sc_activist (1), sc4.7, bots not shooting!!!, rcbots are not shooting enemies in this map!!! Options

[madmax2]

Cheeseh, for the most part the new rcbot1 seems to be working fine . I have only tested a small number of maps though, maybe 1/4th of the sc4.7 maps. So I can't be sure this issue won't show up in some other random map. For now, it's just information for you or anyone else, I know you are busy with rcbot2 and probably many other things. I just thought this should be reported, and if anyone else is running svencoop with rcbot, it wouldn't hurt to confirm this problem, thanks...

Ok this is a wierd one, and I'm not sure if it is actually a bot bug, it seems to be only this map ??? This is the first map I have found that rcbot build feb 2013, seems to have a major problem with. RCbots are not shooting enemies on this map!!

A little history first: I first noticed a problem on this map when I was going to play it on sc4.6. The bots rairly would leave the spawn room & had problems pressing the door buttons. I was using the waypoint from most of the available packs (2014 bytes). And after a short while they would freeze up, quit moving. I checked the waypoint, although incomplete, it looked good, I even added EOL/objective flags to the rcw, but no help. I tested the rcw on sc3.0 & 4.5 with the correct bot version/dll (rcbotbeta on sc3.0, etc). Basicly the same problem. The only version they would shoot temporarily on, was sc4.5. I ended up making a new waypoint for the map, and the bots started working normally again. So i guess the rcw in the packs was corrupt some how, or made on an old bot version maybe?

Anyways, along comes sc4.7 and now the bots don't shoot enemies again , even with the new 4.7 release dll, but they will attempt to run the map with my rcw. They don't freeze up, they just don't shoot enemies! I think this time it is a different problem, here's why and Here is what I have tried so far:

1. throughout this whole process I have deleted bot tmp & vis files etc. Its not that...

2. I just happen to have backup folders of older versions of sven. So just to be sure I wasn't going crazy, I did a couple tests. First I tried the sc4.6 rcbot_debug dll in sc4.7, no change, bots don't shoot. Next I replaced the sc4.7 folder in steam/HL with a sc4.6 folder. I tested both bot dll's in sc4.6, they both work normally, bots move out, shoot & run the whole map. Reinstalled sc4.7, back to the same problem with both dll's, in 4.7 only. Also, note that the map bsp has not changed since before sc3.0...

3. next I made a small/simple test waypoint in sc4.7 with the new rcbot dll release. They still won't shoot enemies.. So I think the rcw can be eliminated as the problem.

4. I figured I might as well try the rcbot debug_bot cmd, maybe it will tell you something useful. I ran it on murks first so I could see what would be normal.

Unlike murks, in activist I do not see the following when bots are being attacked (not in any order):

bot_task_reload
bot_task_normal_attack
bot_look_task_face_enemy
bot_look_task_search_for_enemy

What I did see on activist was this (not in any order):

bot_task_find_path
bot_look_task_next_wpt
bot_look_task_look_around
bot_task_use ............door button
bot_look_task_face_task_edict ...door button
bot_task_face_vector
bot_look_task_face_task_vector

and rarely crouch and run_for_cover

They only seemed to face the general direction of enemies when being shot, somewhat random it seemed. If they could, they run past the enemies it seems... they act like the enemies are invisible, but not thier bullets...

Here's a pic with debug on....
Bad Link Removed

Here is a test rcw I made on sc4.7/rcbot new release.
Bad Link Removed

[edit] Here is a fully playable rcw I made for sc4.6 & sc3.0
http://rcbot.bots-united.com/forums/index.php?showtopic=1753

Oh, I forgot to try putting the rcbot notarget_mode on a bot to see if they really are ignoring the enemy, i'll report on that later. It was kinda funny watching them get slaughtered by the barneys, etc..

[madmax2]

More Info:

Last night, I put notarget_mode on a bot and got an unexpected result. On the first floor of the map the bot ignored all the blue barney's and otis's and ran past them to the next floor. Ignored the barney's on the 2nd floor, before and just beyond the security door. He ran to the elevator, danced around a bit (bots have a little trouble pressing this button), When he saw the black suited swat enemies coming from the other elevator, he started shooting them, killing all of them. [edit] All but 2 barneys were killed, the other barney was either collateral damage or i just lost track of him. [edit]

He freezes on healing this barney ([edit]this has happened several times now[edit]). I shoot barney, he gives health to barney. He froze again, repeat - same thing, I killed barney, he ignored the other one and continued down the lift.

Bad Link Removed

I'll have to retest this in debug & see what happens. I didn't get the bot in position for the dogs, so i don't know about that? [edit] they do shoot dogs [edit]

[madmax2]

After more testing with notarget_mode on a bot, I can say bots never shoot barney's or otis's. Bots will shoot swats(human grunts), dogs, enemy scientists, the turret, and robogrunts.

The first thing they will shoot in the map is the swats here:
Bad Link Removed

Bots will heal damaged barneys, even when being shot. I removed notarget_mode from the bot and he continued to heal the barney shooting him!

With notarget_mode:
Bad Link Removed

Without notarget_mode:
Bad Link Removed

I'm not sure if this is very useful, here is debug_ent on a barney:
Bad Link Removed

Otis was the same info except the fuser2 value was 1.00 .

Here is debug_ent on a swat:
Bad Link Removed

Ok, think I'm done on this for now...

[Cheeseh]

Hmm.. Well I'm working on a version that you can edit the classes that bots will shoot via a cvar, and u can change it via a map specific config which might be useful,

[madmax2]

Last night I was checking out deadsimpleneo2 for waypointing. And I think bots were not shooting these guys too... (enemy bodygaurd with double uzi's)...
Bad Link Removed

Hmm.. Well I'm working on a version that you can edit the classes that bots will shoot via a cvar, and u can change it via a map specific config which might be useful,

That sounds good, thanks , but no rush.

Since you will be coding on this, could I get you to look at a couple other things at the same time? The first one is kinda minor, nit picky...

1. Bots seem to take very wide swings when avoiding other bots/players, this causes some problems at open doors where bots run into walls or edges of doors, and narrow ledges. Maybe that could be tweaked back a little?

2. Bots have a lot of trouble with human towers. I first noticed it on sc4.5/sc_royals2 towards the end of the map [edit] (Correction, sc_royals2, not svencooprpg2, my bad) [edit]. Bots seem to go off path a lot when approaching crouched bot/player, landing on thier shoulder or missing completely. It just seems they were more precice with towers with the rcbotbeta? I thought it might be the new avoidance code, but I'm not so sure now?

I'm not sure if it's avoidance, or just a timing thing? It's really only a problem when the bots have to jump a gap, such as sc_royals2 or sc_persia. But sc_persia is a very difficult tower/jump for the bots, because it's not a straight approach and it's a jump up to a plank first, so not really a fair test? So I made the following tower test rcw's for you to look at if & when you have time? Just add 2 bots and they will continueously circle back to the tower positions...

I tried to make the path in the first test rcw (crystal_towertest1.rcw) perfectly straight. It's just a mid air jump and simulates royals2 or persia. To get the new bots to do this, almost requires a stay close too flag on the HT wpt. Also a mid air jump wpt is needed, remove either flag and they mess up badly. This should show them go off path, especially if the sct flag is removed from the HT wpt. They will then often jump right over the crouched bot, probably because of the mid air jump wpt.

The second rcw (quarter_towertest2.rcw) is just a straight up against a crate tower.

The third rcw (quarter_towertest3.rcw), is like the first, where bots have to jump a gap or over a hole to a higher point. The way it is now, the bots can make the jump to the top of the crate, but every time they move past the HT w/o making the jump, means they just fell in a hole & died. I would like to see something closer to 50% success at least, if possible? I think if the second bot would just land on top of the first bot more reliably, the success rate would be much better? I don't know how common this type of HT is, I've only seen it twice so far?

Of the recent HT's I have set up, sc_persia came out the worst, Its near the beginning of the map if you want to check it out with my waypoint http://rcbot.bots-united.com/forums/index.php?showtopic=1688. Just beyond the HT is a peg jump I also used an HT waypoint at, so bots stop & crouch on the top peg, to make a jump to a roof across a wide gap (about 80% success on that), this took forever to do because of the problems at the first HT.

The other 2 maps I did HT's on came out much better, both are near the end of the maps, sc_royals2 and murks. murks is straight up to a pullup, from the top of a railing, its not easy for them to do, but they can make it. royals2 is a gap jump to a pullup, a crouch wpt at the top helped a lot here. A crouch on persia didn't seem to help, i don't think its a pullup, bots must jump on or over the railing. Seems like many of the HT's are in the worst places for bots or players...

Human towers just seem to be very difficult to waypoint with the latest bot versions. If you want any images or help testing, just say so....

Here are the test rcw's: