CMx2's new movement system explained

[Battlefront.com]

There are some things that are inherently problematic for RT, just as there are things that are problematic for WG. The major one for RT is the difficulty in wearing many hats all at once and doing a good job of it. As I said before, we can not fix this problem. Even CoPlay, where you are in command of a single Platoon, this problem exists to some extent or other based on conditions. There are things we can do to lessen the problem of too many things to do in too many places in too short a period of time... but there is no way to fix it completely.

The way wargame designers deal with RT is to favor one theoretical extreme or the other. On the one hand there is the side that has the AI do pretty much everything and you, the player, need to do very little. To do this there needs to be a jaw dropping amount of extremely good AI programming to be mildly decent. This is what we've been calling a Command Level game.

The other extreme is to have the game consist of turns of fairly short duration. This allows the player to wear all the hats he wants, to the degree he wants, in a timeframe that allows suitable tactical flexibility. However, lots and lots of artificial constructs are needed to minimize micromangement possibilities, there is a highly unrealistic degree of information, coordination, and precision of execution is far too high.

Both of these have their faults, neither one has a grip on the title of "more realistic". The RT extreme tends towards twitch, the Turn extreme tends towards tedium.

CMx1 struck a balance between these two extremes. A very good balance in most people's opinion here. But we feel that it leans too much towards the Turn extreme than we would like. We've taken major steps to change that with CMx2. Whether played in WeGo or in RT I am sure you will very quickly see what we've been talking about for the last two years. Relative Spotting and other features fundamentally alter the gameplay in so many ways, yet keeps the game true to its roots and appeal.

The option for RT is an extension of the work we did to make the game system inherently more realistic. As I've said, the core engine operates in continuous time. Therefore, technically it isn't very difficult for us to allow for RT and WG play styles. Whether or not a majority of CMx1ers will like the RT feature is not of concern to us because the WG option is still very much there. Trying to change RT to be little different than WG isn't any better than trying to change WG to be pure Turn or pure RT. We need to all be very careful to not blur the distinction between the two.

Steve

[ November 10, 2005, 11:06 PM: Message edited by: Battlefront.com ]

[Lee]

This thread has got me thinking about an issue with the AI. I certainly hope that

the tac AI will not be lessened in CMII because of RT play considerations. For

example, the tac AI sophistication (i.e., the complexity of the math the

computer has to figure out) that one can achieve in RT play has a very real

limit to it. Because it RT you have to worry about the affect highly complex

AI routines will have on the frame rate of the 3D graphics, which also

has to be worked on by the computer chip at the same time as the AI. Whereas

in turn-based Wego it doesn't matter, because all the combat math has been

done before the graphics even have to be displayed. If that means that during RT

play the tac AI maybe only has a subset of the full tac AI for turn-based,

then so be it.

The same would also apply for armor penetration calculations. I can see where

this might be a RT frame rate issue in a large armor engagement with lots of hits on

tanks and APC's taking place. Whereas it's not an issue in turn-based combat.

So Charles is free to make the armor penetration calculations about as complex

as he wants in turn-based (within reason).

[Battlefront.com]

I meant to follow up this comment I made:

We need to all be very careful to not blur the distinction between the two.

With this:

Because if we do blur the line we'll likely wind up with a method of play that neither the RT nor the WG players are happy with. Better to make one clearly RT and another clearly WG, and allow both camps to play the way they want to.

Lee, the TacAI will likely perform better in CMx2 than in CMx1 for a variety of reasons. Other things, such as penetration calculations, likely won't have to be sacrificed because the hardware is a lot better and the code is far better laid out. In fact, we now have real time line of fire, which was not in CMx1. We also have detailed damage modeling for vehicle, such as taking out the FLIR of an Abrams. Actually, everywhere I look I see more detail and not less. That's mostly because of the fresh start with the code base.

Steve

[ November 11, 2005, 10:09 AM: Message edited by: Battlefront.com ]

[M1A1TC]

As far as I know Abrams do not have FLIR. They do have thermal sights, though

[Dschugaschwili]

Originally posted by Battlefront.com:

As I've said, the core engine operates in continuous time. Therefore, technically it isn't very difficult for us to allow for RT and WG play styles.

Steve

So, it should not be a problem to offer different turn lengths for WeGo. Will there be an option to use a turn length other than 60s in CMX2?

Dschugaschwili

[aka_tom_w]

CMx2 will have better Tac AI than CMx1...

corrected now in the original post

thanks Steve...

[ November 11, 2005, 10:46 AM: Message edited by: aka_tom_w ]

[Thomm]

Originally posted by aka_tom_w:

Why not better Tac AI in CMx2??

The price of real-time! Muahhh!!!

[Delyn]

After reading the rest several times I would assume Steve meant to say that the TacAI will be better in CMx2 because hardware and coding is better.

Originally posted by Battlefront.com:

Lee, the TacAI will likely perform better in CMx1 than in CMx2 for a variety of reasons. Other things, such as penetration calculations, likely won't have to be sacrificed because the hardware is a lot better and the code is far better laid out. In fact, we now have real time line of fire, which was not in CMx1. We also have detailed damage modeling for vehicle, such as taking out the FLIR of an Abrams. Actually, everywhere I look I see more detail and not less. That's mostly because of the fresh start with the code base.

Steve

[Andrew H.]

Originally posted by Typis:

After reading the rest several times I would assume Steve meant to say that the TacAI will be better in CMx2 because hardware and coding is better.

That's what I think he meant, too - and it would be consistent with earlier statements he's made that, even without coding changes, relative spotting and object permanence will make the AI act smarter.

[Battlefront.com]

Man, talk about typos! Yeah, I accidentally switched around CMx1 and CMx2. Corrected it above to cover my butt

M1A2 SEP has 2nd Gen FLIR on it. Some of the Abrams you will see in CM:SF will be the upgraded SEPs.

Steve

[Lee]

Steve: I have no doubt the tac AI will be better in CMII than CM, naturally

you guys have been thinking of ways to make it better for quite some time.

The concern I have is that the tac AI in CMII will not be *as good* as it would have

been if there was no consideration given to the tac AI's hit on frame rate when

someone plays in RT. Unless, of course, Charles is simply designing the tac AI

and armor penetration modeling as if turn-based wego was the only play option in

the game and whatever affect it has on frame rate for RT, then so be it.

Hopefully that is the case. I don't mind people having the option to play

RT if they want to, but if it starts to cause a lessening of the realism

that we could have going on under the hood for turn-based, then it becomes

a real issue.

[aka_tom_w]

Unless, of course, Charles is simply designing the tac AI and armor penetration modeling as if turn-based wego was the only play option in the game and whatever affect it has on frame rate for RT, then so be it. Hopefully that is the case. I don't mind people having the option to play RT if they want to, but if it starts to cause a lessening of the realism that we could have going on under the hood for turn-based, then it becomes a real issue.

I could be wrong but everything Steve has said so far indicates this potential problem you think they "might" be having or going to have does not appear to an issue for Steve et. al. AT ALL.

Steve has said a few times over EVERYTHING about the game engine has been designed and developed to deliver a high quality RealTime play experience. (at least I think that is what they have been saying)

I guess the demo can give us a clue of what we can expect... and yes I know it is a long way away at this point.

-tom w

[Battlefront.com]

Sure, in theory we could have better AI if the game were just WeGo. We could also have better AI if we spent 3 years working on the AI The point, I think, is that from a practical standpoint the TacAI will be better and that's all that really matters. I don't think Charles would have the time to program the TacAI any better anyway. So in real terms I don't think the continuous time design of the core engine has any negative effect on the game's eventual performance.

Steve

[Lee]

Steve: From what you're saying, Charles is designing the tac AI and such

the same way he would if wego were the only option. I'm not talking about the

underlying RT nature of the combat simulation, but rather not worrying about

frame rates during RT play and just designing the AI, armor penetration modeling

and so on to be the most realistic it can be given the amount of time he has

to work on programming those aspects of the simulation. That's great to hear!

By the way, the very detailed way that vehicles can take damage to specific

systems sounds really cool.

[kipanderson]

Hi,

Referring to the AI in CMSF Steve posted,

“The point, I think, is that from a practical standpoint the TacAI will be better and that's all that really matters.”

Well… in my view the TacAI in CMX1 was the most outstanding of many outstanding features. So if it will be even better in CMX2 games all is well that ends well .

Steve,

Can I ask, has what I know as the Strategic AI been improved in CMX2? The TacAI, the control of the units, given their orders, during the one minute movies was outstanding, but the Strategic AI did struggle a bit . To give an example. In defense it ambushed very well, could be very deadly opening fire at the correct time for the weapons system. However… it insisted on counter attacking way too much, and generally rushing around under the guns of the attacker… in my view. Will this be better in CMX2?

My view of the Strategic AI is that it should be less ambitious, stick to what it does best. i.e. in defense stick more to ambushing and positioning units in the set up.

All good fun,

All the best,

Kip.

[aka_tom_w]

Steve,

Can I ask, has what I know as the Strategic AI been improved in CMX2? The TacAI, the control of the units, given their orders, during the one minute movies was outstanding, but the Strategic AI did struggle a bit . To give an example. In defense it ambushed very well, could be very deadly opening fire at the correct time for the weapons system. However… it insisted on counter attacking way too much, and generally rushing around under the guns of the attacker… in my view. Will this be better in CMX2?

My view of the Strategic AI is that it should be less ambitious, stick to what it does best. i.e. in defense stick more to ambushing and positioning units in the set up.

I agree

and if I may add

I am not so sure the set up and positioning of units should be such a big priority because in any pre-made scenario the scenario designer can position the units at set up best.

I would suggest that because quick battles are just that "quick" then no further Strategic AI work is required as the set up will never really be ideal any way. The quick battle against the AI is the least realistic and most opportunistic way to play so the Strat AI should not be expected to do set up all that expertly or as clever as a scenario designer or human opponent.

BUT on the other hand, while playing solo in a user designed scenario against the AI it would be nice to see the Strat AI put up a better fight.

"My view of the Strategic AI is that it should be less ambitious, stick to what it does best. i.e. in defense stick more to ambushing"

I think this could be tricky in CM:SF because the "things" (concepts principles and tactics) that the America Strat AI will need or use or rely on will be completely different (or "should be" IMHO) completely different than the tactics and strategy of the Syrians. (this alone would almost double the work load of Strat AI coding. (But I could be wrong about that I hope.)

Just my few thoughts.

-tom w

[Battlefront.com]

Oh, the Strategic AI will be vastly superior At lesat for premade sceanrios. I'm not sure how much we can improve it for free form, on the fly type battles. One of the things we are allowing scenario designers is the ability to "spell things out" for the StratAI. Instead of CM trying to figure out how things should go, the designer can give some pretty strong instructions. For example, defining where the best defensive locations are for an Objective or what the best advance route is to an Objective. Stuff like that. Since the worst problems with the StratAI were determining context, its performance should be quite a bit better. It also allows Charles to spend more of his time coding TacAI and OperationalAI (like coordinating platoon elements).

Who knows how it will come out in the end, but we are optimistic that it will be noticably better. As good as we all want it to be? Certainly not. That is just too much to ask for. But beter? Most certainly.

Steve

[aka_tom_w]

One of the things we are allowing scenario designers is the ability to "spell things out" for the StratAI. Instead of CM trying to figure out how things should go, the designer can give some pretty strong instructions. For example, defining where the best defensive locations are for an Objective or what the best advance route is to an Objective. Stuff like that. Since the worst problems with the StratAI were determining context, its performance should be quite a bit better.

That sounds GREAT!

Solo play against WELL made scenario's by "Big Name", Marquee (i.e. experienced) scenario designers could be something to really look forward to, as the player will know he will have a royal battle on his hands because the Strat AI (it would appear) can now be hinted by the designer (human intelligence!) to give the AI a fighting chance.

There should be no end of great scenario's to play as either side against the AI where the opposing side has been designed and hinted by scenario designers.

This is exactly what EVERY married guy with young kids at home dreams about! Some of us don't have very much time to play and prefer solo play (unfortunately) due to RealLife time constraints, so the option to play Solo late night battles battles against the Strat AI that has been hinted and programed by clever humans (he he) would be a dream come true!

(I think early on the favourite scenario would be to design clever defenses and ambush style counter attacks for the Syrian AI player to allow the American human player to feel the "sting" of the low tech Syrian (asymetric warfare??) defensive tactics. But I could be wrong )

Thanks for the update Steve.

-tom w

[ November 13, 2005, 09:09 AM: Message edited by: aka_tom_w ]

[Cpl Steiner]

The Strat AI scripting idea sounds very interesting, but will it be purely linear? I can imagine this making a scenario very predicable after a couple of play-throughs.

It would be nice if the designer could specify some "switches" or "triggers" to alter the AI behaviour depending on events. For instance, perhaps one route of attack could be closed off if it receives a lot of enemy fire, resulting in the AI using another less preferred route.

[aka_tom_w]

Originally posted by Cpl Steiner:

The Strat AI scripting idea sounds very interesting, but will it be purely linear? I can imagine this making a scenario very predicable after a couple of play-throughs.

It would be nice if the designer could specify some "switches" or "triggers" to alter the AI behaviour depending on events. For instance, perhaps one route of attack could be closed off if it receives a lot of enemy fire, resulting in the AI using another less preferred route.

I am not concerned about this.

After a scenario is played once it should be understood the Strat AI behaviour each time the scenario would be re-played would likely be similiar and predictable.

If you want a new challenge play another pre-made scenario.

I would suggest it is unreasonable to expect the Strat AI to be upredictable each time a designed and hinted scenario is played over and over again.

Look at the shear number new scenarios available for CMBO CMBB, and CMAK. After just a few months of the release there will be so many new user designed scenarios available you should never need to play the same one twice.

I disagree completely with the idea that a user made scenario should be coded to be play different ever time you re-play the scenario. The thrill of playing the scenario is always best the first time around and after that it is just "another NOT new, already played" scenario in my mind. For a new challenge either design a new scenario for someone else to enjoy or download another new user made scenario and play it for the first time and see if you do any better. Or you could fall back on the ever popular "Quick battle" and see how you do there.

-tom w

[Cameroon]

Originally posted by aka_tom_w:

I would suggest it is unreasonable to expect the Strat AI to be upredictable each time a designed and hinted scenario is played over and over again.[/QB]

I have to disagree with that; since even the simplest random chance (50/50) could be used to sway the Strat AIs choice at key branches in a way that makes the route planned by the Strat AI different each time.

From a user perspective, the simplest method would be to allow a type of hint that is random as to whether it is followed or not.

There are more complex methods and other options (like being able to place randomly weighted objective points) that could also enhance the replay value of a designed scenario.

None of those options would make a scenario entirely random, but that's fine. Of course, some scenarios should play out the same way each time, but there's no reason that a system couldn't be put in place to allow the scenario designer the flexibility to provide options to the Strat AI.

[c3k]

Ooops, wrong thread.

Carry on.

Ken

[Cpl Steiner]

Originally posted by Cameroon:

None of those options would make a scenario entirely random, but that's fine. Of course, some scenarios should play out the same way each time, but there's no reason that a system couldn't be put in place to allow the scenario designer the flexibility to provide options to the Strat AI.

I agree entirely. The system doesn't have to be perfect but a nod to replayability would be very welcome.

I would imagine this would only be necessary when the computer controlled player was on the offensive, as even in CMx1 the computer is quite capable of setting up an adequate defence that, to my knowledge, isn't the same every time.

It all depends on how heavily scripted a scenario becomes in CMx2. I would prefer hints given to the AI rather than the "squad A, move here at 10:45 and remain for 2 minutes" kind of thing.

[juan_gigante]

Something like "This area has high value" at the simplest, and you could take it to "Counterattack through that valley" is what I would look for.

[Battlefront.com]

We are planning on putting in a little bit of variability, but it isn't a high priority for us. I'm sure most people would rather have a single, tough challenge than several replays of so-so challenge. We were never big on scenario replayability concerns, so I think it is fair to say that is still true.

Also remember, the AI is not scripted to this stuff... it is just there for it to use, or not use, as it sees fit. So even if we had a single set of varibles for the AI to use, based on how the scenario goes it is not for sure going to do what the designer expects. It's just more likely to.

Think of the scenario design stuff as crib notes rather than a script.

Steve