The "debate" about CMBB's Infantry Modeling

[akdavis]

My ONLY issue is with infantry units that attempt to auto-sneak towards cover 100m away instead of sneaking 30m out of LOS. Whether this is something the AI cannot handle, I don't know. It does seem that AI tanks are very capable of reversing out of LOS when taking fire.

BTW Steve, you should start more discussion threads like this for commonly raised issues and topics. If the "Battlefront" name is listed as the topic starter, people will naturally gravitate to the thread. I think it would cut down on the number of seperate threads bringing up the same damn topics. Less clutter, more debate.

[Doodlebug]

Steve,

Thank you for taking the trouble to read through my post. I would still like to see the AI capable of making the analysis of the relative merits of remaining still or sneaking forward or backward to cover when panicked. I suspect judging on the comments you made with regards to the difficulties of assessing walls and the like that it's not for this Game but for a future re-write. On that basis I will say no more on this and wish you well in continuing the good work already put into the game.

[Vanir Ausf B]

Originally posted by manchildstein II:

i believe that the problem with FOs and movement is separate from the other issues here...

Not within the context in which I was using it as an example. Replace the FO with a HMG or a mortar team, then. The numbers may change a little, but the point I was illustrating remains.

[Schoerner]

Originally posted by Boo_Radley:

What is even more frustrating is when I angrily give it targetting orders for the next turn, it keeps the red line for a few seconds, maybe even squeezes out a round or two, then stops.

Same here and i already had mentioned it in the closed thread explicitly.

Nothing new for Steve.

[ November 19, 2002, 05:47 PM: Message edited by: Schoerner ]

[manchildstein (ii)]

Originally posted by Vanir Ausf B:

</font><blockquote>quote:</font><hr />Originally posted by manchildstein II:

i believe that the problem with FOs and movement is separate from the other issues here...

Not within the context in which I was using it as an example. Replace the FO with a HMG or a mortar team, then. The numbers may change a little, but the point I was illustrating remains.</font>

[Cpt Kernow]

Yo doodle bug.

Cant confirm panicked troops crawling forward.Can confirm panicked troops crawling to nearest cover (sideways).

[SuperSulo]

I've had an SS MG-34 HMG panic on a Rocky tile, then proceed to run/sneak/walk in all directions, away from the enemy, towards the enemy (for a patch of wood some 200 m away), to the right... It kept doing this for some 12 turns before it finally became immobile. I kept begging that the russians would kill them . Oh did I say the moved in all directions? Well they didn't. If they had moved to the left, there was a patch of wood right next to the Rocky tile, with another SS MG-34. I guess those guys didn't like eachother or something

So yes, panicked units can definitly move towards the enemy.

[ November 19, 2002, 06:11 PM: Message edited by: SuperSulo ]

[Xavier]

I do like how the infantry is now modelled:)!

Thanx BTS for this nice game :cool: !

[Vanir Ausf B]

Looks like the Dance of Death in super slow motion.

[ November 19, 2002, 06:26 PM: Message edited by: Vanir Ausf B ]

[Michael Emrys]

At this convenient point, I'd like to say that the only thing that causes me real grief is the Dance of Death. I can accept a unit Sneaking, Running or whatever towards what it considers to be safer cover if it would just do that and be done with it. What is absurd, and irritating if it is one's own troops, is when it leaps up and goes running off in one direction, takes six steps, decides to go off in another direction, covers a meter or two, spins around to go off in yet a third direction, et cetera ad infinitum until it is annihilated completely. All this is naturally accomplished in open terrain in plain sight of the enemy, of course, which is delighted to concentrate every gun in the county on the "behaviorally odd" unit.

Okay, maybe this one gets filed under "ain't it hard to program a computer sometimes" heading. Cybernetic devices can have the hardest time trying to do things that a human finds self-evident. Interesting article about that in the current Atlantic Monthly, BTW. Perhaps the existing engine just can't do any better than this for reasons beyond my meager understanding. These things happen sometimes and if that's the case, I'll live with it.

But I do in that case earnestly hope and pray that the engine rewrite will finally and mercifully put a large caliber revolver to the head of this little idiocy and totally blow its f'ing brains to kingdom come.

Otherwise, I'm fine. At least that's what my therapists tell me.

Michael

[Battlefront.com]

Michael,

What is absurd, and irritating if it is one's own troops, is when it leaps up and goes running off in one direction, takes six steps, decides to go off in another direction, covers a meter or two, spins around to go off in yet a third direction, et cetera ad infinitum until it is annihilated completely.

This and the above MG team example are direct results of units having no memory. It is not possible for a unit to remain wedded to something without seriously negative side effects.

Here is how the TacAI works now...

A unit goes bumbling along following the player's orders to the letter. It will continue to do so unless it receives some sort of negative stimulus bad enough to make the unit disregard the player's orders (this is a very complex decision making process and is highly situationally dependent).

The unit is now instructed by the TacAI to perform some sort of evasive action, either controlled or in Panic. Let us say it is to go in Panic from open terrain to a patch of woods 20m to the left.

The unit starts to go to the left, but a turn later receives enough fire that the TacAI decides the unit is in trouble. Depending on where the unit is, its condition, the nature/direction of the enemy fire, proximity to known enemy units, etc. the TacAI will make a decision about what to do. Often it will choose the same exact course of action already being embarked upon. However, it might not.

The TacAI makes a new decision completely and utterly in ignorance of the decision/s that came before it. It has zero concept of what the TacAI previous ordered it to do, not to mention what the player originally intended. This means the TacAI has zero concept that the unit was moving towards that patch of woods, nor the reason for it. The TacAI always thinks in the NOW and never in the PAST or FUTURE. That is what I mean by it has no memory.

OK, so let us say that because of the situation the TacAI now thinks that a different piece of terrain is the best bet for safety. It starts to head towards it. The proximity of the woods is irrelevant now because it has already, for whatever complex reason, been rulled out. The turn after that the unit gets hammered again, and once again the TacAI is triggered to do something about it. In that exact situation it now decides that the woods is a good place to go and so it reverses direction.

And there you have the Dance of Death. There are only two solutions to such a problem once it appears (and it is NOT certain to appear). One is to get lucky and have the unit manage to get out of trouble during one of its TacAI issued orders. This happens quite easily in good cover, not so well when out in the open and swept by the fire of multiple enemy units. The second way out is to distract/suppress/destroy/retreat the enemy units firing at your happless unit. This is obviously situationally dependent, but the player that has uses covering fire better will more chances to do this.

OK... I know what you guys are going to suggest... just have the TacAI make one decision and never override it. Not a good idea more often than not.

The main reason is that the TacAI is not necessarily going to make the right call the first time. Therefore, hardcoding the TacAI to never make a change until the initial order is carried out is asking for trouble. Like this....

Our unit in the above example is caught in the open and is told to go 20m to the left into the woods. It moves 1m towards the cover and an enemy SMG unit becomes spotted. Where? Why, in the woods of course So the Panicked unit starts to crawl TOWARDS a SMG Squad. First couple of meters it takes a casualty or two. The old CM code is screaming at the unit to change direction, but the unit is ignoring it so it moves on. A couple of meters later it is hit again, and again it ignores the pleas for the unit to change course. The Human player is also yelling the same thing A couple more meters, a couple more casualties, the unit is now Broken. Still it continues on and is wiped out.

This MIGHT happen in CM now, but it is not assured. And if you look at the game carefully as you play it next time, I am sure you will find lots of examples far less extreme than mine above that would be horrible if the TacAI was hardcoded to ignore future problems until its current order was acheived.

No folks, until we can afford the computing time, RAM, and file space we can't allow a unit to have memory. And until a unit has memory, there isn't much we can do. Fortunately, the new engine will be written for hardware that can afford this overhead and so it is in the plan to introduce this sort of "smarts" into the TacAI.

Steve

[Berlichtingen]

Originally posted by Battlefront.com:

This leaves only two possible reasons to explain how there can be such diametrically opposed viewpoints:

1. Tactics

2. Expectations

3. Understanding the Orders

Your inability to count is a clear indication that your entire theory is flawed

[Voxman]

In CMBO the troops did not break easily enough

In CMBB the troops break to easily.

What we need is the happy median

Regards,

Voxman

[ November 19, 2002, 09:01 PM: Message edited by: Voxman ]

[SuperSulo]

Re: my dancing HMG. Is it correct that units have no knowledge of friendly units nearby? Safety in numbers, or something? In my pic, 10 m to the left, they had a friendly unit. Should that be taken into consideration when choosing a "retreat" path?

Also, they were in rocky terrain, with "decent cover and concealment". Why start sneaking to steppe in the first place?

Oh and as for helping them get out of the "Dance of Death" by providing covering fire, ALL my 1501 pts of units were on LOW ammo, facing 3750 pts of Russian infantry. Hmm come to think of it, who can blame the poor conscript SS HMG unit for panicing in the first place?

For the record (if anyone is keeping one), this is the 5th time I've experienced the sneak-til-u-freak feature. 2 HMG's, 1 mortar, 2 FO's. I don't see it as a major problem, just a small dent on an otherwise pretty darn good surface.

Memory is good. Both in humans and computers.

[Redwolf]

Originally posted by Voxman:

In CMBO the troops did not break easily enough

In CMBB the troops break to easily.

None of the (IMHO) intelligent whiners says that.

What is rightfully being complained about is the combination of several factors, giving an unsatisfying and unrealistic result for the composition of all factors, even though each factor considered aline would look OK.

Example: taking control away from player (automatically switch to some command), maybe doing it a little too early (which wouldn't hurt in isolation), having this command being heavily tiring (in isolation, that would be OK) and combining all that with giving the same slow tiring command, which may work well for some units (e.g.squads which can also run) but for some other units it can be devastating (heavy weapons which cannot run and hence pass all the way in sneak.

Each of these factors alone would be OK. The combination leaves some rough edges, the auto-sneak-exhaustion for heavy weapons being the worst, but not the only one.

So everybody (and sadly even Steve in this thread) comes along and discusses single aspects. No wonder we get nowhere.

[ November 19, 2002, 07:57 PM: Message edited by: redwolf ]

[murpes]

Originally posted by Battlefront.com:

Fortunately, the new engine will be written for hardware that can afford this overhead and so it is in the plan to introduce this sort of "smarts" into the TacAI.

You're writing the new engine for the X-Box?

[Schoerner]

Steve, thanks for the explanations.

Wouldn't it be meaningful to make the target-location choosen by the TacAI more static with explicit exceptions?

i.e.: if the target location of a panicked unit is "under control" (means very close) of a hostile unit with a certain amount of firepower (say 70% of the panicked unit) and LOS, then the TacAI gets permission to change the direction of the unit like usual (i guess deciding with a algorithm between nearest and best cover and LOS calculations).

For the next choosen location the same is guilty.

If we take the example with the picture above, the Tac AI would have looked for nearest cover, best cover, made it's decision - surely for the wood with the friendly unit on the left - and no matter about hostile fire on the battlefield, it keeps sneaking/going/running torwards the life-saving wood.

Any examples where this wouldn't work?

[The_Capt]

Originally posted by SuperSulo:

I've had an SS MG-34 HMG panic on a Rocky tile, then proceed to run/sneak/walk in all directions, away from the enemy, towards the enemy (for a patch of wood some 200 m away), to the right... It kept doing this for some 12 turns before it finally became immobile. I kept begging that the russians would kill them . Oh did I say the moved in all directions? Well they didn't. If they had moved to the left, there was a patch of wood right next to the Rocky tile, with another SS MG-34. I guess those guys didn't like eachother or something

So yes, panicked units can definitly move towards the enemy.

LOL if you ran into a Pls worth of fire over open ground I am pretty sure we would all be acting like idiots.

Seriously, the best thing you can do when you troops get caught in the open, is exactly what you are suppose to do in Real Life...get em down and under cover..ie hide em. That is if you haven't sent them into a situation so bad that they have panicked.

Once hidden they tend to stand still and also take fewer casualties..THEN if you were doing your job you can bring fire to bear on whomever is making that poor grunts life miserable and then carry on with the advance or get them out of there.

If they panic or rout..well first off they were somewhere they shouldn't have been and second off all you can do is hope they go under cover and under command of an HQ unit.

All of the rest of this reminds me of pool..blame the stick..blame the table..blame the balls..blame the whiskey..blame the lighting...blame the smoke...you get the idea.

[ November 19, 2002, 08:14 PM: Message edited by: The_Capt ]

[Ellros]

Originally posted by Berlichtingen:

</font><blockquote>quote:</font><hr />Originally posted by Battlefront.com:

This leaves only two possible reasons to explain how there can be such diametrically opposed viewpoints:

1. Tactics

2. Expectations

3. Understanding the Orders

Your inability to count is a clear indication that your entire theory is flawed</font>

[Battlefront.com]

Redwolf,

So everybody (and sadly even Steve in this thread) comes along and discusses single aspects. No wonder we get nowhere.

No, we get nowhere because there is no one place people want us to go. Each has their own narrow aggenda, expectations, and desires. Unfortunately, each individual (such as yourself) likes to think that his own pet issues are the ones we should focus on and ignore the rest. To you it is clear and simple. But to us, we have dozens of opinions ranging form "I think this game sucks" to "don't change a bloody thing". If you think that is an easy thing to sort out, might I suggest running for the top position of the UN someday soon

One can NOT discuss a simulation from the higher level. It simply doesn't work that way. A simulation of a higher level event (such as reaction to enemy fire in a specific situation) is based on individual elements. There is absolutely NO WAY to discuss the higher level stuff with even an ounce of productivity without discussing the specifics.

And as always, the "complainers" assume that their opinion is correct and that there is indeed a problem that exists in the first place. Your comments above presume that the current implementation is less realistic than what you imagine, or that your proposed "fix" will result in better behavior. And of course, very rarely does anybody consider the fact that there is ZERO chance of us making a system that doesn't have SOMEONE complaining that it needs fixing.

At some point we are working with diminishing returns in front of us. We were at that point even before the final code was released. People forget that we didn't make the game 2 years ago and sat around inbetween then and now not making any changes, tweaks, or improvements.

Schoerner,

The TacAI already does calculations like this. We have played with the variables and added/subtracted decision trees more than I care to remember. It isn't perfect now, but we feel that tweaking it more will not fix the problem. Like some other issues (spotting comes to mind right away) until we can rewrite the fundamental basis we are pretty much looking at the best the current game engine can yeild.

[Redwolf]

Steve

Originally posted by Battlefront.com:

One can NOT discuss a simulation from the higher level. It simply doesn't work that way. A simulation of a higher level event (such as reaction to enemy fire in a specific situation) is based on individual elements. There is absolutely NO WAY to discuss the higher level stuff with even an ounce of productivity without discussing the specifics.

Any implementable computer model for this kind of human behaviour will have rough edges where it falls short of realism. These edges are really edges, as I said they are a result of "unlucky" units being hit by several aspects resulting in a mess.

My point is that you need to identify a few such edges and patch them up manually.

I know that this is painful and it is easy to make thing worse, but that doesn't mean the need does not exist.

And as always, the "complainers" assume that their opinion is correct and that there is indeed a problem that exists in the first place. Your comments above presume that the current implementation is less realistic than what you imagine, or that your proposed "fix" will result in better behavior. And of course, very rarely does anybody consider the fact that there is ZERO chance of us making a system that doesn't have SOMEONE complaining that it needs fixing.

I followed all the discussions pretty closely and I see a very limited number of these rough edges coming up if you only count people who do not clearly fall into one of the "all-whiner" or "whine-whiner" camps.

The auto-sneak exhaustion is one, the push-out-off-foxholes another, simple bad choice of destination for automatically issued run/sneak a third.

Certainly looks agreeable and implementable to me.

[Silvio Manuel]

Originally posted by SuperSulo:

For the record (if anyone is keeping one), this is the 5th time I've experienced the sneak-til-u-freak feature.

"Sneak-til-u-freak." I like it

[Michael Emrys]

Thank you, Steve, for your reply. I eagerly await CMII (and not for this reason only) and in the meanwhile will grit my teeth and live with the problem. The fact that it seems to be happening to my enemy much more than me makes that task easier.

Michael

[SuperSulo]

The_Capt, read my follow up regarding that pic. They were in rocky terrain, with "decent cover and concealment", in a foxhole, not open ground. The enemy were alot further away when the HMG panicked, that's the end battle pic. As for "bring fire to bear", that's pretty difficult to do when you're all out of ammo. I had even used up my all my AT shells.

As for being where they shouldn't have been in the first place, well, HQ thought my poor reinforced company of 221 men should defend this flat, open, steppe against 1571 fanatic soviets... And I thought that this rocky ground would be better than on the open steppe.

My tactics may stink, but please don't blame my tactics this time, in the end I handle the assignment pretty well, if I may say so myself. 893 enemy casualties, 225 KIA. I suffered 10 casualties, 3 KIA.

I also didn't blame the table or anything , I just provided some info on panicked units moving towards the enemy.

[Schoerner]

Originally posted by Michael emrys:

Thank you, Steve, for your reply. I eagerly await CMII (and not for this reason only) and in the meanwhile will grit my teeth and live with the problem.

Couldn't have said it better.

Me, too.

The fact that it seems to be happening to my enemy much more than me makes that task easier.

Michael

After i finished my learning phase, i'll feel free to contact you to find out