) to make sure AI2 WILL jump even tough it is behind obstacle.
+:*Min run in distance was about 0.5 (00|00|00|3F)
+:I guess this is all
+Overall, while being interesting exeriment, I think this is dead way. If I could, I would change it, so AI2 can jump over gaps and obstacles in pathfinding and in MELE, and this jumping is independent from jump attacks.
+--[[User:Loser|Loser]] 15:11, 21 September 2008 (CEST)