IPB

Welcome Guest ( Log In | Register )

2 Pages V  1 2 >  
Reply to this topicStart new topic
> RCBot 0.79 BETA DLL ONLY, debug dll downloadable from sourceforge
Cheeseh
post Feb 2 2014, 04:11 PM
Post #1


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



If you'd like to try a debug DLL on what I'm working on, have a look

this is a debug dll. to run this you must put -allowdebug into the command line. Also the file is much larger because its a debug DLL. It might be able to give you a crash log if it crashes.

http://sourceforge.net/p/rcbot2/code/HEAD/....dll?format=raw

there are more additions including
CODE

DOD:S improved combat behavior
DOD:S added some improved behaviours for machine gunners
DOD:S machine gunners deploy machine gun at cover points
DOD:S bots crouch and pull out to shoot the enemy at crouch cover points
DOD:S improved bots listening behaviour in DOD:S
DOD:S new cvar rcbot_listen_dist (make larger for bots to hear further)
DOD:S dod bots pick up weapons if out of ammo
DOD:S bot squads : disable by using rcbot_bots_form_squads 0
DOD:S bots use need ammo voice command
DOD:S bots give ammo to players who use the need ammo voice command
DOD:S bots use area-clear radio command
DOD:S fixed possible crash
DOD:S bots pick up weapons in dod:s
DOD:S added some new squad leader voice commands after capturing and defending
DOD:S bots in squads do their own thing if the squad leader is idle, but do not go too far from the leader

DOD:S/TF2 fixed sniper bug

DOD:S/TF2/HL2DM altered waypoint belief system
DOD:S/TF2/HL2DM bots are a little more aware of their environment, investigating enemies etc

for bot listening to work, dod_stats must be working
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
My Time
post Feb 2 2014, 05:45 PM
Post #2


Member
**

Group: Members
Posts: 49
Joined: 17-February 13
From: Southern IL
Member No.: 2,256



Cant wait to try this out ! Sent you some more $$$ for all the hard work here Cheeseh ! Thanks !
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
YuriFR
post Feb 3 2014, 01:33 AM
Post #3


Advanced Member
***

Group: Members
Posts: 51
Joined: 7-January 14
Member No.: 2,318



QUOTE(Cheeseh @ Feb 2 2014, 02:11 PM) *

If you'd like to try a debug DLL on what I'm working on, have a look

this is a debug dll. to run this you must put -allowdebug into the command line. Also the file is much larger because its a debug DLL. It might be able to give you a crash log if it crashes.

http://sourceforge.net/p/rcbot2/code/HEAD/....dll?format=raw

there are more additions including
[code]
DOD:S improved [(...)]


How are you, mr. Cheeseh? smile.gif

I used the "RCBot2 v0.79 BETA DLL", and I could see a VERY GOOD PROGRESS! NICE WORK! More below I report some points you might still improve for the official release for this version 0.79. But before that I need to report that there is some problem going on with my Rcbot2 (both versions 0.76 and 0.79beta), see more details on "oficial" post about the matter, at: http://rcbot.bots-united.com/forums/index.php?showtopic=1861

Mr. "genmac" complained that I was posting too much and got angry ( http://rcbot.bots-united.com/forums/index....amp;#entry12588 ). But I need to post more often, at least until the problem is resolved ... I do not want to be boring and not want to disrupt this forum. I ask for some patience...! I made many posts because I'm new here and needed special attention, to expose those ideas (about Rcbot2 and Sturmbot). It's just a phase: I will not be posting more all the time...

THE PROBLEM:
The fact is that I using files of "genmac" (v.0.76 - see link above). With them, I got to play with normal speed for about 3-4 hours using the Rc-bots (even with sourcemod and metamod!). I played several maps and many of them unofficial. Suddenly, for no reason, during a match on the map "dod_santos_b3", after 3 or 4 hours of normal play (and good matches), the game went back to being very very slow. I change the map and he went slow. I Restarted the Steam (I use the steampipe) and started a new map (deleted all type files .rcb .rcd .rcv), remained slow. Again I changed all the settings for video, audio, skill and number of bots without getting any effect. I played with 6 bots on each team, with "rcbot_skill" = 0.1 and the game continued slow (it's weird because I played 3-4 hours with normal speed, with rcbot_skill = 0.8 and 24 bots in total: 12 in each team... it is no make sense!). I do not know what else to do, after several hours of re-ajustments ... I used now the version of beta0.79-Rcbot and he remains slow, just as it is after the occurrence on map dod_santos_b3 (that could happen in any other map). I'm desperate because I really want the Rcbot2 return to work perfectly on my computer...!

ATTACHED are related materials, such as copies of "console" (on main-screen and inside a game/match), personal-settings of my Rcbot2, technical specifications of my computer, reproduction of my shortcut (to Dod: S with Rcbot2) and more... [etc].

------

The improvements that maybe you can do even to "oficial" version 0.79. I'm already using the 0.79b version, to write below:
- Bots still hit your friends with high frequency ("friendly fire"). I've seen bots deviate from their friends to gunshoot the enemy, but they often ignore the presence of allied bots, just ahead.
- A leader who wants to form a squad, get a lot more often negative responses. I saw a few positive responses...
- Maybe it's a wrong impression of me, but check out the frequency of refills of automatic machine-guns (MGs). Bots perform a few shots and then they begin to reload (soon they repeat the procedure, shooting little).
- Bots still exposes themself to enemy very easily, especially to gunfire of MG-guns. I saw this with great frequency when the bots want to conquer the last enemy flag. In this case they risk much as we were willing to commit suicide...
- Bots also "go back" to the enemy (or "sideways" to the enemy) - even to contradict the improvements of combat-behaviors for this version 0.79. It may also be that some particular configuration, mine, is causing this negligent behavior ... I don't know...
- Looks like the bots better realize the grenades (I liked the MG-gun out of the sandbags to escape the grenades !!!), but not how many they should run, to escape the impact area.
- Consists in describing the version inside the console, still the number "0.75", for the versions 0.76 and 0.79 beta, of Rcbot2. Please change it (or check) of: "0.75" to 0.76 (or to 0.79).

I warn you that these "tests" version 0.79beta with my PC, occurred with the problem of slow(ness). I would like to normalize the speed (fix the problem) in order to make a more "solid" assessment.
I sincerely hope to have helped (with some valid observation), and not disrupt you.


Thank you very much! smile.gif smile.gif


[my apologies: using google translator...]

Attached File  YuriFR_files_of_problem__slowness_.zip ( 28.57k ) Number of downloads: 378
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 3 2014, 06:28 AM
Post #4


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



Hi Yuri.

I cannot say for sure exactly what is causing the slowdown. But i know for sure rcbot will use much more CPU than playing online.

I would suggest keeping rcb rcv rcd files as they are designed to speed up the game, not slow it down.

Your CPU is good enough for online play but might suffer with a lot of bots . I would limit the number of bots, and reduce visrevs in each profile slightly. I believe overheating is causing the problem because my old computer also suffered this before i upgraded. Keeping the cpu cool will help too. AMD processors are notorious for heat and slower threading performance. My new cpu is intel and has far less heat and doesnt lag.

Does it only happen in that map only ? Note that some maps, especially custom maps have rather high r_speeds, (use r_speeds 1 and it will show you your FPS and number of polygons being rendered), even in HL1 with my old laptop it would lag badly in these custom maps. I would suggest running a dedicated server on 1 CPU core and then connecting to it via a client on the other core by means of "set affinity" in task manager to one of either CPU0 or CPU1.

Many people on the steam forums have reported of slowdowns caused by virus scanners running in the background. You could add the steam folders to your exception list in the virus scanner settings just to rule this out.

Defragment your hard drive if it is non-SSD (which I believe it probably isn't anyway) and defrag your steam files using the steam tool.

In the mean time i will search for ways to reduce cpu usage or find problems that might cause lag

PS. Yuri, your rcbot_listen_dist is too small (it is only 40) it should be between 500 and 1000 , if it is at 40 bots will not listen to anything outside melee distance and will drastically affect their behavior to listen for guns and investigate sounds
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
genmac
post Feb 3 2014, 11:07 AM
Post #5


RCBot Guru
*****

Group: Waypointers
Posts: 571
Joined: 11-November 11
Member No.: 2,098



Finally our Bot Master has spoken hehe!
I think you nailed a lot there on Yuri's lag issues.
Tnx for the new dll ....testing..
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 3 2014, 12:23 PM
Post #6


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



There is a bug where, if there is a breakable waypoint in the map, bots will shoot in the sky at a "dod_ragdoll" which appears sometimes, will need to fix that
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
madmax2
post Feb 4 2014, 01:43 AM
Post #7


RCBot Guru
*****

Group: Waypointers
Posts: 956
Joined: 2-March 12
From: USA, WA state
Member No.: 2,162



@Cheeseh

Nice, I will try the new version as soon as I can... Those are good performance suggestions...

@YuriFR

Guess what, our system specs are almost identical blink.gif ... And even though they are well over the minimums for this game, we are some what on the low end... Here is what I have...
------------------------
AMD Athlon 64 3200+ 2.2ghz (slightly faster clock rate, might be factory overclock or a newer revision?)

1.5 gb main memory (slightly less than your 2gb)

Nvidia Geforce 8600gt (overclocked) 512mb Ram ddr3 (better gpu)

Win XP sp3 (32bit)

ATA100 IDE hard drive
------------------------

For me, I usually only play with 7vs7 (or 6vs6, 11 or 13 bots total), for this reason, to avoid lag. It depends on the map. Some maps are ok at 8vs8, but some custom maps can cause lag for me even at 7vs7. I don't think you will get consistant performance, map to map, with that many bots? It will depend on the maps and how well they are optimized. Have you ever had the problem on any of the stock maps?

The Athlon 64 3200+ is single core, so i don't think you can do the dedi/client option on the same PC? If you have another PC, you could set up a dedi bot server on a LAN, thats what I intend to do at some point for dod:s. I have done it for other games with bots, on an older PC (dod classic, natural selections, svencoop, enemy territory, UT & UT2004, etc), and it helps a lot...

Yeah, the problem you discribe is a bit of a puzzle unsure.gif , I'm surprized to hear you can even run 3 to 4 hours without lag or slow down with that many bots, with our specs!

Like my PC, 1.5gb or 2gb main memory could be a bit low, even for winXP? I have had problems with virus scanners in the past, so I don't only disable mine, i completely unload it (some AV may not have that option?). If you are always connected to the internet when you play with the bots, I would suggest unplugging from the internet too, see if that helps.

Usually, cooling problems show up faster than 3 to 4 hours, but it is very possible, and there are a few things you can try to test for that. It could be a slow cooling fan causing a slow build up of heat inside the case, either a case exhaust fan or the CPU fan itself? If each day the first time you run the game/bots, it will run normal speed for a couple hours, then it fails, and the pattern is the same each new day, that is a clear symptom of a heat related problem. Laptops often have heat failures like this, and not all laptops are suitable for gaming. If you have a desktop/tower, you could pull the cover, check the cooling fans, etc. If nothing is obviously wrong, you could run it that way for a while, and even set up a house cooling fan to blow on it... If it stops failing, then it is a heat problem. Memory dimm's often fail this way too...

I noticed in your console [2] txt, the cpu is reported to be at 6.8ghz , yikes !!! I think you may have a hardware problem, heat or memory. Try re-seating the memory sticks too... Of coarse, this is all just a guess, from my own experience...

Good Luck wink.gif
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 4 2014, 02:55 AM
Post #8


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



QUOTE(madmax2 @ Feb 4 2014, 02:43 AM) *

The Athlon 64 3200+ is single core

I noticed in your console [2] txt, the cpu is reported to be at 6.8ghz , yikes !!!


Yeah, These points are quite important!

Also, Yuri in your bot profiles, your visionticks are way too high for your system. They should be between 8 - 20 for your system,edit all of them using the rcbotlauncher and reduce them to 10.

It's okay to have one or two profiles that are higher for better skilled bots though. On my system they use the default which is 10 and they are pretty good but could go higher on my system.

I've gone to using laptops over desktops for less hassle and can cool easier using a cooling pad. These are my specs.

Windows 8.1
Intel Core i5-3337U 1.8 GHz
8192 MB, DDR3 SDRAM
NVIDIA GeForce GT 750M - 2048 MB, Core: 941 MHz, Memory: 903 MHz, GDDR5-RAM
1920 x 1080 resolution
High settings in HL2/DOD:s

I run normally 20 bots for several sessions with default profiles, some custom profiles with higher visionticks. I don't have any slowdown. I'll also be able to run an HD movie on another TV connected to the laptop while playing with not much slowdown.

On my old laptop I got very laggy after adding more than 12 bots with default profiles, but was slightly lower spec than yours.
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 4 2014, 05:14 AM
Post #9


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



I've updated the DLL here again (same place)

http://sourceforge.net/p/rcbot2/code/HEAD/....dll?format=raw

with new changes
+ fixed bots shooting dod_ragdoll in air after a breakable at a breakable waypoint is broken
+ increased listening efficiency and made bots look more in the direction of noises and their previous enemy position.
+ bots go towards team mate if they hear their teammate firing a weapon and look in the direction they are firing
+ sometimes bots will go toward the point teammates are firing in, especially if they are already shooting in the direction of the bot
+ If bots can't see team mates attacking they first turn around to see where they are attacking, then look to where they are attacking
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 5 2014, 10:54 AM
Post #10


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



QUOTE(Cheeseh @ Feb 4 2014, 06:14 AM) *

I've updated the DLL here again (same place)

http://sourceforge.net/p/rcbot2/code/HEAD/....dll?format=raw

with new changes
+ fixed bots shooting dod_ragdoll in air after a breakable at a breakable waypoint is broken
+ increased listening efficiency and made bots look more in the direction of noises and their previous enemy position.
+ bots go towards team mate if they hear their teammate firing a weapon and look in the direction they are firing
+ sometimes bots will go toward the point teammates are firing in, especially if they are already shooting in the direction of the bot
+ If bots can't see team mates attacking they first turn around to see where they are attacking, then look to where they are attacking

a little video showing squad voice commands

http://youtu.be/KJxU8yp59VU

User is offlineProfile CardPM
Go to the top of the page
+Quote Post
YuriFR
post Feb 5 2014, 10:26 PM
Post #11


Advanced Member
***

Group: Members
Posts: 51
Joined: 7-January 14
Member No.: 2,318



QUOTE(Cheeseh @ Feb 5 2014, 08:54 AM) *

a little video showing squad voice commands
http://youtu.be/KJxU8yp59VU


First of all, I want to thank mr. madmax2 and mr. Cheeseh. You are very attentive and are helping me a lot! smile.gif

I Must also commend mr. Cheeseh and his work. I am increasingly impressed and optimistic, with every advance of each version! I really admire your ability to program! There are many very good things in Rcbot2 v0.79beta.
And please , mr.Cheeseh ... parodying mr. genmac in another post: "bot authors should stay away on asian beaches hehe"!
Stay away from these beaches (and Tsunamis), don't do like "Jowo"! smile.gif


Yesterday I reduced further configuration parameters from my PC and tested the dll-0.79 again. The graphics were crude. I also decreases the amount and configuration of bots, as recommended: 6 vs 5 (+1 human: I). EVEN WITH ALL THE DRASTIC REDUCTION, THE GAME REMAINS WITH LAGS ON ALL MAPS . Improved somewhat, indeed, but not like to normal speed, how he came to be before ... ( when I played 3-4 hours, 12X12 bots). On some maps the improvement of run-speed is almost imperceptible.

Settings of bots were as follows (v0.79):

For all profiles of 11 bots:
visionticks_clients = 9
visionticks = 9
pathticks = 50
braveness = 70
aimskill = 75
sensitivity = 14
--
On "config.ini":
rcbot_defrate 0.20
rcbot_skill 0.7
rcbot_anglespeed 0.21
rcbot_avoid_radius 80
rcbot_avoid_strength 100
(...)
rcbot_runplayer_hookonce 1
rcbot_runplayer_cmd_dods 417
rcbot_scoutdj 0.28
rcbot_heavyaimoffset 0.1
rcbot config min_bots 11
rcbot config max_bots 11
rcbot_listen_dist 510
rcbot_bots_form_squads 1


Comments:

- Really, mr. madmax2, be surprised, hehehe [laughter] laugh.gif ! I played with 24 bots in total with virtually all high-settings (bots and graphs), for about 3-4 hours without interruption (with the same speed of a online-game), passing several maps (official and unofficial). I'm not exaggerating or lying ... and it was an amazing experience for anyone who is a fan of bots ... smile.gif I was so happy! smile.gif For 3-4 hours I believed that the speed-problem had been solved on my PC ... ! Until the slowness returned from nothing ... sad.gif

- My PC is " desktop -tower " and single-core [answering to madmax2 and Cheeseh], I have no other PC. I checked the refrigeration (cooling fans etc.) and fittings and saw nothing unusual. Responding to madmax2 : apparently there is no clear standard for the manifestation of the "symptom-slowdown". Most often the first game session is already slow, even with only 6 vs 5 bots (+1 human).

- I used the "ProcessExplorer" software (in "Threads" tab) to identify the point responsible for excessive CPU usage, and is thus: " TID 3348 ; CSwitch 420 ; hl2.exe+0x17e1 ; CPU 58.46 " (the CPU-usage ranging from 58.46 to about 70). Special attention to hl2.exe+0x17e1!

- I wondered the possibility of some " looping " in Rcbot2, some on internal process, which requires the CPU unnecessarily. But however you do not have the same problem as mine ...

- I wondered that there may be some problem in loading the dll-Rcbot2 on my PC, perhaps loading the same processes (and dlls) 2 or more times - or load lots and lots of files. But however I believe that I took precautions and procedures to prevent this (on installation process).

- Viruses can not be, because I checked all the PC again (A.V., anti-spy/malwares and Combofix). Also includes Steam and Dod:S in exceptions-list of my A.V.

- I've defragmented (and optimize) the Hard Drives before returning to play with 6 vs 6 bots. So, the HD fragmentation can't be. But I did not could use the defragmenter of Steam because only now I became aware of it (I'll find out where it is and how it works) - anyway I think the fragmentation within the Steam is not the cause (it's only an opinion).

- The observation of mr.madmax2 (and mr.Cheeseh: "Yeah, These points are quite important!") Surprised me: "I noticed in your console [2] txt, the cpu is reported to be at 6.8ghz, yikes !!! I think you may have a hardware problem, heat or memory.". What does that mean? What are the implications of that "wrong report" about my system? Is dangerous(?), with respect to Rcbot2, or to Dod:S-online, or to the problem of slowness? How to fix it? Is it possible? Is it necessary?
Another related point is that in the main-screen (home) of Dod:Source, the console show a closer value of the reality (processor AMD Athlon 3200+): "1 CPU, Frequency: 2.0 Ghz, Features: AuthenticAMD SSE SSE2 MMX 3DNow RDTSC CMOV FCMOV" (source: file that I have attached previously, named "console [1] with Rcbot2.txt").
The value is changed for some reason, when it comes into a game session. I do not know what causes this and what is the purpose of this modification ... ! ... ?

- Some points worried me in particular (I thought on possibility of being related to slowdowns). When you can, I would like clarification (see more details in the file I sent, named "console [3] with Rcbot2 (strange 'things')":

a) It is normal to appear on the console, on each map, these things? : " *** working out visibility (...) percent ****** ", where "(...)" means a list of numbers ranging from 1 to 100 (or 99 ... can not remember)? There are 2 messages that also do not know if they are normal, or what they mean: "Warning! Client physic props overflow *max 120)" and "Set Gravity 800.0 (0.250 tolerance)".
Also appear with some frequency the following 3 lines: " *** Working out waypoint visibility information... *** " ; " *** Can't open Waypoint Visibility table for reading!" and " *** No waypoint visibility file *** ". The last three I believe to be a lack of files ".rcb ; .rcd ; .rcv" (on folder), when I erased them. That's right, is this normal?

b.) There are also other strange and relatively common messages: "JOY_AXIS_Z: unmapped" ; "JOY_AXIS_V: unmapped" (I do not use joystick). Also very common: "Error reading weapon data file for: weapon_ifm_steadycam", "GetUserSetting: cvar 'hap_HasDevice' unknown". "ISteamUserStats::GetStat failed to get progress value from Steam for achievement DOD_KILLS_WITH_SPRING_STAT" and "ISteamUserStats::GetStat failed to get progress value from Steam for achievement DOD_KILLS_WITH_BAZOOKA_STAT". Too: "Redownloading all lightmaps [;] R_RedownloadAllLightmaps took (...)" where "(...)" = number of seconds and "CGameEventManager::AddListener: event (...)" where "(...)" = list of multiple files. There is also:"Attemped to precache unknown particle system" where "(...)" = list of multiple files.
These messages are related to lags or slowdowns?

--------------------

Changing the subject:

I'll take you to a video posted above, mr. Cheeseh, and include some comments and questions:

- In beta-dll version 0.79, I could not make the bots (who were close) obey my command voice "I need backup!" When I was on a flag that needed more than 1 soldier for capture. Another "symptom" is that they did not approach and did not respond to my voice command with "yes" or "no". The feature already works (from human to bots)? Or is not ready yet...?

To add:

- In addition to answering "yes, sir!" or "no, sir!" (or "Negative, sir!"), the bots could move the head according to the answer ("yes" or "no") - this is only possible if there is already the animation, "embedded" on "vanilla game" (I think...).

- Bots can see through the cloud/dust from smoke-granades (I died several times this way). I think there is a command-line ("rcbot_ ...") can make the bot stop to see beyond the dust/cloud, but I forgot how this command line and if it works ...

- Bots started taking (good) cover on sandbags, more realistically in this 0.79beta version. VERY COOL ! You can add the possibility to reload the gun, crouched, when the bot is behind a sandbag (or piece of concrete etc.). I say this because a bot that was behind the sandbag, took the empty magazine, and reloading him, remained standing-position. He died because of it (had an enemy in front of the sandbag).

- There are some gaps in perception (by bots) of enemies in far windows. It may be that the reason is the waypoint (although the bot did not notice me in a window, on map Donner), or other reason related to the map (eg: mapping). If interested you, mr.Cheeseh, I take screen-shots of these windows (as I discovered) and post here (or send you e-mail).
Environmental perception of bots are much improved in this version 0.79beta. Nice! I think it should be further enhanced (and be more consistent), since the behavior of bots have varied: sometimes are able to perceive the threat of the enemy surrounding (eg: a close/near gunshot him), and they respond; sometimes the bots ignore/oblivious to: ballistic impacts, sounds of gunfire (or footsteps/gibber), the shouts of the enemy's voice commands etc.

- I noticed that the bots perceive me better (more realistically) in holes in the wall or windows near them, in this beta-version 0.79. Are also able to do a search/hunt inside the house where are the hole or window, if I crouch or lie down to hide (prone position). However, they often continue to route and ignore my existence within a certain construction (after seeing me hiding). When they do a search/hunt, it is not very consistent. That is, the bots do not employ efforts to perform the search/hunt, against me (or other bot). The bot that saw me hiding, and did not search/hunt me, in fact is in danger when ignore me: so the bot keeps the usual route, I leave the building/shelter and gun shoot on their backs, killing him.

- I'll be back to that idea, the "improving engagement off route" (exposed in that other thread: http://rcbot.bots-united.com/forums/index.php?showtopic=1858). I repeat this because you achieved greatly improve the environmental awareness of bots in this beta-version 0.79, mr. Cheeseh, and this is probably the time to improve the engagement "off route" too (perhaps along with improving the "bot-understanding" of ballistic impacts and perception of gun-sounds, voice commands, screams and closer melee-fights etc.).


THANKS FOR ATTENTION ! smile.gif

[my apologies: using google translator...]
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 6 2014, 12:25 AM
Post #12


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



your visionticks_clients is still a little high. This can be as low as 1 , 2 or 3 for your system.

With visionticks_clients 3 is perfect for your number of bots

this is a little calculation:

bot reaction time = ( number of players on system * 0.04 ) / visionticks_clients

for you : (12 * 0.04) / 3 = 0.16, an appropriate reaction time for your bots I'd think

however, because your game is smooth until 3-4 hours of play I'm guessing its an overheating problem.

IF there's any looping in rcbot2 -- the whole program would freeze!

QUOTE

a) It is normal to appear on the console, on each map, these things? : " *** working out visibility (...) percent ****** ", where "(...)" means a list of numbers ranging from 1 to 100 (or 99 ... can not remember)? There are 2 messages that also do not know if they are normal, or what they mean: "Warning! Client physic props overflow *max 120)" and "Set Gravity 800.0 (0.250 tolerance)".
Also appear with some frequency the following 3 lines: " *** Working out waypoint visibility information... *** " ; " *** Can't open Waypoint Visibility table for reading!" and " *** No waypoint visibility file *** ". The last three I believe to be a lack of files ".rcb ; .rcd ; .rcv" (on folder), when I erased them. That's right, is this normal?

If you erase the files, it will say those messages, common sense really. I prefer not to delete them, if you delete them it will use more cpu to generate them again.

QUOTE

b.) There are also other strange and relatively common messages: "JOY_AXIS_Z: unmapped" ; "JOY_AXIS_V: unmapped" (I do not use joystick). Also very common: "Error reading weapon data file for: weapon_ifm_steadycam", "GetUserSetting: cvar 'hap_HasDevice' unknown". "ISteamUserStats::GetStat failed to get progress value from Steam for achievement DOD_KILLS_WITH_SPRING_STAT" and "ISteamUserStats::GetStat failed to get progress value from Steam for achievement DOD_KILLS_WITH_BAZOOKA_STAT". Too: "Redownloading all lightmaps [;] R_RedownloadAllLightmaps took (...)" where "(...)" = number of seconds and "CGameEventManager::AddListener: event (...)" where "(...)" = list of multiple files. There is also:"Attemped to precache unknown particle system" where "(...)" = list of multiple files.
These messages are related to lags or slowdowns?

this is nothing to worry about

QUOTE

- In beta-dll version 0.79, I could not make the bots (who were close) obey my command voice "I need backup!" When I was on a flag that needed more than 1 soldier for capture. Another "symptom" is that they did not approach and did not respond to my voice command with "yes" or "no". The feature already works (from human to bots)? Or is not ready yet...?


to make bots follow you with 'need backup' the other bots must be able to see you and are within range, or in the latest version, or a squad leader. more additions are to be added for squad leaders.

QUOTE

- In addition to answering "yes, sir!" or "no, sir!" (or "Negative, sir!"), the bots could move the head according to the answer ("yes" or "no") -

this is only possible if dod:s offers such a thing as you said, the bot cannot add animations to the game

QUOTE

- Bots can see through the cloud/dust from smoke-granades (I died several times this way). I think there is a command-line ("rcbot_ ...") can make the bot stop to see beyond the dust/cloud, but I forgot how this command line and if it works ...


bots know about smoke grenades but sometimes it doesn't work perfectly, its a known issue, it will be improved upon.

QUOTE

- Bots started taking (good) cover on sandbags, more realistically in this 0.79beta version. VERY COOL ! You can add the possibility to reload the gun, crouched, when the bot is behind a sandbag (or piece of concrete etc.). I say this because a bot that was behind the sandbag, took the empty magazine, and reloading him, remained standing-position. He died because of it (had an enemy in front of the sandbag).

yes I will add this, however it requires crouch waypoints or special waypoints behind the sandbags for this to work

QUOTE

- There are some gaps in perception (by bots) of enemies in far windows. It may be that the reason is the waypoint (although the bot did not notice me in a window, on map Donner), or other reason related to the map (eg: mapping). If interested you, mr.Cheeseh, I take screen-shots of these windows (as I discovered) and post here (or send you e-mail).


this is a long known issue that madmax and genmac have debugged in the past

QUOTE

Environmental perception of bots are much improved in this version 0.79beta. Nice! I think it should be further enhanced (and be more consistent), since the behavior of bots have varied: sometimes are able to perceive the threat of the enemy surrounding (eg: a close/near gunshot him), and they respond; sometimes the bots ignore/oblivious to: ballistic impacts, sounds of gunfire (or footsteps/gibber), the shouts of the enemy's voice commands etc.


listening to footsteps will be added , enemy voice commands will also be considered.

QUOTE

- I noticed that the bots perceive me better (more realistically) in holes in the wall or windows near them, in this beta-version 0.79. Are also able to do a search/hunt inside the house where are the hole or window, if I crouch or lie down to hide (prone position). However, they often continue to route and ignore my existence within a certain construction (after seeing me hiding). When they do a search/hunt, it is not very consistent. That is, the bots do not employ efforts to perform the search/hunt, against me (or other bot). The bot that saw me hiding, and did not search/hunt me, in fact is in danger when ignore me: so the bot keeps the usual route, I leave the building/shelter and gun shoot on their backs, killing him.


bots decide whether to follow an enemy based on their current health and other objectives. can you tell me how to decide whether a bot will decide to keep attacking or defending the point or follow enemies?
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
YuriFR
post Feb 6 2014, 12:03 PM
Post #13


Advanced Member
***

Group: Members
Posts: 51
Joined: 7-January 14
Member No.: 2,318



Thank you very much , great mr . Cheeseh ! smile.gif

About the point below :

QUOTE(Cheeseh @ Feb 5 2014, 10:25 PM) *

bots decide whether to follow an enemy based on their current health and other objectives. can you tell me how to decide whether a bot will decide to keep attacking or defending the point or follow enemies?


I recall again the Sturmbot 1.7, where it was possible to note a more consistent engagement of the bot off course.
I could be talking a big silly, or simplifying something that might be very complex. I also do not understand programming ... But the idea is: once the bot sees an enemy (another bot or human) hiding (crouch/prone/evading inside etc.) into a hole in the wall, or next window of a building/construction, the "scale of priorities" of bots should suffer some temporary change. If the highest priority is to capture a flag around, or other purpose, with the sighting of the enemy hiding, the bot must "create" a new "point of objective" (like a "new flag", but able to move up; a kind of "ambulant-flag"), equal to or potentially equal to the weight of the previous objective, which to be "cull the immediate threat". This new target must have the potential to become the "top priority" at that moment (or even not to become, according on battle-situation), and stop being it as soon as the enemy (hiding) die. Logically the bot should be able to some "rational" calculation between two points of much interest (equal or nearly-equal weights), and based on this "calculation", (re)classify its "structure of priorities". That is, the bot should be able to decide between the "old target/goal" and the "emerging goal at moment", whether it is worth replacing the old with the new, based upon another variables ("battle-situation" cited above, or such as cited by you: "current health", "other objectives" etc.), or if not worth it. Something like that...
Maybe you can think about the possibility of developing a "math-scale" of "weights", able to guide the decision of bots, when emerge the "conflicts of objectives/tasks" (like ours "scheduling conflicts" in real life). Conflict, this caused by the unexpected sighting of the threat (bot or human, hiding).

I must stress that there are some simple steps that the bot can adopt under these circumstances, too, for example: throw a grenade in the window (or hole); the bot stand facing the window/hole with fixed sights (so that the enemy bot appears, and be hit by gunshoots); bot may also be able to hide for a while, in order to carry out an ambush against the enemy that was hiding in the window/hole, etc. Or even combinations of these tactics.

Only you can judge whether this idea is applicable, or if it has been discussed before ("is nothing new"), or if it "didn't work before", or even if she is a "huge crap"... !
There are other possibilities to develop this feature ("consistent engagement off route"), for the future. Certainly, with time, you may discover a way to do. smile.gif

All this are just SUGGESTIONS (for the future if it is difficult to apply). Are simply IDEAS. I do not want to burden him or disturb him, with it. Ok?!! smile.gif


Thanks and advance apologies if I did not help!
[I using google translator, sorry]


PS: I do not need to tamper with the following variables in my system (?): "visionticks" (= 9), "pathticks" (= 50), "rcbot_skill" (= 0.7), "rcbot_anglespeed" (= 0.21), "rcbot_scoutdj "(0.28 =)?
One last suggestion: you could eventually include an explanation (as the calculation you suggested to me, "visionticks_clients") for each of these variables within a text file ("readmes") to the most people "lay" understand the they should do. You can include some configuration standards, to set an example to the rest of the user community (due to varying systems)
.
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
My Time
post Feb 6 2014, 06:25 PM
Post #14


Member
**

Group: Members
Posts: 49
Joined: 17-February 13
From: Southern IL
Member No.: 2,256



This is just my two cents worth here but to me this seems like a memory leak problem. Like was said before if he was running on a laptop then maybe it would be a heating up over time problem as they do seem to build up heat the longer you use them but with a desktop/tower this should not be the case . Unless its a SFF case then it would still be enough air moving in them to not slowly build up heat. They would get to a point rather quickly and then stay at that point unless there are no fans running at all. This happening after being fine for 3 or 4 hours just does not point to the CPU overheating . Maybe his video card (fan not spinning up to speed because of dirt in the fan. Maybe then it would start to throttle after time) would cause this . To me it seem like something is robbing him of memory , slowly , until he then starts using his hard drive as Virtual memory. If its a 5400 rpm hard drive then his system would start to crawl . As it is , his system seem to be a little on the slow side(single core cpu , 256 meg video card) it would not take much to bring it to its knees .
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
SPYderman
post Feb 6 2014, 11:45 PM
Post #15


Advanced Member
***

Group: Members
Posts: 59
Joined: 10-August 13
Member No.: 2,299



Has your casual dedicated server crash, but still an amazing job. Keep it up.
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 7 2014, 01:49 PM
Post #16


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



yes It could totally be a memory leak --> virtual memory issue. I will look into possible memory leaks that could cause this. The amount of free RAM (2gb) is enough to possibly cause overload after 3-4 hours. I've got 8GB so it would take more than 4 times the amount of time to see any effect for this issue!
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Ted
post Feb 9 2014, 11:28 AM
Post #17


Member
**

Group: Members
Posts: 13
Joined: 20-February 13
Member No.: 2,257



I've updated the OS X and LINUX builds download links with builds including these changes. Also I found a bug that causes the game to crash when a player joins a server after a bot is kicked. A fake client check is needed before invoking the bot creation function here. The fix is already incorporated into the LINUX and OS X builds.
QUOTE(Cheeseh @ Feb 4 2014, 05:14 AM) *

I've updated the DLL here again (same place)

http://sourceforge.net/p/rcbot2/code/HEAD/....dll?format=raw

with new changes
+ fixed bots shooting dod_ragdoll in air after a breakable at a breakable waypoint is broken
+ increased listening efficiency and made bots look more in the direction of noises and their previous enemy position.
+ bots go towards team mate if they hear their teammate firing a weapon and look in the direction they are firing
+ sometimes bots will go toward the point teammates are firing in, especially if they are already shooting in the direction of the bot
+ If bots can't see team mates attacking they first turn around to see where they are attacking, then look to where they are attacking

User is offlineProfile CardPM
Go to the top of the page
+Quote Post
Cheeseh
post Feb 9 2014, 01:27 PM
Post #18


Admin
*****

Group: Admin
Posts: 3,055
Joined: 11-September 03
From: uk
Member No.: 1



QUOTE(Ted @ Feb 9 2014, 12:28 PM) *

I've updated the OS X and LINUX builds download links with builds including these changes. Also I found a bug that causes the game to crash when a player joins a server after a bot is kicked. A fake client check is needed before invoking the bot creation function here. The fix is already incorporated into the LINUX and OS X builds.

yes I noticed this bug before, but every time I checked for fakeclient it would have no info, although I guess waiting would fix the issue.
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
YuriFR
post Feb 10 2014, 04:53 PM
Post #19


Advanced Member
***

Group: Members
Posts: 51
Joined: 7-January 14
Member No.: 2,318



Since the problem is yet persistent, I think also be caused (maybe) by memory leak, on any Rcbot2 process:

QUOTE(Cheeseh @ Feb 7 2014, 11:49 AM) *

yes It could totally be a memory leak --> virtual memory issue. I will look into possible memory leaks that could cause this. The amount of free RAM (2gb) is enough to possibly cause overload after 3-4 hours. I've got 8GB so it would take more than 4 times the amount of time to see any effect for this issue!



We are anxious for the correction of bugs, and stabilization of version (0.79b), mr. Cheeseh! smile.gif
I also cheer for the lag/slowdowns problem to be solved soon ...


I have some questions, all aimed at the any forum members (I have not found satisfactory answers by the "search engine" here, and other web-sites, sorry) :

1 - In the Rcbot2 profiles, contain the command-line: "model = default".
I know it's possible to change the model of the character (by above command: directory to extension ".mdl").
I wonder if it is possible to change just the skins of each character/bot, per class (trough the files ".vmt", ".vtf" etc.) by the Rcbot2. Is it possible? If is not possible, how do through Dod:Source (v.Steampipe) to determine a skin for each of the soldiers/bots (and each class), preferably without resorting to the extension ".mdl" (and their ligaments: ".vvd", ".vtx", ".phy" etc.)?

2 - I want to limit the number of bots to make them 6vs6 (for now) per match, but often the teams are unbalanced (but without going beyond total of 12). Is there any command like "teambalance" inside of Rcbot2, or inside Dod:Source? I tried "mp_autoteambalance", "mp_limitteams" and "mp_teams_unbalance_limit", "mp_autoconcede", but did not work for me. If Rcbot2 can do a kind of "teambalance", he is able to do so even if the number of available profiles exceeds the maximum number chosen for the each match (eg: 50 profiles = 50".ini", but inside the game-match: limited to 6vs6 game)?

3 - I tried once, slightly increase the duration of dead-bodies on the ground (ragdolls), and also the time of the helmets and weapons that are fallen on ground, but I could not do. I've tried several related commands ("cl_ragdoll_physics_enable", "cl_ragdoll_fade_time", "cl_ragdoll_colide", "cl_ragdoll..." etc.). How should I proceed? The Rcbot2 is also able to keep the dead bodies/ragdolls, helmets and weapons on the floor for longer (with some command-line)?

The next 2 questions is for you, mr. Cheeseh:
4 - The bots of Rcbot2 has the "habit" of taking leaps/jumps vertically for no apparent reasons. This is a bug? Or is there some command-line able to reduce/eliminate these frequency (of jumps/leaps)? [In the video of version 0.79b, provided by you, can notice this behavior: see the bot jumping ahead you, in the begining: http://youtu.be/KJxU8yp59VU ].

5 - Can I force the bots to throw hand-grenades, instead of "shot grenades" with the rifle-grenades? How do I do that (there are some command-line)?

Thanks!

[I using google translator, sorry]
User is offlineProfile CardPM
Go to the top of the page
+Quote Post
genmac
post Feb 11 2014, 04:06 AM
Post #20


RCBot Guru
*****

Group: Waypointers
Posts: 571
Joined: 11-November 11
Member No.: 2,098



QUOTE(Ted @ Feb 9 2014, 11:28 AM) *

I've updated the OS X and LINUX builds download links with builds including these changes. Also I found a bug that causes the game to crash when a player joins a server after a bot is kicked. A fake client check is needed before invoking the bot creation function here. The fix is already incorporated into the LINUX and OS X builds.

nice! tnx Ted!
User is offlineProfile CardPM
Go to the top of the page
+Quote Post

2 Pages V  1 2 >
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



- Lo-Fi Version Time is now: 29th March 2024 - 12:53 PM