AI Discussion (dev branch)

[-Coulum- 35]

Regarding the discussion about how granular AI options need to be I would like to drop my humble opinion:

1. I think all tactical capabilities of the AI should always be at the best level the engine can provide.

This covers things like caring for their own lives / suppresssion, choosing the right weapon for a situation, flanking, handling of formations, searching in the right places, taking cover, driving without crashing, rearming, applying first aid to oneself and ones comrades, ...

This aspect is, which can make AI appear human-like, if it is implemented reasonably, but which also has the potential to totally ruin gaming experience and break immersion if done poorly.

So I would always use the best the engine can give us, and never dumb down AI by disabling or crippling capabilities.

2. It would be completely sufficient to tune AI shooting accuracy and reaction speed for adjusting game difficulty to ones personal liking.

I know about discussions like specific needs of mission designers for scenarios like "Super Human Special Forces vs. a bunch of retarded civilians with fire arms", but I think these can really be dealt with by 2.

IRL even completely untrained and stupid civilians would behave more cleverly than the current level of gaming AI. And this statement is not meant to blame BI. I think the ArmA AI is at least one of the best we can currently see on the market. AI is simply the biggest challenge in designing computer games.

For resembling different training levels and shooting experience it should really be sufficient to simply tune down shooting accuracy (and maybe reaction and detection speed).

So, what is your opinion?

Think that the smarts should be configurable to. Of course right now a 5 year old is probably smarter than the ai but if hypothetically BIS ever got ai to a reasonable "intelligence" level I would want to be able to tweak that depending on the soldier, in addition to accuracy and reaction time (which should be seperate sliders). A spotting speed and distance slider would be cool as well.

Thanks for the new changes devs, good to see the ai getting some love again.

[hbk 0]

1. I think all tactical capabilities of the AI should always be at the best level the engine can provide.

2. It would be completely sufficient to tune AI shooting accuracy and reaction speed for adjusting game difficulty to ones personal liking.

This would seem reasonable but I think the real problem is the underlying complexity of having multiple layers controlling the AI and how unclear they interact with each other.

Even without modding stuff the idea of having a global slider which overrides the AI skill settings set by the mission designers is pretty hard to grasp, especially in how you want to tailor your experience and how you take mission designs into accounts like "how hard was this mission supposed to be?"

Right now I'd day that regardless of future improvements in that regard I think we would greatly benefit from having at least two sliders which would basically cover the two points you're raising: "tactical behavior" and "weapon skill".

Edit: On a side note, I'm one of those guys who where traumatized by AI AK47 headshots at 500m in OFP :p

I mean, I could barely hit them with my sniper rifle (was it the M24?).

Edited October 9, 2013 by HBK

[oldy41 61]

Right now I'd day that regardless of future improvements in that regard I think we would greatly benefit from having at least two sliders which would basically cover the two points you're raising: "tactical behavior" and "weapon skill".

I would really be fine with that.

Of course having a lot of well-documented settings for the different aspects of AI behaviour would allow for endless fiddling and fine tuning, and mods like Zeus AI would really profit greatly.

But then as you stated, the underlying complexity of AI implementation and the ongoing improvements and changes applied by BI make this a moving target, and in particular the documentation would probably need frequent updates, too.

[turin turambar 0]

I don't agree in that, removing player choice in the options, in to choosing how to play the game is not a good idea.

*First, it's not like most of the "noob" players actually edit the game difficulties and change the AI skill values. Only the people who know what they are doing touch the values, usally. They are hidden from the casual user, you have to press "edit" in the part of game difficulties in options, or editing the .cfg.

*Second, said values about game difficulty should not overwrite the AI values from the mission, but both should work together. Playing all these variants :

-An easy mission in easy mode

-An easy mission in normal mode

-A normal mission in hard mode

-A normal mission in easy mode

-A normal mission in normal mode

should be different experiences. For that, maybe the AI skill setting value should go from 0 to 2 and the value of the mission designer is added to the user value. Or maybe do an average between both numbers, whatever method. What I'm trying to saw, if a mission designer put an enemy soldier with a 0.3 AI value and other with a 0.8 value, then the user should notice one is better than the other, AND that still have be the case playing in different difficulties

*Third, removing this fine-grained control is bad because they are sometimes used to fix the work of terrible mission designers and rescue an otherwise mediocre mission. Stand down of the pedestal, devs and modders can also do a bad job, you should take in account that.

*Fourth, from a more general videogame design point of view, it's the user the one who should have final control of how difficult is the game. He is the one who paid for it.

*Fith, if anything, I would like to have control of how good are the AI "senses" (vision, hearing). For that, apart from AI accuracy, I would decouple AI skill into AI skill and AI senses.

*Sixth, after saying all that, I have to confess I want the pure AI (how good is the AI flanking, giving orders, using cover, choosing targets, etc) to be the best possible, so control in that can be taken away. But still leave control for the rest (accuracy, senses, etc).

[.kju 3239]

... giving us a more predictable test-bed and mission makers a more meaningful (and better described) set of configuration options ... but the thought is that it will be, in the long run, beneficial to finally lock down these fundamental aspects of AI so that the ongoing tweaks and refinements can be implemented and tested with more confidence ...

Maybe I get a wrong impression here, yet to me this sounds like the wrong goal and aim.

Sure more descriptive settings and public knowledge about AI settings is very desirable.

However OFP/Arma has never been about a closely controlled experience presented to the player (like the CoD'esque shooters go for).

Instead the series and gameplay has always been been about an open world experience with freedom and choice.

The overall goal should be to make AI as close as possible to player behavior (and to a reasonable degree show military behavior).

As a player you want the AI behave, act and in the diverse skill range of players - as you can assess from yourself and other players what the AI can do.

So in general I agree with the notion from oldy41 there.

While the current skill and accuracy settings could be made more descriptive, this is what you want to tweak as a player (mainly).

Skill could be reduced to AI reaction time probably though. Yet they should be made split settings again.

Also a lot of difficulty settings are missing:

# Add additional server filter: difficulty

# Make it possible to change the default difficulty for a DS

# The create game dialog does not show default difficulty on a dedicated server

# Add option for admin to change difficulty settings for the selected mission

# Add option for admin to change the AI skill settings for the selected mission

In addition the community always went with their AI tweaks for larger engagement ranges and higher weapon use/fire rate, yet considerable lower accuracy (plus quicker reaction and hardly instant kills aside from headshots) - and had great success with that approach given the player feedback at all skill ranges.

For config tweaking BI should take a look at the excellent community tweaks like ASR AI or Zeus AI or SLX.

These core issues aside the other main ones modders cannot tweak are the problems with AI spotting abilities (engine and model issues) and "lack" of AI being affected in their targeting abilities (stance change, being hit, target position prediction, firing through concealment) as well as the very simplified AI info share within groups.

Some of the key issues in the FT:

# BIS_fnc_spawnGroup creates AI with skill 1 by default

# createUnit (array) creates AI with skill 1 by default

# AI can accurately engage through near concealment when it has spotted an enemy which then moves out of LOS

# AI Can still see through Grass

# AI has 360 degree situational awareness and fires before fully turned around./ AI is too accurate,etc..

# AI can see and shoot right through objects that completely obscure them

# AI Still see's player through dense vegetation.

# AI knows about all units of the same side in the direction it is facing (within spotting distance)

# AI can spot you too easily

# AI make LoS detection check on sound source before target is inside FoV (Aka, eyes in back of head)

# AI ignores gunfire for several seconds

# AI should be able to guess direction of gunfire to within 180 degrees

# AI is alerted immediately when a unit it knowsabout is killed

# AI won't attack anything beyond 400 meters

# An AI unit refuses to engage a distant target when ordered to or runs far away

# AI Rate of Fire changes with distance from player

# AI fire at a robotically constant rate of fire (exactly 0.7 seconds between shots) on single shot mode.

# Grouped AI is aware of kills even when they shouldn't notice

# Arma engine doesn't allow you to take out AI units silently if they are grouped to other AI units.

# Killing a soldier reveals the exact killer's position to the rest of AI squad

# No communication between differents AI groups

The CIT also has many good tickets - specific reports with observed vs expected, repro steps and demo mission:

# AI knows about having done a kill instantly

# Improve perceived position design for AI infantry

# Improve AI info share system and AI weapon use on non visible targets system

# Make AI able to recognize a target better when looking at it for a longer time.

# AI does not compensate high weapon sway level

# AI infantry does not compensate (enough) the weapon recoil

# AI infantry in combat mode never sprints (to cover for example)

# AI immediately tries to get into cover when first set to combat behavior

# Make AI infantry to kneel or lay down when exhausted when attempting to return fire effectively

# While AI infantry is moving from cover to cover, it hardly uses its weapon / has a very slow rate of fire.

# AI fixates on perceived source of gunshot, does not adjust based on new information

# Vehicle AI gunners are unable to identify infantry as hostile beyond ~250meters

# AI can even hear slow movement

# AI in vehicles detect enemy vehicles too easily

# AI in a car is deaf - cannot hear an unit to approach it (from the side or behind).

# Make Danger mode overridable and apply to fireteams only

# AI under player command knows about hostile unit ammo state and doesn't engage automatically if target has no ammo

# AI should use only one guided missile per target / do some basic usage evalutation

# Cars/trucks not sticking to roads in combat mode

# Make Tanks drive on roads in Aware mode

# Remove the tank driver's AI collision avoidance when given a move command by the tank commander

More..

With AiA probably even possible to run those within A3 easily.

Edited October 9, 2013 by .kju [PvPscene]

[royaltyinexile 175]

They are hidden from the casual user, you have to press "edit" in the part of game difficulties in options

Actually, they're right there in the 'difficulty' tab of the 'Game' configuration. Right up in any players face that wants to to anything to difficulty. Another option would be having an 'advanced' sub-section underneath the main interface, but more nested menus isn't terribly desirable.

*Second, said values about game difficulty should not overwrite the AI values from the mission, but both should work together.

This is true, but often not the case in reality. Perhaps providing 'advanced' option of some basic params from player point of view could be useful to experienced players (e.g. individual control over accuracy); however, well designed presets should make this requirement less essential.

*Third, removing this fine-grained control is bad because they are sometimes used to fix the work of terrible mission designers and rescue an otherwise mediocre mission. Stand down of the pedestal, devs and modders can also do a bad job, you should take in account that.

Ok, true, we would need to definitely be careful about the 'difference' between the set levels. It would have to feel meaningfully different to be useful.

*Fourth, from a more general videogame design point of view, it's the user the one who should have final control of how difficult is the game. He is the one who paid for it.

I somewhat disagree here, I think it's up to the mission designer to tweak and curate the settings (so, in editor, you have more meaningful access to AI params, but from pbo'd missions, you have less impact). Bad misson? Provide feedback, let the designer update it if he can/cares. It is a more brutal approach, though. The player should definitely retain a certain degree of 'predictable' control, but I don't think he should be able to, essentially, break the AI.

fith, if anything, I would like to have control of how good are the AI "senses" (vision, hearing). For that, apart from AI accuracy, I would decouple AI skill into AI skill and AI senses.

I agree here. Providing access to some simple, understandable controls like spotting ability, accuracy, seems useful. The question is whether is enough to roll this into the player-end AI difficulty preset or to provide individual controls. From a stability POV the presets are more valuable to mission makers.

Overall, the idea is not to take away all control from the player end, but deliver simple, predictable control; rather than the one abstract 'skill' slider which can wildly change the experience. If we went down the road of difficulty presets (e.g. easy normal hard), curating the differences between each level would be important, but difficult (and subjective) work. However, in consultation with the UI element of this task force, exploring 'advanced' AI configuration seems reasonable.

;2530354']Maybe I get a wrong impression here' date=' yet to me this sounds like the [i']wrong goal and aim[/i].

The first aim is to explore and define how AI individual AI config params affect AI behaviour, document and refine these settings. I simply don't see why you would take the view that this is an incorrect approach.

The second goal is to provide the average user more meaningful control over the difficulty of the AI, so that hard difficulties provide a more 'true to life' challenge, but lower difficulty settings enable players more lee-way while learning the game, but not at the cost of giving the AI brain damage (which is what, effectively, the current skill slider does). It's a blunt, unintuitive instrument, which most users simply won't know how to use properly if at all. Again, I simply don't see how you would take the view giving the majority of player meaningful control is a wrong goal or aim.

Let me be clear, this is not the be all and end all for 'improving' the AI - much more work needs to be invested in different areas, all those tickets, etc that you've listed, etc. That is different work but by no means less valid. This work is aimed at cleaning up wonky legacy design, giving mission makers and player better options, and reducing the likelihood of the average user breaking the game.

Best,

RiE

Edited October 9, 2013 by RoyaltyinExile

[Rick E 11]

IRL even completely untrained and stupid civilians would behave more cleverly than the current level of gaming AI.

I stood up and applauded.

This is especially true of AI squad mates. For instance, you can't tell them to take cover. Sure, you can point them to a wall, but inevitably, one or two will go to the other side, exposing themselves to the enemy. When I'm running towards an objective, they saunter along like it's a g.d. Sunday stroll in the park.

I'm used to squad-based game like Rainbow Six and Company of Heroes. Tell your squad to take cover, they will. Tell them to hold back until you tell them to assault, they will. For instance, in RS6, I can position different members at different doors. Then, on cue, we all breach at once, getting the bad guys in a crossfire. I can split CoH squad members to approach and attack from different directions

Another example. I was wounded. I stood right in front of the squad Medic; I selected him with F7 (that was his key); selected "Injured" from the context menu. He made no move to fix me. I tried about a dozen times. My icon was clearly marked that I was injured. In frustration, I shot him in the head to steal his First Aid kits.

I think the ArmA AI is at least one of the best we can currently see on the market.

Disagree. BF3 and Rainbow Six are better. AI in Company of Heroes, Total War series, and World In Conflict (all RTS, I know) is better.

[Smurf 12]

That is you not knowing how to control the squad, which could be streamlined a lot, but that is another problem.

And I sincerely hope that the R6 you are talking about isn't the Vegas one.

[Shagulon 1]

Perhaps there is a middle ground. You could leave or even increase the AI settings available to the player, perhaps in an advanced AI setting sub-menu. However, for the sake of people testing and experiencing the game from a pre-defined collection of settings (easy, medium, hard) we have a simple interface which allows these settings, but behind the scenes if one of these is selected, this then sets all the advanced settings to the pre-defined levels. That way, when helping test AI changes, devs can simply request you set difficulty to e.g. hard level, and we're all testing on the same levels. It also allows advanced users to tweak as they please.

It may also be a great idea for usability, to allow a custom level, into which the user can save their preferred settings, so that switching from these to a different pre-defined level isn't such a ball ache :)

As to how this could be worked into mission designers AI settings vs local users settings - I'm not sure.

[gammadust 12]

Ai settings, 2 additional points...

Why the apparently unwarranted necessity to set difficulty to absurd precision values:

// Recruit
	skillFriendly=0.64999998;
	skillEnemy=0.40000001;
	precisionFriendly=0.36999995;
	precisionEnemy=0.1;

// Regular
	skillFriendly=0.75;
	skillEnemy=0.60000002;
	precisionFriendly=0.54999989;
	precisionEnemy=0.28000003;

This is simply absurd, one even gets wary of changing those values (will somehow AI start being stupid if i set skillEnemy=0.6 instead of 0.60000002). Is this just a mere consequence of UI controls usage? Why not round those values, why not let the user step them (definite increments)...

Diverting the argument a bit but still relevant, the same is true for other unrelated to AI settings. It shouldn't be so. Those UI bars should adopt common user-friendly design practices (stepping), specially since UI bars with that functionality do exist already. An UI bar arrow that requires the user to click 1000 times to span it's length is simply no no in UI land. In addition many times the actual value is not shown, when it should.

---

Mission makers should have script access to values as defined by user. Ie. if user sets general AI skill to 0.8, mission maker should be able to read that value and adopt it in mission. This is an alternative way to connect user intent for difficulty and mission maker intents of difficulty and balance.

Edited October 9, 2013 by gammadust

[instagoat 133]

Is it possible to consider enabling the AI to recover mission critical items within their group? For example, if the AT dies, that someone on his team retrieves his weapon? Ditto for backpacks. I would advocate not having the AI to randomly gather equipment from outside their group for starters, and get a solid system for recovering in-group equipment first, if that makes things easier? Currently the presence of enemy vehicles demands a living AT gunner, which often is hard to do because said AT gunner is dead, but his weapon is still nearby and accessible, but currently will not be recovered for use.

[ProfTournesol 956]

I'm very happy each time i see the team working on the AI :)

[.kju 3239]

The first aim is to explore and define how AI individual AI config params affect AI behaviour, document and refine these settings. I simply don't see why you would take the view that this is an incorrect approach.

As I said: "Sure more descriptive settings and public knowledge about AI settings is very desirable"

There is a lot to learn from Suma's comments in the CIT explaining how the AI works.

You may not be aware of it, yet there has been a very close cooperation and joint effort between Suma, the CIT managers and the CIT users during A2 and OA development to improve the AI in very specific ways.

From this people have gotten a very good understanding how the AI works and therefore have been able to make impressive AI tweaks or scripted AI behavior.

And again of course the documentation is still lacking and it should be very much expanded.

Yet the goal should be in no way to simplify the configuration and options and remove the ability to adjust the AI.

Such would take away a crucial aspect of modding the game and the power of the series.

The second goal is to provide the average user more meaningful control over the difficulty of the AI, so that hard difficulties provide a more 'true to life' challenge, but lower difficulty settings enable players more lee-way while learning the game, but not at the cost of giving the AI brain damage (which is what, effectively, the current skill slider does). It's a blunt, unintuitive instrument, which most users simply won't know how to use properly if at all. Again, I simply don't see how you would take the view giving the majority of player meaningful control is a wrong goal or aim.

Of course not - this was discussed in the CIT and the forums a year ago already.

Why not just check these discussions the community made with the developers back then first to get an idea?

Let me be clear, this is not the be all and end all for 'improving' the AI - much more work needs to be invested in different areas, all those tickets, etc that you've listed, etc. That is different work but by no means less valid. This work is aimed at cleaning up wonky legacy design, giving mission makers and player better options, and reducing the likelihood of the average user breaking the game.

This is not different aspects - these are very closely connected to the player experience (and the technical backend).

The skill setting and it subsettings, as well as the precision setting are fundemental to the key AI issues.

If you look at one without the other, you will not understand the actual problem and how to solve them.

Players have simpler, more predictable UI options

• Some (more advanced/abstract) choices removed from game configuration (e.g. what is this arcane 'skill' black magic?)
  
• Only things that are simple to understand and test
• e.g AI: HARD / DEFAULT / LIMITED
  
  • Value is an abstraction of limited number of predictable inputs (e.g. AI Accuracy and Turning Speed)
    
    • HARD: Default (current state) Accuracy; Default (current State) Turning Speed
      
    • DEFAULT: Lower-than-default Accuracy; Default Turning Speed
      
    • LIMITED: Lower-than-default Accuracy; Lower-than-default Turning Speed
      

    [*]linked to difficulty presets (recruit would be "limited", etc)

    [*]no slider nonsense, just combo-box options; how much precision is really useful from the front-end?

It is a good idea overall - yet certainly an advanced tab is required to allow fine tuning - best down to all the specific settings:

http://community.bistudio.com/wiki/CfgAISkill

People want to adjust the difficulty to their preference.

People like tweaking, people can spend 1000x times the effort you or a mission author can.

This will lead to better results for everyone in the end.

There is no need to lock this down - instead it should be extended, better presented and the yet missing information/context given.

Do players really need to be able to set ability of both friendly and enemy AI?

• Why not simply abstract these settings into one value?
  
• Curate the friendly/enemy settings to what we think is 'best practise' for each default
  

Yes side specific setting is essential. A player want to play with different combinations / adjust it to his skill level or preference:

1. Easy: Enemy weak, friendly strong

2. Middle: Enemy weak/medium, friendly weak/medium

3. Hard: Enemy medium/strong, friendly medium/strong

4. Brutal: Enemy strong, friendly weak

More advanced AI settings should primarily be the mission (official/community) designers choice

• Configurable via the Editor
  
  • AI Skill
    
    • clearly validate and document what existing configuration options - unlock potential of parameters
      
    • Split into more options, if useful
      

To expand the options and provide decent documentation is certainly the way to go.

However as said before most missions/content are not CoD/movie style in OFP/Arma.

A mission designer can never tweak an open world, dynamic mission with freedom to perfection.

It must always be possible for the player or admin to adjust the difficulty level for a given mission.

A mission designer defines the skill of the specific units, and the difficulty with the number and placement of units.

The choice is up to the player if he wants to have an easy time or a challenge.

[xendance 3]

;2530513']

Yes side specific setting is essential. A player want to play with different combinations / adjust it to his skill level or preference:

1. Easy: Enemy weak' date=' friendly strong

2. Middle: Enemy weak/medium, friendly weak/medium

3. Hard: Enemy medium/strong, friendly medium/strong

4. Brutal: Enemy strong, friendly weak

[/quote']

I definitely don't agree with side specific skill selector. If you want harder missions, make harder missions. There are currently way too many variables for essentially the same thing, which makes designing missions with consistent difficulty hard.

As a mission designer I want to have the final say at the balance of the AI in my particular scenario. The user meddling with it might break the balance which results in an unplayable mission.

[gammadust 12]

I definitely don't agree with side specific skill selector. If you want harder missions, make harder missions. There are currently way too many variables for essentially the same thing, which makes designing missions with consistent difficulty hard.

As a mission designer I want to have the final say at the balance of the AI in my particular scenario. The user meddling with it might break the balance which results in an unplayable mission.

You're deflecting the insuficiencies of the whole AI internal configuration problems on the player's shoulders. You can't, it's a whole new can of worms you're opening. A mission design should totaly account for both overall adjustments of difficulty settings as much of balance settings. This is a very positive distinctive characteristic of Arma, throwing it away would be a loss for all.

At best, and recovering some merit in what you mention, the mission-maker should be able to define the default difficulty and balance, but it is his own responsability to assure the player that the mission does not break whitin the range of settings that a player may select for him self, he is not all knowing technician of the intricacies of the game he simply can't fulfil that role. This shouldn't be surprising, it is standard procedure on any game design which includes mission-making.

The problem remains as RiE put it: "Players can unwittingly 'break' the game by setting values that are too low", this is due to problems with the settings themselves and how they extend to mission-making, it should never become a player's problem.

[tmp95 16]

Ai settings, 2 additional points...

Why the apparently unwarranted necessity to set difficulty to absurd precision values:

// Recruit
	skillFriendly=0.64999998;
	skillEnemy=0.40000001;
	precisionFriendly=0.36999995;
	precisionEnemy=0.1;

// Regular
	skillFriendly=0.75;
	skillEnemy=0.60000002;
	precisionFriendly=0.54999989;
	precisionEnemy=0.28000003;

This is simply absurd, one even gets wary of changing those values (will somehow AI start being stupid if i set skillEnemy=0.6 instead of 0.60000002). Is this just a mere consequence of UI controls usage? Why not round those values, why not let the user step them (definite increments)...

Diverting the argument a bit but still relevant, the same is true for other unrelated to AI settings. It shouldn't be so. Those UI bars should adopt common user-friendly design practices (stepping), specially since UI bars with that functionality do exist already. An UI bar arrow that requires the user to click 1000 times to span it's length is simply no no in UI land. In addition many times the actual value is not shown, when it should.

---

Mission makers should have script access to values as defined by user. Ie. if user sets general AI skill to 0.8, mission maker should be able to read that value and adopt it in mission. This is an alternative way to connect user intent for difficulty and mission maker intents of difficulty and balance.

Is this still working within A3 1.02 hot fix patch? The adjusting of A3.UserProfile (Enemy Precision, Enemy Skill values??). It seem to have much more impact before the 1.02 hot fix patch came out. Now I'm not so certain if changing these varibles still work??

Though I think they very much should. i think this is a must for the customer to be able to tweak and adjust these two important skill values within the UserProfile.

[.kju 3239]

Just to be clear - difficulty settings should never break the game.

And currently they dont unless you design your mission so that you need to win these 5 guys over those 3 guys every time.

However as I tried to make clear before this is not the game design this series had from the start - it was always that depending on literally endless different aspects an engagement could go either way, a patrol could be present or not present.

It is the responsibility of the mission designer to design a mission that it can be played with any difficulty level - and the hard/brutal setting may be just that.

The actual problems are the bad values for unit/vehicle/weapon/ammo specific config parameters and the engine issues that make AI too good in some situations and behave very stupid in others.

[royaltyinexile 175]

;2530513']

You may not be aware of it' date=' yet there has been a very close cooperation and joint effort between Suma, the CIT managers and the CIT users during A2 and OA development to improve the AI in very specific ways.

From this people have gotten a very good understanding how the AI works and therefore have been able to make impressive AI tweaks or scripted AI behavior.

[/quote']

Perhaps, yet, the problem is that, some two years on, players - on the whole - rarely recognise any tangible benefits. I would see such endeavour to see this good work translated into better vanilla behaviour/options, rather than arcane black magic! :)

I can only imagine how much dear Mr. Klamacz will relish the thought of opening pandora's box. :cool:

;2530513']

A mission designer can never tweak an open world' date=' dynamic mission with freedom to perfection.

[/quote']

Indeed, but I'd argue that curation can be improved.

;2530513']

It must always be possible for the player or admin to adjust the difficulty level for a given mission.

Indeed' date=' but I'd argue that it should be within reasonable limits.

Why the apparently unwarranted necessity to set difficulty to absurd precision values

Yes, indeed, that one's going on the list.

A mission design should totally account for both overall adjustments of difficulty settings as much of balance settings [...] problems with the settings themselves and how they extend to mission-making, it should never become a player's problem.

Exactly. Front-end AI Difficulty settings shouldn't just be meaningful from the player's point of view, they should be predictable from the mission designer's point of view. It's catching that balance between useful configuration from the player POV and reasonable expectation from the designer POV that's the trick.

Anyway, cheers for the notes, links, etc, so far. I'll enjoy ambushing Oukej with this feedback when he returns from holiday! :cool: He's got the makings of a Task Force AI Config commander in him (he just doesn't realise it yet). :icon_twisted:

Best,

RiE

Edited October 9, 2013 by RoyaltyinExile

[windies 11]

Ideally what I would like to see in a most basic idea is that AI "settings" in the game difficulties controls overall settings. For example, if we opened up the skill slider and made each aspect of the current AI into it's own slider, precision could be set at .5 while everything else could be set to .75. This would be a hard cap for each individual setting, so no AI could ever exceed .5 precision no matter what settings are in the editor or in scripts. Then if I were to play a mission where the mission maker put down every soldier with his skill slider from in the editor set to the max, they would inherit the caps from MY game difficulty settings, so precision would be at .5 and everything else would be at .75.

This way, a mission maker could tailor difficulty of AI to what they feel they should be in their mission, and then the actual player can tailor the hard cap of the AI skills to match their own skills and so it's not as much of a guessing game anymore.

Keep in mind this is mostly how I would like to see editor versus scripted versus game difficulty skill settings to interact with each other, not so much how their difficulty should be configured. I think a lot of the AI are too hard/weak comes from the fact that they aren't programmed to really mimic typical human responses to situations, and this is what you mostly see AI addons trying to fix or make up for.

[Tonci87 163]

Perhaps, yet, the problem is that, some two years on, players - on the whole - rarely recognise any tangible benefits. I would see such endeavour to see this good work translated into better vanilla behaviour/options, rather than arcane black magic! :)

I can only imagine how much dear Mr. Klamacz will relish the thought of opening pandora's box. :cool:

Indeed, but I'd argue that curation can be improved.

Indeed, but I'd argue that it should be within reasonable limits.

Yes, indeed, that one's going on the list.

Exactly. Front-end AI Difficulty settings shouldn't just be meaningful from the player's point of view, they should be predictable from the mission designer's point of view. It's catching that balance between useful configuration from the player POV and reasonable expectation from the designer POV that's the trick.

Anyway, cheers for the notes, links, etc, so far. I'll enjoy ambushing Oukej with this feedback when he returns from holiday! :cool: He's got the makings of a Task Force AI Config commander in him (he just doesn't realise it yet). :icon_twisted:

Best,

RiE

What a splendid surprise for Oukej. Maybe he is reading the forums and now thinking of a way to decline, or become sick :p

I just wish to remind you that currently the players don´t have problems with AI that is too smart (God knows it isn´t) but with AI that is too precise when it shouldn´t be and with AI that knows things it shouldn´t know.

Suggestions to improve the situation:

-Make AI precision depend on Optics used (Snipers deadly accurate, Iron Sights almost useless at longer distances)

-AI Units that are farther away from the player should shoot at a slower rate to simulate aiming. Unless he has a LMG or MG, then he should shoot in short controled bursts. (Longer, less controled bursts if the AI is close to the enemy).

-AI Units should be less precise with the first shots they fire. Their precision should increase with the second or third shot.

-The instant information sharing on the players exact position within an AI group has to go. I think this might be very high on the list of reasons when people speak about unfair AI.

-Make AI less (maybe even much less) precise when it is wounded (Could be tied to the damage the AI received. If AI health 0.4 then decrease aiming by 60%)

-Make AI Ragdoll when it gets hit, it can get back on it´s feet afterwards. This would give players a better and more immersive feedback that they just hit someone. At the same time the player couldn´t be sure that the AI is dead if he doesn´t see the body wich makes for more cautious and interesting gameplay.

And of course you should try to make the AI smarter so that it gets closer to human behaviour :)

Good Night

Edited October 10, 2013 by Tonci87

[Smurf 12]

And make them use\suffer from supressive fire.

[xendance 3]

What a splendid surprise for Oukej. Maybe he is reading the forums and now thinking of a way to decline, or become sick :p

I just wish to remind you that currently the players don´t have problems with AI that is too smart (God knows it isn´t) but with AI that is too precise when it shouldn´t be and with AI that knows things it shouldn´t know.

Suggestions to improve the situation:

-Make AI precision depend on Optics used (Snipers deadly accurate, Iron Sights almost useless at longer distances)

-AI Units that are farther away from the player should shoot at a slower rate to simulate aiming.

-AI Units should be less precise with the first shots they fire. Their precision should increase with the second or third shot.

-The instant information sharing on the players exact position within an AI group has to go. I think this might be very high on the list of reasons when people speak about unfair AI.

And of course you should try to make the AI smarter so that it gets closer to human behaviour :)

Good Night

They already do both of those bolded things.

[OMAC 254]

1. I think all tactical capabilities of the AI should always be at the best level the engine can provide.

This covers things like caring for their own lives / suppresssion, choosing the right weapon for a situation, flanking, handling of formations, searching in the right places, taking cover, driving without crashing, rearming, applying first aid to oneself and ones comrades, ...

This aspect is, which can make AI appear human-like, if it is implemented reasonably, but which also has the potential to totally ruin gaming experience and break immersion if done poorly.

So I would always use the best the engine can give us, and never dumb down AI by disabling or crippling capabilities.

+1. When a subordinate medic gets injured, he should go prone if under fire and treat himself immediately. Then he should go and heal injured squad mates. If a squad AT or AA soldier gets killed, another squad mate not having such heavy weapons should go and pick up the dropped weapons, at least when a vehicle is attacking the squad. A leader can always order the AI to regroup if he deems it unsafe or otherwise unwise to perform these automatic actions. I mention these items as I very often have to order my AI subordinates to do these things, and it would be nice if they did them automatically. It would also be great if there was a setting or function which a mission maker could use to disable/enable such automatic actions.

[maturin 12]

They already do both of those bolded things.

But both features are broken. Now we have MGs using semi auto at range, and robotic streams of shots every .7 seconds that can't possibly allow for aiming inbetween.

And for increasing precision... it just doesn't happen. AI never has to find the range. If they engage from 800m, the first shot is just as likely to hit as the 50th. Except BIS has made accuracy get WORSE over time for some reason. And worst of all, there are strange little quirks built into the accuracy code. The first shot after reloading has an absurdly high hit rate. It comes out of the blue and smacks you upside the head and then people exaggerate about first-round 800m headshots.

Also, don't forget the bulllshit where AI shoot better standing than from prone.

[Variable 322]

...Except BIS has made accuracy get WORSE over time for some reason

If memory serves, I think they have, intentionally. I can't find the changelog that describe it right now.

Accuracy should increase as an AI fires, not drop. Shooters correct their shots according to visible hits.