AI Discussion (dev branch)

[f2k sel 164]

Does anyone else believe this thread is not serving its original purpose. I see very little of the AI devs on here, probably cos it isn't very much use.

I suggest that rather having a free for all, that the devs ask for feedback on specific features that get put into the dev build daily, such as "how is the AI stances in today's build? etc

+1

Last week did see a little improvement at close range but the problem still persists when AI exit a vehicle.

They can kill you but they can also unload a whole clip at 50 meters without a hit.

They Then tend to run off and even if they turn around and clearly see you they run off again.

Also sometimes they get out the vehicle and never fire they just kneel for a while and run off.

[Tyl3r99 41]

please bare in mind that these devs have a deadline to release the full game, given by their CEO.

i am more than sure that issues will be addressed when they can.

the fact that they have opened a thread about AI is a huge step forward.

try and understand things from their point of veiw.

Does anyone else believe this thread is not serving its original purpose. I see very little of the AI devs on here, probably cos it isn't very much use.

I suggest that rather having a free for all, that the devs ask for feedback on specific features that get put into the dev build daily, such as "how is the AI stances in today's build? etc

[antoineflemming 14]

Does anyone else believe this thread is not serving its original purpose. I see very little of the AI devs on here, probably cos it isn't very much use.

I suggest that rather having a free for all, that the devs ask for feedback on specific features that get put into the dev build daily, such as "how is the AI stances in today's build? etc

No, this thread is serving it's intended purpose: to be a discussion thread for THIS thread: Development Branch Captain's AI Log

You can use the feedback tracker to offer up feedback, because, you know, that's its intended use.

[Shagulon 1]

there is precious little discussion about the Captain's AI log that I can see, which was my point. Much more discussion about people's pet peve's and Arma 2 (and how the thread has been derailed - I know I know)

BUT if you all just prefer to bicker then whatever.

[froggyluv 2136]

Shagulon is of course correct but much like the general Dev Branch discussion, it's hard not to try and voice one's biggest gripes over a topic that's been silent for so long.

But again, I completely agree -best to leave this topic to confirmed changes and use the numerous AI threads for outright wishes and general state of AI debate. :)

[-Coulum- 35]

I suggest that rather having a free for all, that the devs ask for feedback on specific features that get put into the dev build daily, such as "how is the AI stances in today's build? etc

Grrr why'd you have to point that out before I got my ramble in (at least my offtopic one, I'm still gonna ramble:))? I guess I'll stay on topic.

I think the new stance system is a huge improvement and I am so glad I will not have to rely on ASR to get my ai to use more appropriate stances. That being said I still think it could be a little more situational (from a players perspective I have no clue from a technical perspective). By that I mean the ai needs to not only take into account what behaviour mode his squad is in but also other factors - in particular how close the nearest known enemies are.

Right now the ai routine in combat mode seems to be:

1. Move at jogging pace half a dozen metres.
   
2. Remain stationary and in same stance for 2 seconds or so (to observe surrounding I suppose)
   
   • ai may remain in this stance longer if they decide to shoot an enemy - this is why we previously had ai standing and shooting at us.
     

[*]Drop to prone.

[*]Stay prone until the rest of the squad has made there move.

[*]Start from step one again.

Basically in step 1 and 2 the ai has a set stance and in step 3, 4, and 5 the ai are prone.

TL;DR ai are always prone unless they are moving or just finished a move in which case they are crouching.

The problem is, moving in crouch is not always the most efficient way to move. In less dangerous situations (Ie. when the enemy is far) it is more efficient to stand up right and move. So when choosing a stance to move in ai should take into account how far (or better yet how dangerous) the enemy is. I've written up a chart comparing what I see now and what I think things should be like: (in the spoiler)

NOW

[TABLE=class: grid, width: 500]

[TR]

[TD]Distance to Enemies[/TD]

[TD]Stance and Speed When Moving in Combat Mode[/TD]

[/TR]

[TR]

[TD]300m+[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]200m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]100m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]50m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]25m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]10m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[/TABLE]

SHOULD BE

[TABLE=class: grid, width: 500]

[TR]

[TD]Distance to Enemies[/TD]

[TD]Stance and Speed When Moving in Combat Mode[/TD]

[/TR]

[TR]

[TD]300m+[/TD]

[TD]Standing, Jog[/TD]

[/TR]

[TR]

[TD]200m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]100m[/TD]

[TD]Crouched, Sprinting[/TD]

[/TR]

[TR]

[TD]50m[/TD]

[TD]Standing, Sprinting[/TD]

[/TR]

[TR]

[TD]25m[/TD]

[TD]Standing, Tactical Pace[/TD]

[/TR]

[TR]

[TD]10m[/TD]

[TD]Standing, Walking[/TD]

[/TR]

[/TABLE]

Notice how the current really doesn't take into account how close the enemy is (from my observations). The suggested chart does. Add a little randomness to the distance in the second chart (ie. +/- random 0-20%) and I think this could make for ai that appear to really be reacting to their situation with the correct stance rather than just sticking to a stance because they are in the almighty danger mode.

Of course there is still one other flaw - the ai don't like to change stance once they start shooting - so if they are running standing up, stop and pause for a second, and see someone before they decide to go prone they will remain standing until they have successfully engaged the target. What should happen is, when the ai decides he wants to engage an enemy, he should FIRST change to the most accurate stance available (That won't obstruct his LOS to the enemy) and then proceed to engage. They do not do this now.

Anyhow, TL;DR, make the AI stance selection more based on the proximity to the enemy, and make ai change stance not only to move or remain stationary, but also to best engage an enemy should they decide to do so.

As always I am not really an expert on the ai and if anyone wants to point Out false observations I appreciate it.

[sqb-sma 66]

Of course there is still one other flaw - the ai don't like to change stance once they start shooting - so if they are running standing up, stop and pause for a second, and see someone before they decide to go prone they will remain standing until they have successfully engaged the target. What should happen is, when the ai decides he wants to engage an enemy, he should FIRST change to the most accurate stance available (That won't obstruct his LOS to the enemy) and then proceed to engage. They do not do this now.

I mostly agree, but I can imagine this going awry. You know how the AI takes so long to turn and engage you currently? Imagine if you turned a corner only to run into the barrel of the AI! But he just crouches and then shoots, or there's a small delay. IMO suppressive fire should come first, if the enemy is very close this fire should hit and kill but it shouldn't be so accurate as to hit at 50+ meters reliably. After 5-10 rounds of this they should do everything you mention.

[afp 1]

I think it's this done wrong, the turn speed isn't fixed but they're working on AI behavior. When turn speed will be fixed it will appear a different AI. The micro management should be done first, then the general behavior. Priorities would be:

1. Turn speed. You are still able to follow an AI who can't shoot at you.

2. Flee conditions. Sometimes AI refuses to shoot at you, sometimes even running away. Is it flee? If so, make it drop the weapon or an explicit surrender animation. Else make it shoot.

3. Gunshot/sounds detection - they don't care if you shoot behind them, they don't turn around to check sometimes. If you shoot after corner they get past the corner not looking.

4. Cover detection - related to enemy direction.

When these will be fixed, a lot of the waypoints behavior will fix itself.

Of course there is still one other flaw - the ai don't like to change stance once they start shooting - so if they are running standing up, stop and pause for a second, and see someone before they decide to go prone they will remain standing until they have successfully engaged the target. What should happen is, when the ai decides he wants to engage an enemy, he should FIRST change to the most accurate stance available (That won't obstruct his LOS to the enemy) and then proceed to engage. They do not do this now.

Anyhow, make the AI stance selection more based on the proximity to the enemy, and make ai change stance not only to move or remain stationary, but also to best engage an enemy should they decide to do so.

I'm not sure, they look kind of natural the way they are now...once the combat starts you don't want them moving around pointless, better wait to shoot. Of course, if they have to move far they should stand, not crawl all around the map. Also, if enemy is very close, they should engage no matter the stance, if they go prone they already waste 2-3 seconds. I agree with you that the proximity to enemy should be one of the most important factors in decisions.

Edited July 30, 2013 by afp

[Tyl3r99 41]

i like this idea

Grrr why'd you have to point that out before I got my ramble in (at least my offtopic one, I'm still gonna ramble:))? I guess I'll stay on topic.

I think the new stance system is a huge improvement and I am so glad I will not have to rely on ASR to get my ai to use more appropriate stances. That being said I still think it could be a little more situational (from a players perspective I have no clue from a technical perspective). By that I mean the ai needs to not only take into account what behaviour mode his squad is in but also other factors - in particular how close the nearest known enemies are.

Right now the ai routine in combat mode seems to be:

1. Move at jogging pace half a dozen metres.
   
2. Remain stationary and in same stance for 2 seconds or so (to observe surrounding I suppose)
   
   • ai may remain in this stance longer if they decide to shoot an enemy - this is why we previously had ai standing and shooting at us.
     

[*]Drop to prone.

[*]Stay prone until the rest of the squad has made there move.

[*]Start from step one again.

Basically in step 1 and 2 the ai has a set stance and in step 3, 4, and 5 the ai are prone.

TL;DR ai are always prone unless they are moving or just finished a move in which case they are crouching.

The problem is, moving in crouch is not always the most efficient way to move. In less dangerous situations (Ie. when the enemy is far) it is more efficient to stand up right and move. So when choosing a stance to move in ai should take into account how far (or better yet how dangerous) the enemy is. I've written up a chart comparing what I see now and what I think things should be like: (in the spoiler)

NOW

[TABLE=class: grid, width: 500]

[TR]

[TD]Distance to Enemies[/TD]

[TD]Stance and Speed When Moving in Combat Mode[/TD]

[/TR]

[TR]

[TD]300m+[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]200m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]100m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]50m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]25m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]10m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[/TABLE]

SHOULD BE

[TABLE=class: grid, width: 500]

[TR]

[TD]Distance to Enemies[/TD]

[TD]Stance and Speed When Moving in Combat Mode[/TD]

[/TR]

[TR]

[TD]300m+[/TD]

[TD]Standing, Jog[/TD]

[/TR]

[TR]

[TD]200m[/TD]

[TD]Crouched, Jog[/TD]

[/TR]

[TR]

[TD]100m[/TD]

[TD]Crouched, Sprinting[/TD]

[/TR]

[TR]

[TD]50m[/TD]

[TD]Standing, Sprinting[/TD]

[/TR]

[TR]

[TD]25m[/TD]

[TD]Standing, Tactical Pace[/TD]

[/TR]

[TR]

[TD]10m[/TD]

[TD]Standing, Walking[/TD]

[/TR]

[/TABLE]

Notice how the current really doesn't take into account how close the enemy is (from my observations). The suggested chart does. Add a little randomness to the distance in the second chart (ie. +/- random 0-20%) and I think this could make for ai that appear to really be reacting to their situation with the correct stance rather than just sticking to a stance because they are in the almighty danger mode.

Of course there is still one other flaw - the ai don't like to change stance once they start shooting - so if they are running standing up, stop and pause for a second, and see someone before they decide to go prone they will remain standing until they have successfully engaged the target. What should happen is, when the ai decides he wants to engage an enemy, he should FIRST change to the most accurate stance available (That won't obstruct his LOS to the enemy) and then proceed to engage. They do not do this now.

Anyhow, TL;DR, make the AI stance selection more based on the proximity to the enemy, and make ai change stance not only to move or remain stationary, but also to best engage an enemy should they decide to do so.

As always I am not really an expert on the ai and if anyone wants to point Out false observations I appreciate it.

[-Coulum- 35]

I mostly agree, but I can imagine this going awry. You know how the AI takes so long to turn and engage you currently? Imagine if you turned a corner only to run into the barrel of the AI! But he just crouches and then shoots, or there's a small delay.

I'm not sure, they look kind of natural the way they are now...once the combat starts you don't want them moving around pointless, better wait to shoot. Of course, if they have to move far they should stand, not crawl all around the map. Also, if enemy is very close, they should engage no matter the stance, if they go prone they already waste 2-3 seconds.

Yes very good point that I didn't think of. Once again, the ideal solution would be to make ai reaction more based on the circumstances - ie. proximity to enemy. FOr example if the ai is only 10 metres from the enemy and decides to engage it shouldn't change stance to do so. But if the a is 300 metres from an enemy, there isn't as much rush, and ai should first to the most effective shooting stance possible and then start shooting.

Having a "one size fits all" will simply not work for a game with so many possible different situations.

BTW my above post was based on an older dev build when ai crouch was first introduced. Now I see that the ai use prone much more and crouch much less. Seems to me the thought process is:

I'm in danger mode and want to make a bound --> 75% chance I will use prone 25% chance I will use crouch.

When I think it should be

I'm in danger mode and want to make a bound + the enemy is 400 metres away --> use standing jog or

I'm in danger mode and want to make a bound + the enemy is 200 metres away --> use crouched jog or

I'm in danger mode and want to make a bound + the enemy is 50 metres away --> use standing tactical pace...

You get the the point - currently ai seems not to consider the enemy when choosing stance and speed. I think they should. Same goes for when choosing a stance to shoot from.depending on the range to the target, the ai should or shouldn't change to a more effective firing stance.

BTW while testing this I found that the ai is horrible at handling recoil right now (even at point blank), anyone else notice this? Also saw great grenade usage by the ai!

I think it's this done wrong, the turn speed isn't fixed but they're working on AI behavior. When turn speed will be fixed it will appear a different AI. The micro management should be done first, then the general behavior.

Yes I agree - start small and then go big. Though my priorities would be a bit different ie. turnspeed, proper stance usage (based on enemy and suppression), ability to recognize and use (properly) all objects as cover and then skill stuff (realistic shot detection, recoil management, shooting moving targets).

Edited July 30, 2013 by -Coulum-

[afp 1]

I think the slow turn speed and all around could be because there is one single aimingSpeed in cfgAISkill

aimingAccuracy = {0, 0, 1, 1};

aimingShake = {0, 0, 1, 1};

aimingSpeed = {0, 0.500000, 1, 1};

courage = {0, 0, 1, 1};

commanding = {0, 0, 1, 1};

endurance = {0, 0, 1, 1};

general = {0, 0, 1, 1};

reloadSpeed = {0, 0, 1, 1};

spotDistance = {0, 0.200000, 1, 0.400000};

spotTime = {0, 0, 1, 0.700000};

Maybe its better to have a "turnSpeed" and an "aimingSpeed" that reffers only to how fast they aquire the target the first time. This will ofer a more fine tunning over AI engagement. They could quickly turn to you but not engage that fast, without aiming.

[Masharra 10]

2. Flee conditions. Sometimes AI refuses to shoot at you, sometimes even running away. Is it flee? If so, make it drop the weapon or an explicit surrender animation. Else make it shoot.

I would rather they behave like trained infantry and when ordered to disengage they dont act like idiots. Unless when the player disengages from battle they are forced to do the same.

Though optionally as like in gl4 if enough of the squad is destroyed and not close enough to a soldier the chance of the ai surrendering/ playing dead should appear.

 // ==============================================================
   // GL4 Enemy A.I. Surrender:
   // ==============================================================
   // Choose if enemy A.I. should be able to surrender if they are fleeing.
   // True / False, default is True
     // GL4_Global set [50, False];

     // Enemy A.I. Surrender Courage Fixed:
     // Choose the enemy A.I. surrender courage.
     // Note: The higher this value the higher the chance that enemy A.I. will surrender.
     // 0 - 1, default is 0.1
       // GL4_Global set [51, 0.1];

     // Enemy A.I. Surrender Courage Random:
     // Choose the random enemy A.I. surrender courage.
     // Note: This value will randomly increase the "Fixed" value given above.
     // 0 - 1, default is 0.5
       // GL4_Global set [52, 0.5];

     // Enemy A.I. Surrender Rank Chance:
     // Choose the chance of how fast enemy A.I. with a specific rank should be able to surrender.
     // 0 - 100, default is PRIVATE: 70% | CORPORAL: 60% | SERGEANT: 50% | LIEUTENANT: 40% | CAPTAIN: 30% | MAJOR: 20% | Colonel: 10%
       // GL4_Global set [53, [70, 60, 50, 40, 30, 20, 10] ];

     // Enemy A.I. Surrender Group Size:
     // Choose how much units of a enemy A.I. group must be alive till a unit of the enemy A.I. group will think about surrendering.
     // 1 - 10, default is 3
       // GL4_Global set [54, 3];

     // Enemy A.I. Surrender Friendly Nearby:
     // Choose how much friendly A.I. units have to be nearby that the unit will not surrender.
     // Note: If there are more friendly A.I. units nearby like set here then the unit will not surrender.
     // 0 - 100, default is 5
       // GL4_Global set [55, 5];

     // Enemy A.I. Surrender Captive Player:
     // Choose how many captives each player can capture at the same time.
     // 0 - 10, default is 3
       // GL4_Global set [56, 3];

     // Enemy A.I. Surrender Fleeing Chance:
     // Choose the chance of how enemy A.I. which already was captured by a player will try fleeing.
     // 0 - 100, default is 5%
       // GL4_Global set [57, 5];

   // ==============================================================
   // GL4 Enemy A.I. Simulate Dead: ( New )
   // ==============================================================
   // Choose if enemy A.I. should be able to simulate dead if no friendly A.I. is nearby.
   // True / False, default is True
      GL4_Global set [58, False];

     // Enemy A.I. Simulate Dead Chance:
     // Choose the chance of how enemy A.I. should simulate dead.
     // 0 - 100, default is 15%
       // GL4_Global set [59, 15];

     // Enemy A.I. Simulate Dead Friendly Nearby:
     // Choose how much friendly A.I. units have to be nearby that the unit will not simulate dead.
     // Note: If there are more friendly A.I. units nearby like set here then the unit will not simulate dead.
     // 0 - 100, default is 5
       // GL4_Global set [60, 5];

Edited July 30, 2013 by Masharra
inserting code

[Varanon 892]

So, if I fire from, say, 300m away, AI's guess of my x, y position is based on a default guess error (of 100m) plus my distance (300m) multiplied by a coefficient (0.25), plus a random value (between 0 and 1, applied to x and y).

This individual calculation doesn't currently factor in my weapon type, but my visibility remains a product of that, so - if I've used, say, an enormous sniper rifle instead of an smg - the AI is more likely to spot me, when looking in the right place.

I wonder why this is only done when one of the team was killed. Wouldn't the same rules apply, regardless of whether they hear a shot, or one of them gets killed ?

And isn't that the reason for this problem ? They are made aware of the player when someone is killed, regardless of whether they actually detected the shot or not ?

I'm sorry, but it does not make sense to base knowledge of the killer based on the fact that someone in the group was killed. The only sensible thing to do is base the information about the shooter on the fact that they heard the shot/saw the muzzle flash/had the ability to somwhow detect the shot.

How is this handled when a member of the group is killed through a satchel charge ? Same information about the killer ?

[Sniperwolf572 758]

So, if I fire from, say, 300m away, AI's guess of my x, y position is based on a default guess error (of 100m) plus my distance (300m) multiplied by a coefficient (0.25), plus a random value (between 0 and 1, applied to x and y).

Can we get some parentheses or a more specific formula? :D

Going by what RiE wrote, it seems a bit strange. I'm assuming I came up with a wrong formula, because the results below would mean that shooting a weapon on top of AI would end up them having a minimum of 25 meter error in the best case of shooting right on top of the AI.

(100 + 0) * 0.25 = (25 + 0.5)
(100 + 300) * 0.25 = (100 + 0.5)
(100 + 400) * 0.25 = (125 + 0.5)
(100 + 500) * 0.25 = (125 + 0.5)
(100 + 1000) * 0.25 = (275 + 0.5)

100 + (0 * 0.25) = (100 + 0.5)
100 + (300 * 0.25) = (175 + 0.5)
100 + (400 * 0.25) = (200 + 0.5)
100 + (500 * 0.25) = (225 + 0.5)
100 + (1000 * 0.25) = (350 + 0.5)

[royaltyinexile 175]

Can we get some parentheses or a more specific formula? :D

Indeed, I hoped to offer a general description of the behaviour, rather than a precise formula. That being said, the random value is a multiplier applied to both x and y coordinates.

I could dig out a more precise formula, but feedback on the behaviour in game will generally will be more useful to our programmers, who should better know how best to adjust values based on feedback to actual situations.

I wonder why this is only done when one of the team was killed.

Apologies, this also applies to when they're hit or bullets impact close to AI Units. Community reports mostly were dealing with kill repros. Updated Changelog info.

How is this handled when a member of the group is killed through a satchel charge ? Same information about the killer ?

No, this is handled in a similar way to the AI discovering a dead body.

As they have 0 visibility of shooting, AI takes random position around explosion spotter (they will check surroundings, default error used is 200m)

Best,

RiE

Edited July 30, 2013 by RoyaltyinExile

[Variable 322]

Apologies, this also applies to when they're hit or bullets impact close to AI Units. Community reports mostly were dealing with kill repros. Updated Changelog info.

Thanks for clarifying this. What about bullets passing overhead? And audible shots fired?

[benedykt 10]

BIS are you going to change increased error of AI crouching and crawling all the time while in combat? I can't stand that. Playing this game is not fun for me anymore because of current AI.

[Alwarren 2767]

Apologies, this also applies to when they're hit or bullets impact close to AI Units. Community reports mostly were dealing with kill repros. Updated Changelog info.

How does proximity and communication assets influence this? My ticket that Varanon mentioned has the AI on opposite sides of the islands with radios removed, and it still seems to know about the incoming fire. Since this is very disruptive towards stealth missions, would it be possible to limit this information exchange by proximity and availability of communication? I.e. being in earshot range and/or having a radio?

(Speaking of communication, a big "Thank you" for the increased volume of communication from you devs. It's really appreciated).

[royaltyinexile 175]

BIS are you going to change increased error of AI crouching and crawling all the time while in combat? I can't stand that. Playing this game is not fun for me anymore because of current AI.

There was an initial change that increased chance of prone and crouch generally, but then an update which modified this behaviour so that they're mainly selecting prone when in stealth mode.

I'll dig out some revisions via Dr Hladik, but the current goal/target state is to have AI taking crouch/prone while stationary and under fire (so that they are more difficult to hit), but not doing this when moving to cover or given a movement waypoint (so that they remain mobile).

Notice how the current really doesn't take into account how close the enemy is (from my observations)

As Coulum correctly points out, the AI do not currently take distance of enemy into account when selecting stances when moving/moving to cover. However, it should be possible to change this behaviour, and tweak the values so that when enemies are closer, e.g. CQB, they are less likely to go prone.

Dr. Hladik is a bit busy fixing some crash opportunities atm, but he'll take another look at the state when he gets a chance.

Best,

RiE

Edited July 30, 2013 by RoyaltyinExile

[Zorg_DK 10]

I just played a mission with heavy fog, I thought the AI was a little too good at spotting, and start firing at me.

I also think the ai accuracy should be decreased, longer lasting firefights are more fun imo. The Ai should be quick to start firing though.

[petek 62]

There was an initial change that increased chance of prone and crouch generally, but then an update which modified this behaviour so that they're mainly selecting prone when in stealth mode.

I'll dig out some revisions via Dr Hladik, but the current goal/target state is to have AI taking crouch/prone while stationary and under fire (so that they are more difficult to hit), but not doing this when moving to cover or given a movement waypoint (so that they remain mobile).

As Coulum correctly points out, the AI do not currently take distance of enemy into account when selecting stances when moving/moving to cover. However, it should be possible to change this behaviour, and tweak the values so that when enemies are closer, e.g. CQB, they are less likely to go prone.

Dr. Hladik is a bit busy fixing some crash opportunities atm, but he'll take another look at the state when he gets a chance.

Best,

RiE

Thanks for the update RiE - look forward to a fix. I'm seriously finding the Dev branch AI so "un-enjoyable" at the mo that I'm considering switching to the stable build for the first time since release!

"I'll dig out some revisions via Dr Hladik, but the current goal/target state is to have AI taking crouch/prone while stationary and under fire (so that they are more difficult to hit), but not doing this when moving to cover or given a movement waypoint (so that they remain mobile)." Sounds like what we need. I keep seeing AI crawling in the open while taking fire rather than moving to take cover

[Tyl3r99 41]

There was an initial change that increased chance of prone and crouch generally, but then an update which modified this behaviour so that they're mainly selecting prone when in stealth mode.

I'll dig out some revisions via Dr Hladik, but the current goal/target state is to have AI taking crouch/prone while stationary and under fire (so that they are more difficult to hit), but not doing this when moving to cover or given a movement waypoint (so that they remain mobile).

As Coulum correctly points out, the AI do not currently take distance of enemy into account when selecting stances when moving/moving to cover. However, it should be possible to change this behaviour, and tweak the values so that when enemies are closer, e.g. CQB, they are less likely to go prone.

Dr. Hladik is a bit busy fixing some crash opportunities atm, but he'll take another look at the state when he gets a chance.

Best,

RiE

i produced x2 tickets regarding the combat/stealth mode so im guessing these will be resolved shortly?

11998 danger

12003 stealth

[runekn 3]

Dr. Hladik is a bit busy fixing some crash opportunities atm, but he'll take another look at the state when he gets a chance.

Well tell him to hurry up. Reading the lastest SITREP has made me nervous about the data lock ;)

[metalcraze 290]

RiE was this fix from 1.63 beta added to A3?

Since I seem to be noticing the similar pre-fix behavior of AT soldiers where they run point blank to tanks and other soldiers in general try to close in point blank to enemy at times

The ticket with repro is here: http://feedback.arma3.com/view.php?id=12029

This was a very important change in 1.63 making AI many times more survivable and since the release is almost imminent this is a rather big issue

[fabrizio_t 58]

I think it's this done wrong, the turn speed isn't fixed but they're working on AI behavior. When turn speed will be fixed it will appear a different AI. The micro management should be done first, then the general behavior. Priorities would be:

1. Turn speed. You are still able to follow an AI who can't shoot at you.

2. Flee conditions. Sometimes AI refuses to shoot at you, sometimes even running away. Is it flee? If so, make it drop the weapon or an explicit surrender animation. Else make it shoot.

3. Gunshot/sounds detection - they don't care if you shoot behind them, they don't turn around to check sometimes. If you shoot after corner they get past the corner not looking.

4. Cover detection - related to enemy direction.

When these will be fixed, a lot of the waypoints behavior will fix itself.

I completely agree with your points.

Turn speed is indeed a key problem.

Let me add that fleeing mechanics are just completely broken, for example:

* Whenever a single unit within a group starts fleeing, the entire group instantly flees as well;

* Follow-up to the point above: groups try fleeing while keeping formation ... which is absurd.

* Fleeing behaviour is useless: sometimes units lock in place ignoring any enemy which is not exactly in front of them. Other times they (slowly) wander around while standing up (no use of combat stances). Either way ... that's not fleeing.

Finally, not only the game deserves some better fleeing routines, it also needs tactical withdrawal routines and possibly surrender routines.

Edited July 30, 2013 by fabrizio_T