Cmx2 universal tactical AI Rules.

[Other Means]

I thought I'd try and invoke some debate on what universal tactical rules we can think of that may be of use in CMx2. Simple, atomic things that can used as parameters for the fuzzy logic of the AI.

BFC have obviously been through this process themselves but it might pass us some time while we wait.

Some initial things i can think of:

<ul>

[*]MG's placement (although this is an area they are good at now)

<ul>

[*]Need to be in cover

[*]Need to be as far back as possible

[*]Need to be able to see at least 20% of the horizon

[*]Better to be covered from the front with a diagonal view of the map

[*]Attack should go through covered ground

<ul>

[*]Infantry should not get closer than 25m between squads

[*]If attacking through no cover a call to arty for smoke should be made (arty will be handled according to need)

[*]Squads should advance in formation

[*]Formation should change according to threat level etc

<ul>

I add more later, after the pub.

[junk2drive]

"If attacking through no cover a call to arty for smoke should be made (arty will be handled according to need)"

There are times that I question the tactics that we use in these simulations. I have been in CM battles where I thought "in RL I would pull back my troops into cover and wait for an artillery strike to help me out, or retreat altogether".

Instead we take our 20 to 60 minute turns and move across open ground in the face of mgs and he guns. Sometimes there are other tactics, sometimes not. I would like to see more options for covering open ground, like more smoke.

[YankeeDog]

"Universal Tactical Rule" and "fuzzy logic" are inherently contradictory.

Fuzzy logic does not work by "Universal Rules," but rather by a complex web of interacting variables. Each individual logic routine is quite simple, (in CM's case, something like "prefer to be at least 25m away from other friendly units"), but absolutely nothing is hard and fast.

The best layman's description of fuzzy logic I've ever heard is that each individual subroutine is like a little "demon", screaming into the primary routine's ear. The more closely a variable matches (or goes against, in the case of negative feedback) that subroutine's preference, the louder it screams. The master routine then makes its decision based on which demons (i.e., subroutines) it "hears" the loudest.

Additional complexity is added by creating intermediate level demons that combine the "voices" of lower-level demons before reporting to the next-higher echelon, and also by sometimes creating postive- and negative feedback loops between demons on the same level.

At least, that's what the coding guys tell me; it's not really my area of expertise. At any rate, So I'm not really sure the kind of rules you really mention are all that helpful; I think what's more important is how they interact.

Cheers,

YD

[Sequoia]

Originally posted by junk2drive:

There are times that I question the tactics that we use in these simulations. I have been in CM battles where I thought "in RL I would pull back my troops into cover and wait for an artillery strike to help me out, or retreat altogether".

I'm sure you're aware of many instances where the orders were to seize the objective regardless of cost or lack of support, even in the American Army.

[Other Means]

Simple, atomic things that can used as *parameters* for the fuzzy logic of the AI.

pub time.

[flamingknives]

I can think of several instances where I disobey some, many, or all of those rules. It's usually to my advantage.

Where the AI needs to be improved, IMHO, is in it's combined arms approach. Using MG and mortars to lay down cover fire (or at least be in a position to do so) until squad infantry (operating in platoons) have cleared up to the next patch of cover. Adding in higher 'levels' of AI might be a way forwards, so sections function as part of platoons, platoons as part of companies or detatchments, and all of them functioning as part of the whole.

Coding that sort of behaviour would be time consuming to say the least.

[junk2drive]

"I'm sure you're aware of many instances where the orders were to seize the objective regardless of cost or lack of support, even in the American Army."

Yes and that is when I continue playing and read that into the briefing provided. I imagine in my mind (a little one) that for whatever reason there is no support available and the objective must be taken within the timeframe allowed.

Such is the goal of the scenario designer, to make it challenging and believable.

[Other Means]

Originally posted by flamingknives:

I can think of several instances where I disobey some, many, or all of those rules. It's usually to my advantage.

Where the AI needs to be improved, IMHO, is in it's combined arms approach. Using MG and mortars to lay down cover fire (or at least be in a position to do so) until squad infantry (operating in platoons) have cleared up to the next patch of cover. Adding in higher 'levels' of AI might be a way forwards, so sections function as part of platoons, platoons as part of companies or detatchments, and all of them functioning as part of the whole.

Coding that sort of behaviour would be time consuming to say the least.

Yes, I think you're right. But before we can start to comment on /exactly/ what the AI should do to act against the general rule, I think it might be useful to define what that rule is. And then we can start to talk about how to combine them.

You see what I'm trying to do I think, define the tectical background against which to have these discussions.

[YankeeDog]

I appreciate the sentiment, but I think what you're talking about has already been done.

Basically what you're talking about is a battalion-level tactical operations manual. BFC actually sells several.

So I don't really think there's much need to re-invent the wheel; the hard part is getting from the concepts ideas expressed in said manuals, to an AI that can actually execute them. I know nowhere near enough about AI programming to discuss that subject with any intelligence.

Cheers,

YD

[Tero]

Originally posted by Other Means:

I thought I'd try and invoke some debate on what universal tactical rules we can think of that may be of use in CMx2. Simple, atomic things that can used as parameters for the fuzzy logic of the AI.

The "universal soldier" approach.

(NOTE: My take is rooted in WWII)

MG's placement

Only the Germans have a easily portable, belt fed GPMG, the rest have other arrangements (BAR, DT, Bren etc LMG's and Maxim etc HMG's)

Need to be in cover

The Germans dug in in front of the forest tree line, the Red Army and the Finns inside the forest tree line. The Western Allies ?

Need to be as far back as possible

Depends on the terrain. And the cover SOP.

Need to be able to see at least 20% of the horizon

Does not work in places like Hürtgen, Karelia, the Pripet marches etc

Attack should go through covered ground

Leading to lucrative predeterminable choke points ripe for preplanned arty hot spots.

Infantry should not get closer than 25m between squads

That would automatically push leading elements too far beyond the covering base of fire in certain situations (cresting a ridge line etc).

If attacking through no cover a call to arty for smoke should be made (arty will be handled according to need)

Finns did not use arty deployed smoke.

Squads should advance in formation

See the "no closer than 25m" rule.

Also, there were differences between the use of walking fire and squad/half squad/smaller element rushes.

Formation should change according to threat level etc

Dependant on the small arms demographics of the squads.

[securityguard]

Improving attack AI will be a huge undertaking. I'm not totally sure if in a game like this it can be totally improved, theres just so many ways to do things.

I'd personally like to see them up the defender AI though a lot. I'm sure it can be done. It's probably just as hard, but with less focus on movement and more on placement maybe it wouldn't be.

[flamingknives]

Originally posted by Other Means:

Yes, I think you're right. But before we can start to comment on /exactly/ what the AI should do to act against the general rule, I think it might be useful to define what that rule is. And then we can start to talk about how to combine them.

You see what I'm trying to do I think, define the tectical background against which to have these discussions.

There are parameters governing AI behaviour already, the problem, as I see it, is the lack of combined arms tactics.

My impression is that the strat. AI picks the approach line, while targetting and detailed movement is controlled by the squad AI (different to the TacAI), so it attacks as a mass, each squad making as good as it can until it meets opposition, where it continues attacking as an individual rather than acting according to role (MGs/mortars moving to the fore of an advance before squad infantry have cleared the ground) or organisation (platoons split up rather than pausing to rally and regroup)

[hoolaman]

It would be good if scenario designers could use SOPs, designate areas of advance and lay out a few different options for the AI to produce more realistic behaviour before the AI calculations even come into it.

It is possible to prod and coax the AI to attack through sensible routes now by making trails of flags or using no flags or putting reinforcements in strange unexpected places. If this could be factored in to the game properly through the behaviour of AI, the work would be half done.

If you think about how a human plans an attack, firstly they will identify where the enemy will be likely to deploy. Then they will split their forces to advance through planned routes, to deal with the areas that he hopes the enemy might be in.

If they see enemy units, they will try to assign them to a likely formation, and note that formation on the map. If the AI responds to all these factors, and identifies enemy units as "blocks" just like a human, then that might help simplify tactics.

[ November 28, 2004, 02:41 PM: Message edited by: Hoolaman ]

[aka_tom_w]

Originally posted by Hoolaman:

It would be good if scenario designers could use SOPs, designate areas of advance and lay out a few different options for the AI to produce more realistic behaviour before the AI calculations even come into it.

It is possible to prod and coax the AI to attack through sensible routes now by making trails of flags or using no flags or putting reinforcements in strange unexpected places. If this could be factored in to the game properly through the behaviour of AI, the work would be half done.

If you think about how a human plans an attack, firstly they will identify where the enemy will be likely to deploy. Then they will split their forces to advance through planned routes, to deal with the areas that he hopes the enemy might be in.

If they see enemy units, they will try to assign them to a likely formation, and note that formation on the map. If the AI responds to all these factors, and identifies enemy units as "blocks" just like a human, then that might help simplify tactics.

YES

I agree with this completely

I dream of a game where two players could set up a scenario and an attack and defense plan, with SOP's and flags and hints and zones of control and possible paths of advance and THEN let the game play out minute by minute and JUST WATCH the Game. (AI vs AI)

HEY for CMX2 maybe the game could play out AI vs AI (over time) and then RENDER a Movie (FULL blown movie replay) and the you sit back and watch the WHOLE movie.

The Basic Concept here is that the player ,in this OPTIONAL way of ENTERTAINING your self, is completely HANDS OFF.

Two players agree to set up a scenario where each player plays one role (say attacker and defender) once the scenario is designed (each side deployed set-up and give orders and SOPs INDEPENDENT of the outher side) then the Game Resolves every minute of the battle without any player intervention and you JUST SIT BACK AND watch those poor little units do the best they can with the orders you have given them....

(its JUST an idea BUT I think it would be REALLY entertaining to watch.....)

-tom w

[kipanderson]

Hi,

My view is that the Tactical AI is the most outstanding, of a number of outstand features of the current series of games.

When I first downloaded and played the demo for CMBO…as soon as it was released, the thing that struck me most was the quality of the AI controlling the manoeuvre units themselves. How they behave during the one minute movies given the orders they have, or have not received. No doubt BFC will find tweaks… but if there is a current problem with the AI I do not think it is with the individual unit tactical AI.

I think the Strategic AI has most room for improvement. It will never be as cunning as a smart-ish human, but in my view, others will differ, it is currently far too aggressive for its own good in defence.

I am a very slow player… I also try to take care of my virtual men… casualties matter to me;) When I attack against the AI I find the computer Strategic AI often “rushes about” in front of my men too much. Exposing the defender to my fire. The defending Strategic AI tries too hard to plug gaps too close to the front line… plus counterattacks too much.

Two recommendations.

The defending AI should rely more on ambush, which it does very well… Plus if it feels the need to plug to gaps, it must send units round behind cover, under cover, to plug the gaps. Not move them in the open in front of the attackers units. This includes less counterattacking.

A cunning initial placement, plus ambush, can in itself produce a fine AI control defence.

Overall, less movement by the defending AI would be better. But of course, there is no perfect solution.

All the best,

Kip.

[aka_tom_w]

Thats sounds good

I agree with Kip

-tom w

Originally posted by kipanderson:

Hi,

My view is that the Tactical AI is the most outstanding, of a number of outstand features of the current series of games.

When I first downloaded and played the demo for CMBO…as soon as it was released, the thing that struck me most was the quality of the AI controlling the manoeuvre units themselves. How they behave during the one minute movies given the orders they have, or have not received. No doubt BFC will find tweaks… but if there is a current problem with the AI I do not think it is with the individual unit tactical AI.

I think the Strategic AI has most room for improvement. It will never be as cunning as a smart-ish human, but in my view, others will differ, it is currently far too aggressive for its own good in defence.

I am a very slow player… I also try to take care of my virtual men… casualties matter to me;) When I attack against the AI I find the computer Strategic AI often “rushes about” in front of my men too much. Exposing the defender to my fire. The defending Strategic AI tries too hard to plug gaps too close to the front line… plus counterattacks too much.

Two recommendations.

The defending AI should rely more on ambush, which it does very well… Plus if it feels the need to plug to gaps, it must send units round behind cover, under cover, to plug the gaps. Not move them in the open in front of the attackers units. This includes less counterattacking.

A cunning initial placement, plus ambush, can in itself produce a fine AI control defence.

Overall, less movement by the defending AI would be better. But of course, there is no perfect solution.

All the best,

Kip.

[aka_tom_w]

bump

I wanted to re-read this thread in light of the latest CMx2 AI "hope its BETTER" thread.

-tom w