Brand New Platoon Comands for CM:WWII?

[Cid250]

Hi.

One of the things that worked worse with any AI design, is the handle of the bigger units as a whole, like Companys and Platoons on the field.

May be is too much complex to develop "platoon" AI, or "company" AI... but you can't expect good results with AI in attack missions if all the AI weight his options based on the squads paths as individual units.

That's not only a benefit for the Computer AI, but it also can add more to the players that doesn't want to micromanage everithing at any time (specially if the unit is not involved in combat).

The system can be built with a new and different set of comands only available to the Platoon HQs and Company HQs. If you give a platoon level comand to a Platoon HQ, then all the units attached to that platoon will move and follow the general objetive.

Examples of Platoon Comands:

March:

Squads are moved as a single row, one after another, with a separation of 50 meters between squads. The advanced guard is formed up by detaching a squad, and that half-squad moves ahead for recon aproximately 600 meters ahead (as a patrol). Any attached HMG goes to the rear of the column.

Aproximation:

Squads increase security and deploy in a wedge formation, increasing the separation in between squads in depht and width to aproximately 150 meters. The recon advanced guard is increased/reinforced adding the second half-squad as a second element in the forward, but with enought separation to the first patrol. Any attached HMG or mortar increases his separation to at least 200m in the rear of the formation for security reasons.

Air warning / Artillery danger

This comand makes that each man in a squad doubles his separation, and also the squads double his separation, due to possible enemy air presence, or for security reasons when you expect to cross an area beat by enemy artillery or mortars. (This comand works just like a switch, or the button up comand in tanks, but for infantry).

Attack

Any attached HMG or mortar finds a place in the rear with line of sight of the target waypoint (and maximun terrain height), if that is not possible within a circular area of 100 meters (with his current position as center), then those heavy weapons deploy close to the starting path of the attack (at the maximun terrain height possible at that place). Once the heavy weapons are in position, the squads starts moving in bouncing and overwatch, providing cover to each one (if one squad moves the others fire). When the full platoon sucess in moving forward... If the heavy weapons attached are more than one, they also proceed to move forward half of his strenght to a new position as soon as the platoon is about to reach the maximun effective range of those weapons, or also in the event of a new attack order placed over a terrain that has no Line of Sight from the current deployment.

Assault

Is an Attack but the bouncing overwatch is modified to keep more squads on the move simultaneously, and less squads in the firing support role. The intended result is a faster move but with less firepower.

And much more... Delaying action, Pursuit, etc.

Overview

The current 1:1 model, handle each individual soldier of a big group, that's already Done by the Tactical AI. This is the next step up, considering each squad as an element of the platoon.

The benefits are that you can use this new option at your risk, some player will love the micromanagement, but other will delegate responsibilities in the platoon HQs at least in the earlier phase of any battle... you can do micromanagement again when the situation doesn't fit to any of those platoon comands.

You add a new depht to the value of HQ units, since if they are destroyed, you lose the chance to use those platoon comands.

The computer AI, can use those new commands to a better result of the support weapons in attack... and AI can easily switch automatically from March to Approximation when the advanced guard meets the enemy. And if an enemy mortar barrage starts, your platoon then turn "ON" the Artillery Danger to spread out more in the Approximation... and finaly when more enemies are spotted, they can launch an Attack.

The comands can be chained in a natural order, and can be scripted with triggers, switching from one comand to another when an event happens. That can make the AI plans look more natural, like a real unit following his training regulations or combat experience.

If you can go even more deeper in the simulation, then you can go straight to the historical national training manuals or field regulations, to handle different options for british, germans, americans, etc...

Please, discuss....

[Battlefront.com]

Higher level AI, which coordinates the actions of units, is extremely easy to design on paper and very straight forward. I have dozens of pages of designs for AI that I've made over the years. The problem is that a one paragraph description or a single feature can mean 3 months of programming.

There will be improvements made to the AI player over time, but it will always be "sucky" in the eyes of almost all players. Which we're happy with, quite frankly, because usually gamers think far worse things about AI than just "sucky"! In other words, "sucky" is actually a compliment because it's about as good as it gets

Steve

[Vergeltungswaffe]

When there is a tactical AI worth playing against, we'll have to start worrying about Skynet taking over.

[Pvt. Ryan]

There will be improvements made to the AI player over time, but it will always be "sucky" in the eyes of almost all players. Which we're happy with, quite frankly, because usually gamers think far worse things about AI than just "sucky"! In other words, "sucky" is actually a compliment because it's about as good as it gets

Steve

You should print that right on the package - "Now With Sucky AI!"

[dan/california]

Steve is correct, sucky is all that can be hoped for in terms of AI.

Which is why co-play and TCIP WEGO need to be higher up the list. Complete with an easy to use lobby attached to this forum. He did just say it would be less work.

[SlapHappy]

I think Cid is referring to broader "macro" commands that can be issued at the Platoon level and then are carried out by the TAC AI. He's not referring to the enemy AI.

On hindsight, it appears he might be referring to BOTH.

[Battlefront.com]

SlapHappy,

At first I thought that was what he was asking for and actually wrote up a different response. I reread the post and decided he's talking about the AI player. But rereading it AGAIN I can see he's talking about both. Wish I hadn't deleted my original response

Having higher level commands for the Human to use is definitely out. We're not ever going to try to do that. Why? Because the amount of AI programming is (as stated above) huge and the player's expectations for perfection (make NO mistakes that is exactly what customers will expect) completely dwarf his expectations for the AI Player.

Put another way... we don't think we can make an AI that the player will respect in the hands of the AI Player. It's just not practical. It's even less practical to make such an AI for the player's own commands. It's basically a suicide mission for us that would take months away from development.

And don't even try to tell me it's done well by another tactical wargame, because that would be a bold faced lie In fact, it's hard to find any other tactical game of ANY sort that does this sort of higher level unit management. The ones I can think of are also more-or-less RTS which is comparatively easy to program for and player expectations are inherently lower.

In short... for purely practical reasons this is not going to happen. Those who disagree are welcome to go make their own games with really great higher level AI working on behalf of the Human player. Then when the gaming audience rewards you with high praise and sales proportional to the development effort, I'll gladly say "I am not worthy of being in your presence." It's a safe challenge for me to make

Steve

[Uedel]

well all i realy hope for the WWII CMX2 Whatever Gameset is following:

-if on defense the AI should NOT shuffle arround Troops needlessly after deployment phase, i hatd it when i played vs the AI in CMX1 and the AI decided to dance arround the troops on the Battlefield. Heads down everyone should be the maxime of the AI. Deploying troops and setting them to good positions is done in deployment not in middle of battle.

- if on Attack: Mortars and HMG´s etc are not meant to walk towards enemy Trenches, thats what squads are made for, i very often seen the AI Walk with happy weapons right into known postions of my forces where they got slaughtered. if they even halfassed could seek overwatch positions behind the Advancing squads it would be a huge step towards a more challanging AI.

[Redwolf]

Well, technically speaking formation commands that are executed well are possible in a computer game. You just have to dedicate your whole game and all your energy to it And do a 2G game with no 'no-go' terrain - such as Airborne Assault.

What would be nice/possible would be to have

• Formation movement commands
  
• Replace some TacAI with SOPs
  
  
  

But worth the time of the only coder at this time? Probably not.

[Thomm]

... with no 'no-go' terrain ...

What about water?!

Best regards,

Thomm

[Battlefront.com]

Redwolf,

Well, technically speaking formation commands that are executed well are possible in a computer game.

I don't know of any other games with the tactical level of detail CM has that also has successfully implemented higher level "formation" commands. By successfully I mean something which people don't endlessly gripe about.

You just have to dedicate your whole game and all your energy to it

Which can be done exactly one time unless people are willing to pay for it Many, many years ago I remember us having a lengthy discussion about a "command level" game experience for CM. Where you, the player, would sit back and issue higher level commands to nothing smaller than a platoon (under normal circumstances). You could even give commands at the company level and the AI would handle everything below.

Even *if* we could pull that off as a commercially viable product, and I'd argue we can't, how many people would actually value that degree of abstracted control over their tactical assets? The discussion proved what we already knew... a small number would LOVE it provided the AI did a very, very good job. If the AI was even mediocre they would hate it. The majority would reject it even if the AI was near perfect simply because that's not what jazzes them about wargaming. Which means there's no way to make a commercially viable product out of such an idea even if we somehow managed to have the resources to get all the normal CM stuff in AND AI that we don't think exists anywhere else on the face of this Earth.

Put another way, if some billionaire said to us "I'll give you $10,000,000 and 5 years to make the perfect tactical wargame with higher level AI controls, and I don't care if I'm the only one that buys it" we might very well be tempted to do it I'm sure that we could produce a product that would knock people's socks off AND profit well from it. We have the inherent skills and the money would be high enough that we could hire people to keep our viable commercial work going while we catered to the rich and foolish investor

What would be nice/possible would be to have

• Formation movement commands

• Replace some TacAI with SOPs

We've discussed SOPs at length over many years. We like them in theory, we don't like them in practice because the UI is inherently a distraction from the rest of the game.

The truth is CM doesn't "need" any of this. In a perfect world it might be better for having such options, but everything comes with a pricetag. We don't think anybody, you guys or us, would be happy with the results if we started losing our focus and introducing "Scope Creep" into CM. CM is inherently a very low level tactical wargame, not anything else. Period.

Steve

[Cid250]

SlapHappy,

At first I thought that was what he was asking for and actually wrote up a different response. I reread the post and decided he's talking about the AI player. But rereading it AGAIN I can see he's talking about both. Wish I hadn't deleted my original response

Having higher level commands for the Human to use is definitely out. We're not ever going to try to do that. Why? Because the amount of AI programming is (as stated above) huge and the player's expectations for perfection (make NO mistakes that is exactly what customers will expect) completely dwarf his expectations for the AI Player.

Put another way... we don't think we can make an AI that the player will respect in the hands of the AI Player. It's just not practical. It's even less practical to make such an AI for the player's own commands. It's basically a suicide mission for us that would take months away from development.

And don't even try to tell me it's done well by another tactical wargame, because that would be a bold faced lie In fact, it's hard to find any other tactical game of ANY sort that does this sort of higher level unit management. The ones I can think of are also more-or-less RTS which is comparatively easy to program for and player expectations are inherently lower.

In short... for purely practical reasons this is not going to happen. Those who disagree are welcome to go make their own games with really great higher level AI working on behalf of the Human player. Then when the gaming audience rewards you with high praise and sales proportional to the development effort, I'll gladly say "I am not worthy of being in your presence." It's a safe challenge for me to make

Steve

Understood.

Mainly i was thinking in the broad AI available in COTA (Conquest of The Aegean), that game was done by only 2 coders, and handles the concept at the "operational" level in a 2D map with abstract elevation model and lines of sights.

The feature is very usefull to avoid micromanagement when you still doesn't expect contact with the enemy... what is enought usefull when you have more than 20 units in the field and you want to move them in an organiced way without selecting all to give a general comand.

Selecting all units to give a "move" comand, doesn't allow to keep correct spacing between squads and slowly support weapons. And by sellecting all you also can't expect to draw a "90 degree turn" in a move comand, keeping a good shaped formation (with the units in the inner side of the turn, being intentionaly slower than the units deployed in the outer side of the turn path)

Can we simplyfy it to just one "move" comand at the platoon level that allows us to keep current spaces between fast squads and slowly support weapons (attached to that platoon)?. Is a syncroniced turning in formation too much code?. Keep in mind to be used only in the early stage of any battle... when there is no contact, or when you get reinforcements that need to travel a long distance to reach the main action zone. As player, you can use this feature at your own risk if micromanagement of paths is boring for those first minutes, or when reinforcements enter far away of the zone already controlled by your troops.

[dan/california]

That little "2D abstract" bit changes the coding involved by at least two orders of magnitude. It is not even a discussion about the same thing, AT ALL.

[akd]

Although you command on a 2D map, elevations and LOS are real in the Airborne Assault engine, not abstracted height bonuses and such. The devs (now working on the 4th in the series, Battles from the Bulge) have even mentioned that a 3D view is on their list of future enhancements so they can stop struggling with 2D representation of 3D terrain (although the fan community has provided a wonderful stop-gap in the form of Cota of the Aegean maps that you can open in Google Earth and view in 3D).

There is a great deal of synergy between CM and the AA series. I have often wished they could be combined into one Ultimate Wargame. The two teams' approaches to development are remarkably similar.

Our approach has always been to work from the ground up. Perfect the base game mechanics like movement and combat, layer onto this the ability to develop basic tasks and then develop complex plans and so on. Once the engine can simulate just about any WW2 operation, then I reckon we should look at things like multi co-op play.

http://www.matrixgames.com/forums/tt.asp?forumid=414

A very interesting Powerpoint presentation on AI development given by Panther Games' Dave "Arjuna" O'Connor (the Australian Steve) can be found here:

http://hosted.wargamer.com/dropzone/articles/index.html

[Battlefront.com]

dan/california,

That little "2D abstract" bit changes the coding involved by at least two orders of magnitude. It is not even a discussion about the same thing, AT ALL.

Bingo AI is bad enough when comparing apples to apples, but comparing the AI of an orange to the AI of an apple is a complete waste of time. Doesn't work at all.

AKD,

Although you command on a 2D map, elevations and LOS are real in the Airborne Assault engine, not abstracted height bonuses and such.

Terrain resolution is of massive importance when talking about higher level AI controls. Why? Because it basically comes down to scale of resolution. In CM it matters if your guys are in front of a wall, behind a copse of trees, below a ridge by 1m, etc. Getting the AI to coordinate individuals within such an environment is magnitudes more difficult than getting an abstracted unit to move over abstracted terrain. Even CMx1's terrain was far less abstracted than Airborne Assault's. Which makes sense because CM is at a level or two lower than Airborne Assault, so CM's terrain resolution is inappropriate for a game at AA's level. Just as their terrain abstractions are inappropriate for a game at CM's level.

Steve

[akd]

dan/california,

Bingo AI is bad enough when comparing apples to apples, but comparing the AI of an orange to the AI of an apple is a complete waste of time. Doesn't work at all.

AKD,

Terrain resolution is of massive importance when talking about higher level AI controls. Why? Because it basically comes down to scale of resolution. In CM it matters if your guys are in front of a wall, behind a copse of trees, below a ridge by 1m, etc. Getting the AI to coordinate individuals within such an environment is magnitudes more difficult than getting an abstracted unit to move over abstracted terrain. Even CMx1's terrain was far less abstracted than Airborne Assault's. Which makes sense because CM is at a level or two lower than Airborne Assault, so CM's terrain resolution is inappropriate for a game at AA's level. Just as their terrain abstractions are inappropriate for a game at CM's level.

Steve

I wasn't suggesting equivalence, but only countering the claim that it is a "2D" environment. Obviously scale and resolution are different, and the AA A.I. is tailored to that scale. Personally, I feel CM picks up at just about the level where AA stops, thus "synergy" rather than "similarity."

You could take a snapshot of 1-9 cell grid (1-3 square kilometers) in an AA battle and pretty easily create a CM battle. Terrain resolution is down to the meter, so it would really just be fleshing out the lower resolution details, e.g. buildings, walls, foliage, etc. Once the later CMx2 WWII modules and Battles from the Bulge are both out, the latter might provide an excellent platform to run a meta-campaign.

[Battlefront.com]

Ah! Got you. From our perspective a game like AA is 2D. Yes, it has more terrain level subtlety than most 2D games, but it's still the same thing as Hex 2,3 is at Y height, Hex 5,8 is at Z height, then trace LOS from the two hexes to see if something inbetween is blocking LOS. This is the stuff found in tactical wargames going back decades. What AA brings to the table was having a far more "organic" treatment of height and terrain. Still very much 2D, just "better" in terms of execution.

This reminds me of way back when Charles told me he wanted to do CM in 3D instead of 2D. I basically described the AA style map system (which I first encountered with Chris Crawford's "Patton vs. Rommel") and said "what's the difference between that and a 3D?". Well, a ton of things However, almost none of them are important for a game at AA's level of command while they are critically important at CM's level. But it's still correct to say one is 2D and the other 3D.

Agreed that for people looking to play a game at both operational and tactical levels, a mating between a game engine like AA and CM would be ideal. Ideal because to do that sort of environment right you really are talking about requiring two entirely unique games with full development support behind each. That's not something we're capable of doing, obviously. Working with another developer to achieve the level of interaction between the two is a pile of work on its own. We don't have the resources or incentive to pursue it.

Steve

[SlapHappy]

Put another way, if some billionaire said to us "I'll give you $10,000,000 and 5 years to make the perfect tactical wargame with higher level AI controls, and I don't care if I'm the only one that buys it" we might very well be tempted to do it I'm sure that we could produce a product that would knock people's socks off AND profit well from it. We have the inherent skills and the money would be high enough that we could hire people to keep our viable commercial work going while we catered to the rich and foolish investor

Steve

Well, there is a certain ex-major league baseball player who could possibly meet that description, but that doesn't mean he's going to hand over $10,000,000 just to satisfy a whim.

P.S. Also, wouldn't he kinda rate a free copy or two under those circumstances?

[akd]

Ah! Got you. From our perspective a game like AA is 2D. Yes, it has more terrain level subtlety than most 2D games, but it's still the same thing as Hex 2,3 is at Y height, Hex 5,8 is at Z height, then trace LOS from the two hexes to see if something inbetween is blocking LOS. This is the stuff found in tactical wargames going back decades. What AA brings to the table was having a far more "organic" treatment of height and terrain. Still very much 2D, just "better" in terms of execution.

This reminds me of way back when Charles told me he wanted to do CM in 3D instead of 2D. I basically described the AA style map system (which I first encountered with Chris Crawford's "Patton vs. Rommel") and said "what's the difference between that and a 3D?". Well, a ton of things However, almost none of them are important for a game at AA's level of command while they are critically important at CM's level. But it's still correct to say one is 2D and the other 3D.

Agreed that for people looking to play a game at both operational and tactical levels, a mating between a game engine like AA and CM would be ideal. Ideal because to do that sort of environment right you really are talking about requiring two entirely unique games with full development support behind each. That's not something we're capable of doing, obviously. Working with another developer to achieve the level of interaction between the two is a pile of work on its own. We don't have the resources or incentive to pursue it.

Steve

Hah, not an actual suggestion, just a fantasy in my brain. Anyways, you'd obviously have to move to Australia, and who wants to do that?

[Battlefront.com]

AKD,

Hah, not an actual suggestion, just a fantasy in my brain. Anyways, you'd obviously have to move to Australia, and who wants to do that?

New Zealanders? Wait, wrong answer. Uhm... well, there's a head scratcher there for sure. The good news is that Australians are perfectly happy with the fact that nobody wants to move in. I have the same feeling where I live. Let them visit and drop all kinds of money, then let them go back to wherever they came from

Steve

[LRC]

Steve,

Is it possible to clarify the following " for the game AI, what is the difference between [a set of movement orders issued by the player ] and [ a set of movement orders issued by a group movement GUI] ?

If difference is minimal, group movement should just be considered to be way a to automatize waypoints creation. But because it would be imperfect, individual waypoint editing would become a must....

[Battlefront.com]

Hmmm... not quite sure I understand what you want clarified, but here's my best shot...

There are many different things which we all collectively think of as AI, but in coding terms they are completely different concepts. Here's a list of the most important ones:

1. PathingAI - responsible for figuring out how to get a unit from A to B. Involves negotiating the terrain and determining when to use optimize the "best path" for speed or for cover. Often the results are very different.

2. TacAI - lowest level of unit behavior, which is itself a collection of smaller sub-routines which kick in on an as-needed-basis. Things such as determining when to duck, when to switch a player Command to something else, when to panic, when to risk exposure to engage the enemy, what type of ammo to use, etc.

The TacAI often calls upon the PathingAI to help it achieve it's goals. For example, "this unit is taking fire here and need to move. This other spot over here looks good. PathingAI, please take this unit from here to there emphasizing speed over cover".

3. OperationalAI - for the AI player only, this determines coordination of groups of units in order to achieve some sort of goal. In CMx1 this AI was completely on its own, in CMx2 it's using scripted behavior to influence what it does and when it does it.

4. StrategicAI - for the AI player only, this determines overall strategic use of units to achieve the overall goals of the scenario. Like the OperationalAI, in CMx1 was generic and quite limited. In CMx2 it is mostly following the script that the scenario designer sets up for it.

Now, with that in mind I'm going to give a shot at answering your question

When the player puts down a Movement Command, no matter if individually or as part of a group, none of the AIs are involved initially. There are some simple binary checks to see if a waypoint can be placed where you clicked, nothing more than that. After a successful waypoint has been placed, the PathingAI takes over to get your units to their waypoints. The TacAI kicks in as needed, but otherwise there's no other AI going on.

Did that answer the question?

Steve

[LRC]

Steve,

Here is a respin on the group command question.

Consider the sequence for a group of 5 units "

1{deploy in V formation oriented North} + 2{move 400 meters North } + 3 {deploy in compact line oriented West}

It is possible to code this sequence of orders by giving each unit a sequence of waypoints but it is not practical.

With this in mind, group command should only be considered to be a way to simplify waypoints editing (like {deploy in V formation oriented North} or {deploy in compact line oriented West}) .

[Battlefront.com]

OK, got you. What you're asking about is similar, but not the same, as what the original poster was asking about. You're request is to have "formation commands" without additional behaviors, such as those described on page 1 of this thread. While much easier to do in theory, it's just about as hard to do.

The problem is, once again, with AI coding. The AI has to understand what a "V formation" is, what sort of spacing it should have, when it should have it, when it should abandon it, what to do if abandoned, how to handle insufficient/excess elements within the formation, etc. It's a ton of logic that has to be written, debugged, tested, debugged, etc. And that's just for a single formation type

On top of what I just mentioned, there has to be a realistic treatment of a formation. Such as how easy/difficult is it for each individual within the formation to keep formation. This is behavior which, basically, determines how much a single unit should deviate from the theoretically optimal positioning. This is an absolute MUST, otherwise you'll have Conscript tank crews keeping a perfect Echelon Right formation in rough, broken terrain while going top speed and being engaged by the enemy.

In order to do this each element of a formation must have it's own set of instructions and communicate with some sort of central AI specific to that formation. The centralized AI decides how to handle the formation as a whole based on how the individual components are doing. The individual components, in turn, must constantly try to carry out the instructions from the centralized AI.

On top of that, we will need new UI for higher level commands. That itself takes time to develop, test, tweak, and debug.

It's a huge mess and we're not going to touch it with a 10' (3.048m ) pole. Massive time sink we simply can not afford to get tangled up in.

Steve

[-Pv-]

I can understand not wanting to tackle formation command UI and AI. It is however something I miss from M1 Tank Platoon. I thought it was done as well as could be for an inexpensive game several years ago.

Urban maneuver I would think to be so difficult as to be nearly impossible. I miss the reduced workload for the player at least when my units were not engaged in heavy fighting and I was mostly working to get a large number of units in position.

-Pv-