AI Discussion (dev branch)

[roshnak 41]

Not sure this is a very well thought statement even if meant in jest. I will not going to start a debate about what zombies should be like, lets just all agree that "smart" zombies present a challenge and make for better experience

I think you're fully misreading the intention of that statement. It was almost certainly intended to mean "DayZ's AI doesn't have to work as hard as Arma 3's AI does," not "zombies should be dumb."

[.kju 3245]

How about disabling lean until the more basic problems get sorted?

As alternative the lean action should at least take the current target of the AI in mind.

The more advanced approach would be to check if lean makes sense in the first place,

and if other known threats are present in other directions.

But as said while it creates a bad impression seeing it, personally it seems to me a minor issue,

with other far more important ones to be focused on first.

[Variable 322]

Oukej, I have the utmost respect to you guys and I am certain that you are at least frustrated from the state of AI as we are. My criticism is directed to the decision making echelon at bis that is in charge of allocating overwhelming and sufficient resources in order to fix the AI and lift them to reasonable standards.

Apart from teaching the AI all the new functionality (slingloading, firing from vehicles... ref. roadmap)

That's exactly my point, BIS is only treating the AI when it must - in order to support new DLC content. As a customer, I would expect treating the AI first and only then add more features and content that the AI needs to handle. That would be the fair thing to do. It might not be the most financially beneficial to BIS (I guess announcing an effort to fix a very significant part of the game doesn't bring new customers, although might return old ones) but sometimes a business need to acknowledge that its product is in severe state. I work as a software product manager, and my product is now releasing Service Pack number 2 for the latest version, because some things cannot pass. Arma 3 needs a service pack for the AI now. Not new DLC version with sling loading.

the focus so far has been mainly on fixing issues. Most recently it's been low level AI optimizations and the pathfinding in buildings (which had been broken for a long time in buildings without the floor level). There are more improvements coming and some "new" features are in the process or under investigation. (I know, I know - and you've hinted it - some of them have been an industry standard and "should have been there from the start").

that's all very promising and still I don't hear the AI revolution bells we should be hearing. We are not talking about too much really - Get the AI moving when ordered to, make them move straight, teach them to drive properly and suppress and be suppressed.

But to put into manpower perspective - it's been only relatively recently that we were able to expand the team and even now our programming dept. is about 3 times smaller than the one from OFP-DR.

And that is exactly why it's so distrusting. Weighing the effort that was put into so much stuff since Arma 2 (many DLCs, Arma 3, more DLCs) it is clear that even if some of that effort would have been put into the AI, we would have been in a completely different place now.

AIs are hilarious! ... Ok fine no AT weapons, but what about tyres? Chuck a grenade maybe?

Sorry, but given the current state of the AI, expecting them to shoot-out tires of vehicles when not equipped with AT is like expecting a 2 year old to build a spaceship. In the list of the Arma 3 to-do list, it should be somewhere deep down in task number 1643. Let's have them walk straight first...

[roshnak 41]

That's exactly my point, BIS is only treating the AI when it must - in order to support new DLC content. As a customer, I would expect treating the AI first and only then add more features and content that the AI needs to handle. That would be the fair thing to do. It might not be the most financially beneficial to BIS (I guess announcing an effort to fix a very significant part of the game doesn't bring new customers, although might return old ones) but sometimes a business need to acknowledge that its product is in severe state. I work as a software product manager, and my product is now releasing Service Pack number 2 for the latest version, because some things cannot pass. Arma 3 needs a service pack for the AI now. Not new DLC version with sling loading.

First of all, you understand that those features are not really "DLC content," right? They are just updates to the core game.

Second, those features are some of the most requested features in the history of the series. People have been begging for the abiliity to fire from vehicles for nearly a decade now.

Third, BIS needs to have a continuing revenue stream in order to pay the developers of the game to work on things like AI. DLC does that.

We are not talking about too much really ... suppress and be suppressed.

That alone sounds like it would be enormously complicated to do properly, given the variety of unit sizes and types that may be involved in any specific engagement.

Don't get me wrong, it's a feature that I would love and think should be in the game, but I wouldn't say that it's "not talking about much."

[Variable 322]

First of all, you understand that those features are not really "DLC content," right? They are just updates to the core game.

Think bigger and don't over complicate what I'm saying. (a portion of) the money and time that was put into whatever BIS released since, say, 2008 could have been resourced differently. Sure, that might mean recruiting algorithm developers instead of artists. If you look closely into Arma 3 threads you can see even BIS personnel complaining silently about the lack of programing personnel to address issues that are very bothering them.

Second, those features are some of the most requested features in the history of the series. People have been begging for the abiliity to fire from vehicles for nearly a decade now.

The AI is so fundamental for so big portions of the community activities that even this cool feature (which I'm excited about) pales when compared to.

Third, BIS needs to have a continuing revenue stream in order to pay the developers of the game to work on things like AI. DLC does that.

Give me a break, they made a ton of money from Arma 2 sales peak due to DayZ and Arma 3 sales weren't bad. All the options in the world in terms of finance and time to fix such a core issue.

That alone sounds like it would be enormously complicated to do properly, given the variety of unit sizes and types that may be involved in any specific engagement.

A scripted mod (and a very successful one) has proven this claim to be false. If you want to speculate then it's more reasonable to assume that a core engine implementation would be even less complicated and less resource demanding.

--------

You are a customer. Don't try to speculate excuses for your vendor. Nothing that has been suggested here is out of any reasonable scope. It's only a matter of management decision.

[.kju 3245]

@ roshnak

BI could just add new features without AI support, and focus on the AI basics instead.

One can have different views what should get priority, but there are different approaches possibles.

[2nd ranger 282]

;2774886']How about disabling lean until the more basic problems get sorted?

Currently, seeing the AI lean around cover is the only thing they do that makes me think they're halfway intelligent.

Also I think what this thread needs is video. Short clips of the AI being consistently and reproducably dumb. The last time someone posted a really embarrassing AI video, it resulted in the close quarters reaction/turning fix.

[alky_lee 279]

Lately we've been fixing bugs related to AI behavior in the cargo buildings and towers. Some of it was quite... interesting.

Does that mean that this sort of behaviour is a thing of the past? https://www.youtube.com/watch?v=FYQVTb0eRj0

[Alwarren 2767]

I'm replaying Operation Flashpoint: Dragon Rising again, and seeing how well the AI behaves there truly frustrates me in regards to the shameful state of Arma 3 AI.

A console game, no less-

With a better First Aid system too..

[Tonci87 163]

A console game, no less-

With a better First Aid system too..

Sadly this....

How much fun we had bashing that broken POS and now look at us.....

Regarding the Arma 3 AI. It is actually quite amazing. I can have hundreds of AI on massive islands, even custom ones, and it will still work. No other game can do that. That for me is part of Arma s beauty. However there are problems, quite serious ones too.

Infantry is relatively OK, especially if you use Mods.

They do need improvements.

A sense for danger would be nice. Ai simply sprinting out in the open during a firefight should not happen that often.

Tactical sense. Currently the AI always seems to attack move the enemy. They need to know how and when to defend.

Buildings. Well they simply need to get better while navigating through buildings.

In my mission yesterday I had set up a few AI inside a building and some of them actually did a decent job keeping players out of it, some others though......

The biggest AI problems show with vehicles.

AI Drivers are absolutely horrible. One car parked on the side of the road can stop a whole convoy and make it one giant clusterfuck of vehicles slamming into walls and each other. Drivers of armored vehicels that don´t turn the heavy armored front towards the enemy but the back. Crews that bail out into sure death if you shoot out tires or tracks while the Weapon is still operational. Helicopter pilots who can only do strafing runs (rule#1 of combat helicopters: Don´t fly over your enemy, break off before) and many more issues......

[f2k sel 164]

I can only conclude that AI isn't a priority and hasn't been for a very long time.

If it was possible to fix AI path finding and Driving I'm sure it would have been sorted years ago, attempts have been made but I don't see any real improvement. Usually it makes things worse.

[Variable 322]

If it was possible to fix AI path finding and Driving I'm sure it would have been sorted years ago

I don't think this sort of problems fall into the definition of impossible. It's only a question of resources and time and demand by the community. What certainly doesn't help is a community that have grown accustomed to the horrible state of the AI to a point that any objectivity has been lost towards it.

[das attorney 858]

Long post ahoy.

I'm pretty disappointed by the AI's use of the environment.

In terms of driving, even simple waypoints placed on road junctions seem to get them stuck on corners or they'll rip a tyre off as they go past an object.

For infantry, it seems there isn't enough information available for them to make good judgement on how to navigate buildings or use the various floors of buildings.

DayZ prebaked solution wouldn't work as user placed objects and vehicles driving around the place or abandoned wouldn't be accounted for. AI would run straight through them as they wouldn't be in the data baked into the map. Plus it's just for running in straight lines ala Dijkstra algorithm.

http://en.wikipedia.org/wiki/Dijkstra's_algorithm

However, more information should be in the building/object configs. At the moment, there is building positions and that's it. I've been working on a mod which was supposed to be for MANW, but I'm not going to make that date as the AI were a mess and I had to get involved changing a lot of things to get them to "think" smarter for what I want to do.

The AI can now use info in the building configs to find good places to go, and watch useful angles from the building. I'm working on a breaching/searching behaviour that improves room clearing and again it's all based on pre-baked info placed in the building configs. (Find the doors, stack up, search that floor, secure staircase, stack up and rinse and repeat for next level etc). I'm still working on the scripts but it's looking pretty good so far.

The price for this is lots of boring evenings spent in the editor (for me) and more CPU use scripting the AI behaviour. I'm surprised that this sort of stuff wasn't done during the games development though.

There should be a synergy between:

• Info in the building models
  
• Info in the building configs
  
• Scripts the AI use to process this info.
  

As for unit's in water, it's so hopelessly broken that it's not worth doing anything with divers. It's a shame as this was supposed to be a new interesting environment to play in for Arma 3, but looks half finished and abandoned.

Here is a ticket I submitted on 14 November 2013 regarding divers not being able to move in combat mode. Nothing's been done to address these sort of issues so it's remained worthless trying to use divers in missions.

http://feedback.arma3.com/view.php?id=16058

Also, because the Arma 3 dev team refuse to release any sort of roadmap for AI development, then it's possible that I'll log on to this forum one day to find they've added all this stuff. On one hand this would be great, but also annoying as the time I've spent trying to fix up the AI would be wasted.

Glad I didn't bother working on a logistics system given they've just added slinging from choppers which would mean a massive rewrite for any sling/logistics mods. (OK they did announce that, but haven't announced anything concrete for AI).

Plus, any increase in AI using the environment effectively means more CPU use so we're back to the problem of the game crapping out on FPS when the AI is doing computationally intensive stuff.

[bad benson 1733]

DayZ prebaked solution wouldn't work as user placed objects and vehicles driving around the place or abandoned wouldn't be accounted for. AI would run straight through them as they wouldn't be in the data baked into the map. Plus it's just for running in straight lines ala Dijkstra algorithm.

http://en.wikipedia.org/wiki/Dijkstra's_algorithm

that's too bad. so that is probably what oukej meant. i guess then this effort will be useless for arma. although i don't see why there can't be a combination. i mean are you basically predicting that dayZ zombies will run through barricades, cars, tents and anything player placed?

i'm also wondering if saying it's for running in straight lines is a bit simplistic. isn't it more a matter of what commands are issued to the unit? i mean let me put it like this. the zombie gets one position (player position) and then goes there using the nav mesh. the soldier gets several positions to flank around the player and uses the nav mesh to get to each of them one by one. isn't it just a new method to find paths quickly in general?

There should be a synergy between:

Info in the building models

Info in the building configs

Scripts the AI use to process this info.

yea. but it seems that BI rather have backwards compatibility than some AI breakthrough. sounds like an excuse to me. i mean peopel are literally making houses from arma 1/2 enterable due to mlods being available. saying it would break old or usemade content is highly underestimating the community's capabilities and enthusiam.

Exactly! And who says Cry can't handle scale with ease:

ok first of all. that is not cryengine.

second of all you are completely missing the point of what i said. how is showing AI fail in an engine that isn't made for that kind of engagements proving that using predefined data is not useful? pretty pathetic fanboy post, if you ask me.

Cryengine ain't the end all be all

it's sad that you can't even bring up things just to make a point without someone getting into faction war mode. you totally ignore the part where an actual dev just said that they are outnumbered by those teams. but yea you are right. we should just laugh at other engines and never even bring them up in a discussion because no matter what the point was, it's heretic and arma has the best AI ever. that's why we're talking about it in the first place, right? it's also funny how you are willing to accept stuff that is used in dayZ (which turns out to be useless?) but not any outside engine. let the discussion happen. don't be afraid :p

Edited September 14, 2014 by Bad Benson

[froggyluv 2136]

ok first of all. that is not cryengine.

Ok your right it's an offshoot of CE mea culpa -Ill look into large scale AI Crysis battles.

second of all you are completely missing the point of what i said. how is showing AI fail in an engine that isn't made for that kind of engagements proving that using predefined data is not useful? pretty pathetic fanboy post, if you ask me.

Actually your right in that I don't see predefined data as a negative, quite the contrary actually. I'd like to see every house, wall, rock, roof and bridge tied into the AI cover system but that in itself is not the answer. Many games feature cover nodes everywhere but the AI are still extremely easy to beat as they play a mindless game of pop up/down like a whack-a mole. RO2 is a good example of that, the AI are great at using indoors, setting up MG's in windows, jumping thru windows etc. but can't manage to stay alive for more than a few seconds.

Now look at your second point, "not made for that kind of engagement", what does that tell you? Arma was never built for indoor or any real CQB of any consequence as well. They're trying to build that layer of cake after the fact and thats not an easy thing to do. I mention other engine fails for that exact reason -what were they designed to do initially.

As for the "fanboy" label well now that should be beneath you now shouldn't it :rolleyes:

It's sad that you can't even bring up things just to make a point without someone getting into faction war mode. you totally ignore the part where an actual dev just said that they are outnumbered by those teams. but yea you are right. we should just laugh at other engines and never even bring them up in a discussion because no matter what the point was, it's heretic and arma has the best AI ever. that's why we're talking about it in the first place, right? it's also funny how you are willing to accept stuff that is used in dayZ (which turns out to be useless?) but not any outside engine. let the discussion happen. don't be afraid :p

Pretty sure if you took an office poll, BI would consider themselves a success over the buried OFP:DR legacy -matter fact I remember Maruk stating they were quite pleased ;)

When DR came out I bought first day and came on these forums to compliment the use of their's AI's ability to use indoors and windows. I've posted multiple AI troubleshooting/feature bugs that I feel are flawed or missing on the Dev tracker as well as have criticized the lack of AI indoor advancement so many times I've lost count. The difference is I see the strengths of these AI as well so no need to just bash them all the while praising the triumphs of other really shitty games just to rub it in. Trust me, if I could find other shooters whose AI could entertain me for more than an hour I'd be all over it.

[bad benson 1733]

Ok your right it's an offshoot of CE mea culpa

no it's not http://de.wikipedia.org/wiki/Far_Cry_3 it's a different engine entirely. and it's also very much besides the point. arma uses "nodes" inside houses too. it just uses them in a very limited way to mask the huge problems that AI units have when interacting with objects.

Ill look into large scale AI Crysis battles.

well please go do that to prove my point that you are not getting it at all...please keep wasting your time with finding videos, if it keeps you from posting.

Actually your right in that I don't see predefined data as a negative, quite the contrary actually.

then why do you have to post lines after lines of nonsense?

RO2 is a good example of that, the AI are great at using indoors, setting up MG's in windows, jumping thru windows etc. but can't manage to stay alive for more than a few seconds

that made me laugh. did it ever cross your mind that AI in other games in general don't survive as long because they don't aim like super humans and the engagement range is far shorter? arma's AI keep their distance and shoot very well and in addition the player has the exact amount of health as an AI unit.

time of survival says nothing about how smart they are. i don't know why you think comparing games like that makes any sense at all and i fail to see how that has any relevance in a discussion about technical methods.

Now look at your second point, "not made for that kind of engagement", what does that tell you?

what does any of that have to do with baked data? are you really that narrow minded that you can't get off the game vs game trail? this whole post of yours just proves my point. i mean the video you showed is from a game that has no focus whatsoever of AI engaging eachother. showing the lack of such AI says nothing about that game's pathfinding technology.

I mention other engine fails for that exact reason -what were they designed to do initially.

see? YOU brought up "cryengine" in that context. I brought up ACTUAL cryengine as there was a discussion in here some time about nodes (you were probably involved by the looks of it) where someone (probably you) said that "but they just use nodes". good to see that you seem to accept the usefulness of baked data now burried inside all the useless butthurt.

As for the "fanboy" label well now that should be beneath you now shouldn't it :rolleyes:

it's literally what you are. taking other people's statements out of context to serve your personal faction battle they have no part in. some people just embody that term. it's not even meant as an insult. it's a mere observation.

Pretty sure if you took an office poll, BI would consider themselves a success over the buried OFP:DR legacy -matter fact I remember Maruk stating they were quite pleased ;)

see? sad...people were actually describing things from that game that are well inside the scope of arma, which arma doesn't do and could use very well. but it's invalid info because it's OFP:DR, right? (i'm talking about the static gun example again). it's also invalid because Maruk said something somewhere. blablablabla. ugh...

it's a huge misconception many people have. when a dev says "but they have to be able to use vehicles in arma" then they don't mean that AI are smarter than in other games that don't have vehicles because of that, because news flash, arma's AI does NEVER use vehicles on its own, it only "can drive" them after you the player made them mount them via command or mission editing. it's solely a matter of AI devs having to spare resources for separate driving AI and does say nothing about actual intelligence of the AI in combat.

the reason why arma's AI is more entertaining is because they can flank and there is actual room for doing so. that's about it. i'd like someone to show me another example of arma's AI being smarter than just "go there shoot that". they don't even rearm for themselves whne they run out like in several arma 2 AI mods.

The difference is I see the strengths of these AI as well so no need to just bash them all the while praising the triumphs of other really shitty games just to rub it in.

this is what it comes down to. you assume that about me and probably about everyone who dares to merely mention other engines, the context is irrelevant to you. the rubbing in part is hella funny. this thread is about AI feedback you know? but i guess i will make sure to always add praise about things that aren't broken, ok? that won't be feedback or meaningful discussion but it might keep some people from going off.:rolleyes:

you are literally repeating the same old shit just because my post contained some of your personal trigger words. all you are achieving is poisoning the discussion. you would have a point, if people would say "just switch to cryengine to solve all your problems". it's not happening though so it's kind of all in your head mate...

Edited September 14, 2014 by Bad Benson

[Bouben 3]

AIs are hilarious! No seriously! I was testing something, so I spawned a buddy rifleman and an enemy Ifrit, unarmed though. Ordered my buddy to attack Ifrit, nothing. 2, Attack! Nothing, 2, Fire! Nothing. 2, Move there! 3 sec later "Enemy MRAP!". Really, I just ordered you to attack it you dumbass. So he refused to shoot. Ok fine no AT weapons, but what about tyres? Chuck a grenade maybe? Anyway, he goes to the Ifrit and starts circling it, while still refusing to engage. The other guy in Ifrit notices him standing in front of the vehicle, so he starts reversing, then turns around and starts to flee. It is like watching dumb and dumber, really.

OK, so the AI would shoot the tires off. But what next? The car will be still able to use its gun (and don't say that you want an AI gunner to leave the vehicle when its tires are destroyed - one of the most stupid behaviours currently in the series). The best thing the AI can do is to be completely stealthy when they have no efficient means of destroying a vehicle. You guys clearly have no idea how many variables there is. There is so much the AI must do even if you think it could be simplified.

[CaptainObvious 95]

OK, so the AI would shoot the tires off. But what next? The car will be still able to use its gun (and don't say that you want an AI gunner to leave the vehicle when its tires are destroyed - one of the most stupid behaviours currently in the series). The best thing the AI can do is to be completely stealthy when they have no efficient means of destroying a vehicle. You guys clearly have no idea how many variables there is. There is so much the AI must do even if you think it could be simplified.

While I agree that engaging MRAPs with small caliber weapons is stupid, the subordinate should still do what he is told to, be it stupid or not.

[f2k sel 164]

While I agree that engaging MRAPs with small caliber weapons is stupid, the subordinate should still do what he is told to, be it stupid or not.

I'd agree, when we tell them to do something just do it.

One thing I've noticed with AI driving.

AI will pass each other on the road (both moving towards each other) without an issue but if one vehicle stops it becomes extremely difficult for the moving vehicle to pass the stationary one.

When a vehicle is parked the moving AI vehicle needs a large clearance .

In a test I parked two Identical vehicles on opposite sides of the road to see how much gap a moving vehicle needs to pass between them.

moving vehicle Hunter AI

ATV <5m> ATV

Hunter <15m> Hunter

HEMTT Drive Unit <19m> HEMMT Drive Unit

I assume objects are measured center to center so in reality it's a bit less.

If a HEMTT is parked off to the side of the road an AI vehicle still needs 9.5meters clearance or it stops or tries to move off road.

In towns this is a disaster as there just isn't the room to move off road.

[roshnak 41]

Think bigger and don't over complicate what I'm saying. (a portion of) the money and time that was put into whatever BIS released since, say, 2008 could have been resourced differently. Sure, that might mean recruiting algorithm developers instead of artists. If you look closely into Arma 3 threads you can see even BIS personnel complaining silently about the lack of programing personnel to address issues that are very bothering them.

Okay, but those things didn't happen, and complaining about them now doesn't do any good. The Arma 3 dev team isn't going to get a huge influx of developers at this point.

I mean, if you really want to complain about the allocation of development resources, you should probably recognize that DayZ is pretty clearly the focus of development at this point, with a much larger team than Arma 3.

Give me a break, they made a ton of money from Arma 2 sales peak due to DayZ and Arma 3 sales weren't bad. All the options in the world in terms of finance and time to fix such a core issue.

Money doesn't last forever.

A scripted mod (and a very successful one) has proven this claim to be false. If you want to speculate then it's more reasonable to assume that a core engine implementation would be even less complicated and less resource demanding.

That mod only addresses the "to be suppressed" portion of your request to allow the AI to "suppress and be suppressed." It's much more difficult to teach the AI when and how to suppress other groups. There are a lot of potential factors involved.

You are a customer. Don't try to speculate excuses for your vendor. Nothing that has been suggested here is out of any reasonable scope. It's only a matter of management decision.

Not thinking BIS should put everything else on hold until they solve my pet issues doesn't make me an apologist.

;2774919']@ roshnak

BI could just add new features without AI support' date=' and focus on the AI basics instead.

One can have different views what should get priority, but there are different approaches possibles.[/quote']

I don't think I posted anything that would indicate that would indicate otherwise. I simply said that the DLC is likely what is funding the continued development of the game, up to and including AI fixes.

[CaptainObvious 95]

One thing I've noticed with AI driving.

AI will pass each other on the road (both moving towards each other) without an issue but if one vehicle stops it becomes extremely difficult for the moving vehicle to pass the stationary one.

When a vehicle is parked the moving AI vehicle needs a large clearance .

In a test I parked two Identical vehicles on opposite sides of the road to see how much gap a moving vehicle needs to pass between them.

moving vehicle Hunter AI

ATV <5m> ATV

Hunter <15m> Hunter

HEMTT Drive Unit <19m> HEMMT Drive Unit

I assume objects are measured center to center so in reality it's a bit less.

If a HEMTT is parked off to the side of the road an AI vehicle still needs 9.5meters clearance or it stops or tries to move off road.

In towns this is a disaster as there just isn't the room to move off road.

Now that you mentioned it, I noticed this too just the other day when I recorded the video to demonstrate the doppler effect in the Audio Tweaking -thread, the sport hatchbacks didn't even slow down when they overtook the Offroads at considerable speed, interesting stuff.

[bad benson 1733]

OK, so the AI would shoot the tires off. But what next? The car will be still able to use its gun (and don't say that you want an AI gunner to leave the vehicle when its tires are destroyed - one of the most stupid behaviours currently in the series). The best thing the AI can do is to be completely stealthy when they have no efficient means of destroying a vehicle. You guys clearly have no idea how many variables there is..

i think KK's point was more about what actually happens in arma, which is the AI circling around the vehicle doing fuck all apparently. i mean, fair point about the tires and stuff but i'd have them rather try something meaningful or say "can't do that" than do something totally retarded.

the thing is, you say there are so many variables and on the other hand you immediately present a solution to the problem. going stealth or depending on distance or if they are being detected flat out running for their lives. it's something the series hasn't been doing well for ages eventhough it's supposed to be a simulation of military combat. reacting to a very obvious threat they can't handle.

there are not too many variables. the type of enemy/vehicle an Ai can encounter and the way to react to that is pretty limited and predictable by the devs. is the target armored? do i have AT? done.

the AI currently acts like there is no variable for evaluating their target at all. and even if this was a case of too much thinking with the result of disaster, is that what we want? something that doesn't work but the intend behind it was good?

There is so much the AI must do even if you think it could be simplified.

what are these things? i mean people keep saying that but what outstanding things does arma's Ai do that keeps it from performing simple life preserving measures when confronted with a superior enemy?

Edited September 14, 2014 by Bad Benson

[killzone_kid 1331]

While I agree that engaging MRAPs with small caliber weapons is stupid, the subordinate should still do what he is told to, be it stupid or not.

This. AI should do what it is told. If it is ordered to jump it should jump. There could be ambiguos commands like Engage This Enemy, so that AI can do own routine, and direct commands, like Fire At This Enemy, where it should point and shoot no questions asked. For once it could solve some driving issues. When I command it to reverse it should shift to reverse and press gas until I say stop. It is very simple. All AI brain needs to do is to know when to think and when to comply. It is not a democracy.

[Bouben 3]

i think KK's point was more about what actually happens in arma, which is the AI circling around the vehicle doing fuck all apparently. i mean, fair point about the tires and stuff but i'd have them rather try something meaningful or say "can't do that" than do something totally retarded.

the thing is, you say there are so many variables and on the other hand you immediately present a solution to the problem. going stealth or depending on distance or if they are being detected flat out running for their lives. it's something the series hasn't been doing well for ages eventhough it's supposed to be a simulation of military combat. reacting to a very obvious threat they can't handle.

there are not too many variables. the type of enemy/vehicle an Ai can encounter and the way to react to that is pretty limited and predictable by the devs. is the target armored? do i have AT? done.

the AI currently acts like there is no variable for evaluating their target at all. and even if this was a case of too much thinking with the result of disaster, is that what we want? something that doesn't work but the intend behind it was good?

what are these things? i mean people keep saying that but what outstanding things does arma's Ai do that keeps it from performing simple life preserving measures when confronted with a superior enemy?

Mate, going stealth is one thing if there is only the enemy vehicle present. Easy (if you don't consider the going-stealth procedure itself). Now what if there is an enemy vehicle with infantry? What if there is an enemy vehicle with infantry and another enemy infantry coming from behind? What if the enemy vehicle is out of LOS for 15 seconds and the enemy infantry is still chasing you? Should they now engage the infantry or not? Etc. There is so many things to consider and it is very difficult to do basic priorities with AI if there is too many variables going on. Just think about it for few minutes. I am sure you will soon or later find some situation that would not be so clearly solvable.

---------- Post added at 22:07 ---------- Previous post was at 22:04 ----------

While I agree that engaging MRAPs with small caliber weapons is stupid, the subordinate should still do what he is told to, be it stupid or not.

I definitely agree with this. But that was not the point I was talking about.

[roshnak 41]

This. AI should do what it is told. If it is ordered to jump it should jump. There could be ambiguos commands like Engage This Enemy, so that AI can do own routine, and direct commands, like Fire At This Enemy, where it should point and shoot no questions asked. For once it could solve some driving issues. When I command it to reverse it should shift to reverse and press gas until I say stop. It is very simple. All AI brain needs to do is to know when to think and when to comply. It is not a democracy.

I understand where you are coming from with this statement, but how do you want the unit to comply with your orders? If you tell a rifleman AI to engage an MRAP, should it just start shooting the vehicle's center of mass? Should it be smart enough to try to engage weak points like tires? Should it try to use grenades if it has them? What if you tell it to engage a tank? There is nothing the AI can do to damage a tank, should it still try to engage weak points? Waste grenades on it? What if you tell the rifleman to engage a jet? Should he chase it down trying to shoot it with his rifle?

I'm just saying that it's not quite as simple as, "Do what you're told. Always."