Mythbusters ArmA edition.

[NeMeSiS 11]

Very good question!

The reason I prefer to test AI-View-Blocking with a sea of bushes is that when I place various individual bushes in the editor to examine them, I (as the human player), can always see through the vegetation to the other side, to varying degrees based on the vegetation object. Seems to me that if I might be able to see AI on the other side (even only a few pixels), then it's really not a 100% complete test.

Now, I know that AI-View-Blocking is actually suppose to be based on view-lods. (Allegedly, along with other factors like surface, which may be at the heart of the problem.) But if I place a ton of these bushes in a very dense area, and still get detected and shot... well then imo it presents a much stronger case that there's a real underlying issue.

Unfortunately, when using the default terrains as they exists (Chernarus and Utes), no matter the location, there is always some ambiguity as to:

1. How thick was the vegetation?

2. Could I see some pixels through it?

3. Did the underlying view-lods actually block the entire view? And how would I even know if the view-lods did block the entire view, as I can't see them in-game?

Etc, etc, etc...

However, with a TON of hand placed bush objects packed tightly together, all doubt is removed. ;)

Right on... Fair enough. :)

Well, to be honest i assumed that MAP_EU added some bushes that were bugged or something, but looking at the bushes that were a problem im pretty sure that these dont have a viewlod at all. After all you cant behind one or two of them BI mst have thought that it isnt worth it i guess.

Which basically means that if one of those bushes doesnt block the AI's view, a million also wont. Ingame this isnt really a problem because you will never see a million of these tightly packed together, giving you the illusion of concealment.

For those not liking easyshare(I just used it because it was the first link google gave me :p ):

http://www.mediafire.com/?awld5afadoxmxx5

Edited July 27, 2010 by NeMeSiS

[walker 0]

Hi Bouben

I think you are confusing searching for you with knowing where you are.

Kind Regards walker

[NeMeSiS 11]

Hi Walker

I really dont understand why you dont see what we are trying to say.

Kind Regards NeMeSiS

[Guest]

I understand both you and walker mate.

In my opinion, I think perhaps you and Bouben are misunderstanding the AI work, because myself, walker, and the BIS staff are all unable to replicate your claims.

However, of course there is any possibility you are correct, but it may only affect very few machines, because most people can't replicate this.

Richie.

[Bouben 3]

I understand both you and walker mate.

In my opinion, I think perhaps you and Bouben are misunderstanding the AI work, because myself, walker, and the BIS staff are all unable to replicate your claims.

However, of course there is any possibility you are correct, but it may only affect very few machines, because most people can't replicate this.

Richie.

I suppose you are kidding...a lot.

Once again, the problem IS REPRODUCED, HAS IT'S TICKET AND IS ASSIGNED.

And your opinion that it may happen on only few machines, so it is computer based is absolutely ridiculous and may showing how superficial is your understanding of what the engine does.

And please, don't speak for BIS, they've never said anything like that.

You simply don't get it and I suppose you even have not read my special comment for you to clarify.

---------- Post added at 03:16 PM ---------- Previous post was at 03:14 PM ----------

Hi Walker

I really dont understand why you dont see what we are trying to say.

Kind Regards NeMeSiS

Yes...perhaps we're from a different planet or something...

---------- Post added at 03:17 PM ---------- Previous post was at 03:16 PM ----------

Hi Bouben

I think you are confusing searching for you with knowing where you are.

Kind Regards walker

No, I am not, because they don't have to search for you in grass. They will instantly pinpoint you.

Edited July 27, 2010 by Bouben

[Guest]

Bouben, I try and be polite, but it's obvious your just rude and arrogant.

You are WRONG, because we have PROOF and you do NOT.

And yes, if you search there are comments from BIS clearly stating that the AI can NOT see through grass.

Clear enough???

[Bouben 3]

Bouben, I try and be polite, but it's obvious your just rude and arrogant.

You are WRONG, because we have PROOF and you do NOT.

And yes, if you search there are comments from BIS clearly stating that the AI can NOT see through grass.

Clear enough???

What proof you have? Nothing? Or some chaotic gameplay experience where you THINK it is how it is? If you look at that ticket, there is a proof that you are wrong.

And if you read that BIS statement, they're basicly saying that yes, grass areas has some camouflage bonus, but that bonus is simply based on numbers in configs and has nothing to do with LOS of AI. Moreover - these numbers stops their function once a player have been spotted.

I've written exact examples of behaviour we're discussing here but you're simply still re-playing that old tape.

As if I was talking with a little child whose toy I've touched. This is incredible how ignorant can some people be. Please, study the engine a little and try to go over it despite your big love for the game.

Reproduced and confirmed. Sorry for touching your toy.

[4 IN 1 0]

You guys are confusing me, because it seems to me that you guys are arguing over things that are different to each other, can someone shine some lights on who is talking about what????

[walker 0]

Hi Walker

I really dont understand why you dont see what we are trying to say.

Kind Regards NeMeSiS

Hi NeMeSiS

Comments

I do understand your frustration and I can see what you are talking about but you are just misinterpreting the results of your experiment by not looking close enough at the excellent data it produces.

Experimental Errors in your test mission.

1) I tested your improved mission and once again you are putting the AI too close to the teleport spot.

Such that all you are in fact testing is the AI's spotting range when stimulated to aware and scan for enemies status. In fact when you look at the data from your test mission that is what is happening including your being spotted by the search party.

2) The ground at your chosen location slopes too much so I chose a flatter and clearer bit of ground for the experiment.

Corrected Test mission.

I have therefor produced this version of the test mission correcting the errors:

http://www.downloads.thechainofcommand.net/zips/AI_See_You_in_Grass_Mythbuster.Chernarus.zip

Alterations made

1) I found some flatter ground with no obstructing bushes or terrain between the AI and either the players initial position or the players teleport position.

2) I put the 2nd team member for the player back in the mission as per my previous edit of your mission. I only put it there for my convenience as I find it easier and more intuitive to use the command camera than the camera SQS. If you want to remove it by all means do. I do not believe it affects the experiment though.

3) I placed the camera SQS spawn point over the AI to get their perspective.

Suggested improvement to the experiment.

The experiment could be improved by adding a script to see through the eyes of any AI in the unit.

Method

I varied the initial player and teleport distances to the AI and observed the data lines and the AI's resultant actions closely.

Findings

I find:

The AI's ability to spot you on flat ground is ~100m

The AI ability to interpret you as an anomaly ~200m.

If you are beyond the anomaly spotting distance, search parties will not be sent to investigate you and will instead scan and search for you in a general manner.

Conclusions

The AI searching and scanning for you is what allows the AI to spot you, not X-Ray or Radar vision.

Kind Regards walker

Edited July 27, 2010 by walker

[Bouben 3]

Hi NeMeSiS

Comments

I do understand your frustration and I can see what you are talking about but you are just misinterpreting the results of your experiment by not looking close enough at the excellent data it produces.

Method

I varied the initial player and teleport distances to the AI and observed the data lines and the AI's resultant actions closely.

Findings

I find:

The AI's ability to spot you on flat ground is ~100m

The AI ability to interpret you as an anomaly ~200m.

If you are beyond the anomaly spotting distance, search parties will not be sent to investigate you and will instead scan and search for you in a general manner.

Conclusions

The AI searching and scanning for you is what allows the AI to spot you, not X-Ray or Radar vision.

Experimental Errors in your test mission.

1) I just tested your improved mission and once again you are putting the AI too close to the teleport spot.

Such that all you are in fact testing is the AI's spotting range when stimulated to aware and scan for enemies status. In fact when you look at the data from your test mission that is what is happening including your being spotted by the search party.

2) The ground at your chosen location slopes too much so I chose a flatter and clearer bit of ground for the experiment.

Corrected Test mission.

I have therefor produced this version of the test mission correcting the errors:

http://www.downloads.thechainofcommand.net/zips/AI_See_You_in_Grass_Mythbuster.Chernarus.zip

Alterations made

1) I found some flatter ground with no obstructing bushes or terrain between the AI and either the players initial position or the players teleport position.

2) I put the 2nd team member for the player back in the mission as per my previous edit of your mission. I only put it there for my convenience as I find it easier and more intuitive to use the command camera than the camera SQS. If you want to remove it by all means do. I do not believe it affects the experiment though.

3) I placed the camera SQS spawn point over the AI to get their perspective.

Suggested improvement to the experiment.

The experiment could be improved by adding a script to see through the eyes of any AI in the unit.

Kind Regards walker

I don't know if I understand your article well, but the point is, that you should test AI's seeing capabilities after you've attacked them. Nothing else. Stealth test without showing yourself to enemy is pointless.

I was told by Nemesis recently that he's preparing a explanation video at the moment. So lets wait for it.

[NeMeSiS 11]

RLVW0qSgXPw

As you all couldnt reproduce it (Except for Bouben and the people on Devheaven) i decided to make it a video instead. 720p is adviced if you want to read the globalchat.

Suggested improvement to the experiment.
The experiment could be improved by adding a script to see through the eyes of any AI in the unit.

I disagree, raw data is much more important and percieved position location and knowsabout value is already provided, i dont know what else you would want to know.

[whisper 0]

Mmmh, I didn't follow closely the "I'm holier than thou" fight above, but is the fact that the perceived position after teleporting is 300m away from real position a problem for you, Nemesis? That sounds about right to me :)

[NeMeSiS 11]

but is the fact that the perceived position after teleporting is 300m away from real position a problem for you, Nemesis? That sounds about right to me :)

It is supposed to be that distance. The problem is in the second part of the video when i am hiding in the grass. ;)

[whisper 0]

To make your experiment easier, add this to the init line of your soldier :

this addEvenHandler ["handleDamage", {0}]

It will make it invicible and allow you to really see AI reaction after they saw you.

for me, they should track you without problem (a bit like when you're not aware, you don't see that guy hidden, but once you look closely, you know where he is)

If there is an issue, it may be in the time it takes for them to spot you (ie, it is instant), but that is relatively minor

[NeMeSiS 11]

To make your experiment easier, add this to the init line of your soldier :

this addEvenHandler ["handleDamage", {}]

It will make it invicible and allow you to really see AI reaction after they saw you.

for me, they should track you without problem (a bit like when you're not aware, you don't see that guy hidden, but once you look closely, you know where he is)

If there is an issue, it may be in the time it takes for them to spot you (ie, it is instant), but that is relatively minor

The problem is that i teleported to a location i couldnt be seen in. If they wanted to track me they should have tracked me to the position i was firing in, making myself invicible will still make them attack me at my teleported position while they shouldnt know i am there in the first place.

(Knowsabout stays at 1.4 when they only hear my shots, but jumps higher when i am spotted, laying down or standing up in the grass has no effect on this)

[whisper 0]

You TP where you couldnt be seen, and AI was thinking you were still at your old position, what is wrong with that? oO

I fear I'm missing smtg :(

Edited July 27, 2010 by whisper

[NeMeSiS 11]

You TP where you couldnt be seen, and AI was thinking you were still at your old position, what is wrong with that? oO

No, they know my TP'ed position. A position i am not visible in if i am not known about beforehand. A position far away from when i was firing.

[whisper 0]

They don't know, they perceive you 300m away from there. LOL you lost me.

They say :

My knowsAbout is 1.4

I think you are at a position 300m away from your real position.

Ie, they look for you at a very wrong position, not your current one

[4 IN 1 0]

I think he is talking about that once you open fire and show your location, the AI will find you and even if you teleported yourself out of LOS and into hiden place before they will still pinpoint you out.

The one thing the video did not show is that if there is any unit scanning for incoming fire at different direction and spot you just as you teleport out infront of their LOS

Edited July 27, 2010 by 4 IN 1

[Guest]

I think he is talking about that once you open fire and show your location, the AI will find you and even if you teleported yourself out of LOS and into hiden place before they will still pinpoint you out.

But that is completely untrue, which can be proven by this post:

http://forums.bistudio.com/showpost.php?p=1698296&postcount=101

Once you teleport the AI will only know about the last position they saw you in.

[NeMeSiS 11]

They don't know, they perceive you 300m away from there. LOL you lost me.

They say :

My knowsAbout is 1.4

I think you are at a position 300m away from your real position.

Ie, they look for you at a very wrong position, not your current one

That is the last part of the video and that does not show the problem

The second part of the video shows the problem.

0:00 till 0:28 shows that my teleportation location is invisible.

0:28 till 0:47 shows that if i fire, then teleport to my invisible location i am somehow visible. This is the problem. The AI ignores the grass once they have a knowsabout about you, no matter how small.

0:47 till the end shows that if you hide somewhere else (not grass), everything works fine.

Richie did you even watch my video? It is very different compared to the video you posted. I am hiding in grass, he is hiding by distance, terrain and objects.

Edited July 27, 2010 by NeMeSiS

[walker 0]

I don't know if I understand your article well, but the point is, that you should test AI's seeing capabilities after you've attacked them. Nothing else. Stealth test without showing yourself to enemy is pointless.

I was told by Nemesis recently that he's preparing a explanation video at the moment. So lets wait for it.

Hi Bouben

I have altered the layout of the post I made in order to hopefully make it more easy to understand and to conform to a more formal scientific format. I must have been doing it while everyone else was reading and posting. Seems this is a popular thread :D

Bouben, I am not stupid. I am of course testing by firing at the enemy from intial location as well as doing the control test.

Abstract from the test

The findings are that:

1) An AI or AI Group Stimulated to combat mode can see and identify an entity as enemy at ~100m; but will not scan for you or spot you if you do not stimulate them by firing at them from the original location first unless you are under ~100m.

2) An AI or AI Group Stimulated to combat mode by firing from the initial position can see and identify an entity such as a teleported entity, that is not moving or firing as unknown at ~200m and under.

3) An AI or AI Group Stimulated to combat mode seeing an entity identified to be unknown will send out a search party toward the unknown in order to verify it as enemy and at ~100m identify it as enemy and engage it.

These AI spotting and target identification distances seem IMHO reasonable. I can do the same against camouflage cloths in open flat ground and can spot anomalies actually further than AI, in both real life and ArmA.

Kind Regards walker

Edited July 27, 2010 by walker

[whisper 0]

That is the last part of the video and that does not show the problem

The second part of the video shows the problem.

0:00 till 0:28 shows that my teleportation location is invisible.

0:28 till 0:47 shows that if i fire, then teleport to my invisible location i am somehow visible. This is the problem. The AI ignores the grass once they have a knowsabout about you, no matter how small.

0:47 till the end shows that if you hide somewhere else (not grass), everything works fine.

Got it, thks

In terms of position, is your TP position in the frontal arc of AI looking at your old position? If yes, can you try tp behind them?

[NeMeSiS 11]

Got it, thks

In terms of position, is your TP position in the frontal arc of AI looking at your old position? If yes, can you try tp behind them?

No, i am done with this.

[whisper 0]

awww kaayyy

....