User Details
- User Since
- Sep 2 2013, 12:56 PM (585 w, 3 d)
May 10 2016
Please, guys, don't go off topic in this ticket. This ticket is not about mods and their performance. This ticket is about an AI problem. Thank you.
Thank you, vlad 8011,
of course I am talking mainly about gunners targeting soft non-lockable targets. Also, there is a lot of crew manned weapons in the game (without any real-life aim-assist systems) that are suffering from the same problems I have described.
I guarantee, you would not be able to do what AI is doing in the repro mission (not in first person anyway). Have you tried it? If no, please take a look and you will know what I mean.
Also, this is a fundamental, gameplay AI problem and, IMO, this is no place to argument by mods. This should be fixed in vanilla.
Thank you.
EDIT: Also, this is a problem exclusive to crew manned weapons in the game. Infantry units do not have this behaviour.
Yes, thank you for addition.
Yes, three separate tickets with detailed steps to reproduce them. Otherwise it would be very hard to fix those issues.
This is not a helpful ticket. Wrongly structured, mixing many separate issues together without enough debug detail. Please, consider changing this.
Thank you very much for taking care of the rest! :o))
That is a nonsense, Illuminous. Your hands still have to obey physical laws. This is clearly a bug that can be exploited a lot during MP.
I am not telling you that your ticket is a spam. It is simply wrongly defined. That is all.
Generally, all the clutter that is spawning on a ground (I am not talking about actual bushes as objects) on vanilla islands should give you concealment when prone and AI should not see you through it. It gets tricky, though, if a part of your body isn't concealed as AI seem to not get confused by clutter as we, humans, do.
Also, if you give away your position and go prone AI will nevertheless fire on that last known position quite heavily so it looks as if they actually see you while they actually not. If you are on a completely flat ground (try grassy patches on an airfield) and you move away from that position while being in a prone stance, you will find out that AI does not see or track your movement and still fire on the last known position with some amount of dispersion. However, it is often hard to stay alive long enough to actually spot this behaviour so "player EnableDamage False" is necessary to test it properly.
As I said, though, it is possible that some parts of clutter does not represent this concealment properly and you will get spotted (happened to me with ported Arma 2 islands that seems to be bugged in this way in some places).
Also, some clarification from the devs about how clutter actually works would be nice because there is a chance it is just a simple formula based on a terrain type and not on actual clutter and LOS (that would mean that laying on a road would not conceal you even if it was on its edge where high patch of grass is as you would simply be spotted on a certain terrain type that does not provide any concealment in actual numbers).
I will upload a mission where the script is used (the script is not written by me) so you can experiment with it and tell us your observations.
You should have a private message from me on the forums.
Demongod, I really appreciate your tone. Very polite and constructive.
I have made an actual testing (that is results in actual numbers) about the grass thing and AI does not see through it. Have you made an exact testing or you just guess it based on your gameplay experience?
Put an exact repro-mission in the ticket. Then we will all see what is it about exactly. You might be right in some particular cases but until you deliver a proof this is just a speculation.
You cannot simply force AI going less prone "and that is it" because it will make them even more stupid and vulnerable. If you want to make them going less prone, you will have to change another parts of their behaviour in order to compensate for that (that is: make them take cover more efficiently, move more cleverly etc. - not very easy task to accomplish).
Also, the example with a Pawnee against AI infantry has nothing to do with AI going prone too much. It is all about AI not moving to cover. It has nothing to do about their stances. It is simply caused by a non-existent or not-triggered part of AI behaviour that would tell them to do so (that is, to move to a cover instead of just being static).
You seem to don't know much how the game works. Make some actual research and then come back because this is just cluttering of feedback tracker with a non-exact impressions about a general gameplay. Not very helpful.
EDIT:
Again, your repro-steps are totally irrelevant to the prone-position problem. You are mixing different problems into one.
You should have created separated tickets. For example: "AI is not taking cover when threaten by an enemy armor", "AI seeing through grass", "AI going prone in CQB too much" and "AI being too static in a combat".
You have to be very careful and precise while describing undesirable AI behaviour because it is a very complicated realm and each misinformation is not helping its development.
I down-voted the ticket. I absolutely don't agree with the 1 and 2 points. I believe you have managed to put a very poor arguments in there.
- Is not true on Arma 3 islands and if true in some scenarios (therefore bug) it should be fixed by fixing the LOS system and not by nerfing the AI behaviour.
- Absolutely ridiculous and unacceptable reason you put there (sorry). We should fix the AI shooting skills itself so that they are less super-human and not making them more stupid in their position selection in order to make them easier to hit.
3 and 4 are relevant points and could be fixed by timeouts before going prone or making cities CQB zones for which AI would have special behaviour. This would be very difficult to do properly IMO because there will be inevitable situations when you would love to have AI going prone even in CQB.
Guys, please, read the ticket description and submit your findings according to that.
You will have to design and upload a reliable repro-mission as these kind of issues are very tricky to debug. Typically, there is a lot of misunderstanding and myths involved.
There is no bug reproduced in the video.
The video shows an absolutely correct behaviour. You are playing the game wrong. Use the "copy my stance" instead of the "stand up" command that basically just forces units to stand up no matter the circumstances.
To others, please investigate a ticket before you vote it up.
This is not a proper ticket. You should make it clear and specific. Admins should close this one.
supercereal4: Please, note that the "Disengage" command has no other purpose than to cancel the "Engage" and "Engage at will" commands. It it not meant to be as a general retreat command or anything else apart from what I have written.
I have the same problem since beta.
They also use wire fences as solid walls. This are all problems since Arma 2.
A major issue indeed.
I don't agree with the proposal. This would mean AI would move forward and back all the time, taking too much time in the process. I think the solution would be more careful planning from the AI leader which unit to use for a particular engage order. There could be some distance check.
Finally someone made a ticket about this. I agree with the description. The stealth mode really needs to be slow and low. Sprinting and unnecessary movement is killing the AI, especially when facing a scanning APC.
After some random gameplay I believe even more that blocked LOS could have major influence on this bug. But I can be wrong.
EDIT: I will try to test this theory when I have some free time.
Here you go:
AI_engage_bug_far.Altis.rar [^] (87 bytes) 2013-09-05 00:31
AI_engage_bug_AT_test.Stratis.rar [^] (3,147 bytes) 2014-01-09 08:13
AI_engage_bug_AT_test.Stratis.zip [^] (4,465 bytes) 2014-01-09 08:32
Thank you.
OK, what I have found out so far is this:
- the engage command is apparently broken in many ways
- if there is a terrain or object blocking LOS of an unit which is supposed to engage a target, the engage process is not started at all (most of the time*)
- the distance between the engaging unit and its target might be an irrelevant factor after all but it will need more testing to confirm
- this bug is probably going to be rather complicated to test and fix**
- Sometimes the unit is able to start executing the engage process even if a terrain or object is blocking its LOS but will fail if the unit is not able to get direct LOS in few seconds after the start of the process. Sometimes the unit will run far far away in irrational directions.
- I am asking anyone with enough scripting knowledge if he/she is able to prepare better, exact and more analytical repro-mission to test this.
OK, I have made a test case for this and can finally confirm that it is NOT connected to the HOLD FIRE command.
It seems as if having no direct LOS is the reason for a unit to not to engage.
Also when he finally had direct LOS he started to engage but after a few steps reported "Negative" and stopped the engagement.
So definitely, HOLD FIRE is not the reason.
Another thing I have observed is that a unit with no LOS and far away from the target will say nothing and do nothing and just keep quiet after you give him an order. Then, when you move him closer but still with no LOS he will say "Engaging" but reports "Negative" after few seconds. With direct LOS it is similar but he actually tries something.
EDIT:
Also, the engage command is still not correctly ended when unit reports "Negative". The unit is apparently given only a silent "No target" order but "Engage" is still active. See the info in the ticket.
EDIT2:
Would it be possible that he is reporting "Negative" because he lost LOS again?
Uploaded a new repro mission and updated the info and steps.
EDIT: Updated the new repro mission.
EDIT2: Please, ignore the repro missions not mentioned in repro steps. I don't know how to delete obsolete uploads.
The settings you have described are deliberate and wanted. The behaviour you see is happening even without those settings set up in a normal single player gameplay. It is set up only for better control over the tests. You can order him to open fire when he is in the engage mode. Sometimes he will attack the tank, but with the infantry target it is even less stable.
Interesting. Thank you for the info.
EDIT:
Well, I have recently experienced situations when my AT gunner refused to engage an enemy armored car with his AT weapon. He was ordered to engage a spotted car but did absolutely nothing (no response, no movement). Then after a while said "Negative". Only when his LOS was completely clear he accepted the order but it took him quite a while to do something. HOLD FIRE was not active this time.
Thanks for heads up. Re-uploaded.
They also clip through the walls on the edge of the tower when prone.
InstaGoat, you should separate the ticket into two. One for crew being able to repair vehicles and one for crew not abandoning vehicles.
While I generally agree with the ticket, I suggest that a gunner should never leave the vehicle if there is a threat around and his weapons are able to operate. Returning fire should be of higher priority to damaged tracks even if the gunner is the only guy left. A driver and commander could try to repair a vehicle under fire but gunner not. He should be firing until his guns are destroyed or he is killed in his vehicle. This way it would be relatively easily programmable and rational enough for the good quality gameplay.
Also, I think a turret and an engine should not be able to be repaired in the field by the crew (there are support vehicles and engineers for this). Only tracks should be repairable by the crew but it should take some time.
ThePredator, I am sorry, but you can be the most accurate shooter in the universe but if you really deny there are muscles involved in holding and shooting a gun, then you really are an ignorant person. You cannot possibly mean it seriously. No way.
Is this ticket still relevant?
Down-voted, because I don't want BIS to waste a single valuable second of work with this.
I am for this to be improved only if it does not bring new AI problems to the game.
This feature was already implemented in America's Army (a PVP multiplayer game) and worked well. There is absolutely no gameplay reason not to implement it to the game. It would be a gameplay improvement adding some tension.
The ambient sounds in ARMA 3 are lacking compared to ARMA 2. I have never been to Limnos but I would say it is too quiet.
May 9 2016
I am always for options but it has to be cheater-proof.