Is RT the fundamental problem with TacAI?

[kipanderson]

Hi,

TacAI is the very heart of CM, it is what made CMX1 the great series it was.

Is the current problem with TacAI down to the RT engine? If so is the solution smaller games/scenarios? Charles programmed outstanding TacAI/ stunning TacAI in CMX1 so there must be some other issue causing the problem.

All the best,

looking forward to many more CMX2 games,

Kip.

[Itael]

My thoughts exactly.

Itai

[Elmar Bijlsma]

Gah!

[SgtMuhammed]

I seem to recall the TacAI being the number one complaint about CMx1. That and the curvature of the turret armor on the T34.

[Runyan99]

When you mean the Tactical AI, I assume you mean the AI which tells each individual soldier where to stand, who to fire at, what door to enter the building, etc. This is as opposed to the strategic AI, which tells the computer player where to move units, and which objectives to attack.

If the TacAI is rough, that is no suprise to me, and the reason isn't realtime, but the lack of abstraction which 1:1 modeling brings to CMSF. Handling the behavior of every individual on the battlefield is a whole new world of complexity which CMx1 didn't concern itself, and if soldiers and Strykers are exhibiting some odd behavior in this first CMx2 release, that's hardly suprising to me. I expect the TacAI to get refined and improved gradually over time.

If you really mean the Strategic AI, well, I always thought that the AI in CM sucked. In fact I think the AI in all games sucks, and that's why I wargame almost exclusively against humans.

[Sergei]

Originally posted by kipanderson:

Is the current problem with TacAI down to the RT engine?

No, you're out on a limb. AI programming is an arduous task, and the way it worked in CMx1 was rather unsatisfactory especially with panicked troops. Running out of good cover, doing the chicken dance etc. That had nothing to do with RT either.

[Panzer76]

Originally posted by Runyan99:

If you really mean the Strategic AI, well, I always thought that the AI in CM sucked. In fact I think the AI in all games sucks, and that's why I wargame almost exclusively against humans.

I agree, and I played cmbo/cmbb/cmak almost excusivly (after learning the game) on TCP/IP. Unfortunally, the WEGO system is now N/A for rhis mode, effectivly, making, for me, CMSF a single player game. And as such, even iof they fix the problems with it, it will go off my HD rather fast.

[Runyan99]

I think RT and CPU cycles are being used as scapegoats to explain pathfinding and TacAI CODE which is, in my estimation, still in it's adolescence.

Modeling the behavior of individual soldiers on the battlefield seems to me like an almost infinitely complex undertaking which can never be perfect, but only approach perfection the more time is spent on the issue. The programming will surely improve.

[Redwolf]

I think Kip is confused about the TacAI and the strategic AI, aka computed opponent.

The TacAI, along with it's evil brother automatic redrawing of user-instructed movement paths, actually wasn't any better or worse in CMx1. This problem is not as big in RT since you can instantly correct a wrong path or wrong reaction and don't have it up to 59 seconds move the wrong way.

The computed opponent in CMx1 was capable of coming up with a plan of it's own on the simpler CMx1 maps, which often enough was enjoyable to play against. This is gone and replaced by a directed computed opponent and I don't think it works too well yet. By their own admission the beta testers and scenario designers say that only few of them mastered the art of doing this.

This leaves us with 18 scenarios of which the majority have poor computed opponents. Remember that you cannot use autogenerated maps. If you don't have a hand-drawn map with a hand-drawn plan you can't play against the AI on it.

But worse, this seems to completely disable Quickbattles (even the ones on the standard maps) against the AI. Since the Quickbattles have their plan and custom setup zones in them, they seem never to be able to put the map-integrated plan into good use with a new set of units for the AI.

Hardly surprising, any of this, but look at how much this lowers the amount of "new stuff" you can play against.

Considering that BFC publically made playing against the AI the most important play mode for CM:SF there is a problem to solve here.

[Ronn]

This problem is not as big in RT since you can instantly correct a wrong path or wrong reaction and don't have it up to 59 seconds move the wrong way.

Sorry Redwolf,

tryed RT very hard .. but the action takes place in many parts of the battlefield. In RT I can correct one or two of this situations not more. I can't get back the WEGO-minute to see, what has happend on on the other side of the battlefield.

RT ist time pressure pur, and the ESC-pause a nice feature. Not more.

just my two cents

By the way: I love this game! I enjoy every minute, but only the WEGO-aspekt

[Stew]

Adding to Runyon's point above, people seem to forget that in creating an entirely new game engine, you can't simply cut-and-paste the old TacAI and pathfinding code. Surely BFC learned lessons from CMx1 that can be applied to the new code, but this is a much more complex environment and it's going to take awhile to get it right.

It's simply unfair to compare the TacAI of CMx1, which featured 12 patches for CMBO and two major upgrades in CMBB and CMAK, to a game that is less than a week old.

Already the new engine does all kinds of things that CMx1 couldn't (1:1 representation, relative spotting, semi-dynamic campaign, etc.). I totally agree that it's frustrating when it falls down on basic stuff like pathfinding. But that stuff is hard and it's simply going to take some time to get it right in the new engine, just as it did for CMx1. I'm glad to have the game now, if nothing else to fool around with the new interface and the editor. I'm going to hold off on the campaign, however, until there's been a patch or two.

One final point: the highly lethal nature of the modern battlefield means CMSF is going to have much less tolerance for error, human or computer. How many times in CMBO did you have units blunder into enemy fire only to escape back to cover while the first shot went wide or while the enemy tried to rotate his gun? There is MUCH less margin for error in modern combat. Screwups, whether due to bad tactics or bad AI are going to be much more noticable in CMSF than in CMBO and its siblings because they will almost always be fatal.

[Dirtweasle]

One final point: the highly lethal nature of the modern battlefield means CMSF is going to have much less tolerance for error, human or computer. How many times in CMBO did you have units blunder into enemy fire only to escape back to cover while the first shot went wide or while the enemy tried to rotate his gun? There is MUCH less margin for error in modern combat. Screwups, whether due to bad tactics or bad AI are going to be much more noticable in CMSF than in CMBO and its siblings because they will almost always be fatal.

That makes sense.

[kipanderson]

Hi,

No…I do understand the difference between TacAI and Strategic AI. I think .

TacAI is that which controls the individual units, when they stop if they spot an enemy, when they fire if they spot an enemy…and so on. The Strategic AI is now handled by Group Orders scripted by the scenario designer.

In CMX1 the Strategic AI was terrible, the TacAI…when a Sherman stopped and turned the turret and such…was outstanding . It is the behaviour of the individual unit/squad/AFV that worries me. I am happy to have humans scripting the Strategic AI.

If RT is no the problem then I am confident Charles will crack the TacAI problems. He did in CMX1 so why not in CMX2.

The behaviour of the individual units, given their orders either from a human or the scripted Strategic AI, is what makes CM for me.

The most immersive games ever developed .

Looking forward to many more CMX2 games,

All the best,

Kip.

[Fly Pusher]

Originally posted by sgtgoody (esq):

I seem to recall the TacAI being the number one complaint about CMx1. That and the curvature of the turret armor on the T34.

Not to mention the shocking omission of Bren gun tripods.

[Michael Dorosh]

Originally posted by Stew:

Adding to Runyon's point above, people seem to forget that in creating an entirely new game engine, you can't simply cut-and-paste the old TacAI and pathfinding code. Surely BFC learned lessons from CMx1 that can be applied to the new code, but this is a much more complex environment and it's going to take awhile to get it right.

It's simply unfair to compare the TacAI of CMx1, which featured 12 patches for CMBO and two major upgrades in CMBB and CMAK, to a game that is less than a week old.

Already the new engine does all kinds of things that CMx1 couldn't (1:1 representation, relative spotting, semi-dynamic campaign, etc.). I totally agree that it's frustrating when it falls down on basic stuff like pathfinding. But that stuff is hard and it's simply going to take some time to get it right in the new engine, just as it did for CMx1. I'm glad to have the game now, if nothing else to fool around with the new interface and the editor. I'm going to hold off on the campaign, however, until there's been a patch or two.

One final point: the highly lethal nature of the modern battlefield means CMSF is going to have much less tolerance for error, human or computer. How many times in CMBO did you have units blunder into enemy fire only to escape back to cover while the first shot went wide or while the enemy tried to rotate his gun? There is MUCH less margin for error in modern combat. Screwups, whether due to bad tactics or bad AI are going to be much more noticable in CMSF than in CMBO and its siblings because they will almost always be fatal.

Deserves to be quoted in full. "Welcome" to the forum, by the way. Two posts in 5 years ain't bad.

[Fly Pusher]

Originally posted by Ronn:

. I can't get back the WEGO-minute to see, what has happend on on the other side of the battlefield.

This is the KEY point that breaks RT - its impossible to know what is going on all over a complex battlefield from one perspective during one run through.

Also, WEGO being both processed and displayed in 60 sec means that corners need to be cut to get all the maths done in that time. I would happily look at the blue bar for 2-3+ min if it meant getting a turn where things worked...

What happened to 'its done when its done' ?

[Stew]

Thanks Michael. I'm generally more inclined to play the games than talk about them, but reading some of the traffic on the forum this week has just made me shake my head. There are plenty of bugs that need to be worked out in CMSF. Moreover, I'm not sure how well many CMx1 gamers (self included) will take to the demands of modern warfare. Increased lethality doesn't necessarily mean increased fun and perhaps just the opposite. But I'm certainly willing to give it a try and some of the conspiracy theories about RT and WEGO seem laughable. The history of excellent support for CMx1 gives me faith that BFC will eventually improve (if not perfect)the TacAI and pathfinding. Whether they can make modern warfare as fun as the rumbles in the French hedgerows is something I'll wait and see.

[Steiner14]

Stew,

thanks for remembering us about these important facts.

Indeed we need to give BFC time.

But that brings up another question for me: they knew they had an excellent brand name with Combat Mission. They also knew, they have a community not many game developers have.

What i don't understand is, if they knew how much work still needs to be done, why did they make the community believe, that a completely finished product, continuing where CMx1 stopped, will be released?

A word from Steve, that they will offer a buyable beta, because the game needs to grow over time, would have been understood by everyone and i think the majority of the old CM-players would have taken the chance to get the beta in their hands. But how they did it, could only lead to the high expectations - built on Cmx1 - they have to cope with now.

[RMC]

Originally posted by sgtgoody (esq):

I seem to recall the TacAI being the number one complaint about CMx1. That and the curvature of the turret armor on the T34.

And the thickness of the Tiger mantlet.

And Bren tripods.

And bad juju for Panzer IVs when hulldown.

[rlg85]

Originally posted by Stew:

One final point: the highly lethal nature of the modern battlefield means CMSF is going to have much less tolerance for error, human or computer. How many times in CMBO did you have units blunder into enemy fire only to escape back to cover while the first shot went wide or while the enemy tried to rotate his gun? There is MUCH less margin for error in modern combat. Screwups, whether due to bad tactics or bad AI are going to be much more noticable in CMSF than in CMBO and its siblings because they will almost always be fatal.

The major difference here is that in CMx1 they would actually *try* to get away (pop smoke, reverse, take cover etc.)

Here in CMSF I've seen:

Strykers that go after T-72s with machineguns instead of popping smoke and running away.

Lemming marches to windows under fire where one soldier after another gets killed.

Soldiers next to doors running around to the door on the other side, which once again features a lemming march into a machine gun.

Strykers given straight, parallel paths that inexplicably run into each other.

Soldiers seemingly oblivious to being under heavy fire and just stand up out of nowhere (and die)

And my favorite, when two sides meet in the same building, it usually involves a few seconds of staring at each other followed by troops running in circles, and the defending force butchering the attackers without any fight returned to them.

Sure, the CMx1 tacAI wasnt perfect, It pretty much didnt do a whole lot other than follow your waypoints, and keep in mind self-preservation. I dont know how many times I yelled at squads who turned and ran back into the open when only 10m away from cover.

But, at least it did keep in mind self preservation, and for the most part follow your waypoints the way you drew them. At least the way I see it (And I only speak for myself) the current tacAI basically prioritizes targets (sometimes) and thats pretty much it.

Heck, I had no problem dropping some of my speculations about the whole WEGO thing and the plan for the future, as I admittedly don't have any proof about BFC plans beyond my personal guess. (Although ive been thinking, had they dropped PBEM, which they openly considered, where would WEGO be? But eh, thats off topic.)

But the more I try to enjoy this game for what it is, the more I run into stuff like i've posted above, a lot with a lot of minor stuff I didn't mention. There is something fundamentally wrong with the tacAI, is it caused by RT, bugs, space lobsters? Who knows, I just know that these problems are seriously hurting the ability of this game to be great. Those who want to rationalize and say "oh you'll get used to it" can do so all they want, it will just hurt the product. All I'm looking for is a straightforward admission that the tacAI and pathing is a problem and will be worked on, instead of beating around the bush reasons why we aren't doing things right, or "just aren't used to it."

[Stew]

Re: an admission from BFC of at least some of the current problems , see here among other places:

http://www.battlefront.com/discuss/ultimatebb.php?ubb=get_topic;f=52;t=002086;p=2#000046

As to WHY things like TacAI and pathfinding are so buggy, I'm not the developer, but it seems pretty obvious to me: we are playing version 1.01 of a highly complicated game running on an entirely new engine.

[CRourke]

But you know, I never minded when CMx1 troops panicked and did something bizarre. I can always write that off as in the realm of possible under the stresses of combat. But show me a tank commander who thinks spinning in circles is a credible response to ambush by an AT team, and I'll show you the one tanker who survived my Arnhem campaign in CC2.

I agree AI programming is beyond hard. But most of the problems we're seeing are specific to pathfinding. Pathfinding AI is hard as well, especially when you need to consider enemy fire, but I've also heard its very CPU intensive. So it still seems logical to me that the code might benefit from a little more CPU time.

[stic.man]

Originally posted by Redwolf:

The computed opponent in CMx1 was capable of coming up with a plan of it's own on the simpler CMx1 maps, which often enough was enjoyable to play against. This is gone and replaced by a directed computed opponent and I don't think it works too well yet. By their own admission the beta testers and scenario designers say that only few of them mastered the art of doing this.

This leaves us with 18 scenarios of which the majority have poor computed opponents. Remember that you cannot use autogenerated maps. If you don't have a hand-drawn map with a hand-drawn plan you can't play against the AI on it.

Which scenarios have proved to generate an acceptable AI performance?

I've only played through the first few (in alphabetical order) as both sides so far.

Abu susah: I'd suggest you play as the US first to avoid any spoilers. Playing as the Syrians is interesting, however. After I destroyed a stryker(or more) the AI just sat around the destroyed stryker(s). 15 minutes later, 30 seconds left to go and they finally started moving toward the objective.

I think the Al Amarah scenario would be really good in a human vs human match.

Has anyone started reviewing any of the scenarios?

[Guest]

More CPU time is always great. However, if you're not limiting your pathfinding correctly (which CM:SF seems to have done), all the extra cycles in the world aren't going to help you. Your game will just be a hog, and this game, while intensive, does not seem to be a hog.

Therefore, I'd suggest that the pathfinding issues are not caused by hard choices imposed by resources, but rather that it's just buggy because pathfinding, especially on such nasty terrain and on such a small scale, is hard as hell to debug.

Unfortunately, CM:SF seems to create situations where every pathing corner case is hit in the course of a battle, so that hard-to-catch stuff seems to come up in players' games all the time.

Phil

[Grumbling Grognard]

IMHO (to answer the title) "yes".

But then again I never play RT so I would not miss it a lick if it was not there. Others...