So... AI in wargames (Full Version)

All Forums >> [General] >> General Discussion



Message


Perturabo -> So... AI in wargames (3/31/2011 10:33:24 PM)

I bought two FMs (a in infantry battalion and infantry company manual) a few months ago and I have found it interesting how there's a ready procedure for almost every normal situation in Wargames and how much stuff gets placed on maps that is supposed to help commanding the troops.

It makes me wonder if it would be possible to write an AI (and generally a whole command/information/scenario creation system for a wargame) that would be based on that stuff.




KG Erwin -> RE: So... AI in wargames (4/1/2011 5:00:15 AM)

Unfortunately, a good AI is the toughest part of programming, which is why you often see it being sacrified for the sake of cute graphics and multi-player involvement. While not a wargame, Shaun Sullivan's baseball game (Pure Sim Baseball 4) has one of the best AI engines in the business. If only the wargame designers would follow his example. As a working man, I don't have the time or inclination to get involved in an elaborate online multi-player megamonster. I simply want a good historical game that will hold my interest.




jomni -> RE: So... AI in wargames (4/1/2011 5:01:45 AM)

Most game developers have given up the search for the AI holy grail and instead focuses more on multiplayer since HI is always better than AI.




sterckxe -> RE: So... AI in wargames (4/1/2011 8:16:30 AM)

quote:

ORIGINAL: jomni
Most game developers have given up the search for the AI holy grail and instead focuses more on multiplayer since HI is always better than AI.


Most, but not all. The games produced by Panther Games and SSG have a wickedly clever AI

Greetz,

Eddy Sterckx




Perturabo -> RE: So... AI in wargames (4/1/2011 9:39:22 AM)

quote:

ORIGINAL: sterckxe

quote:

ORIGINAL: jomni
Most game developers have given up the search for the AI holy grail and instead focuses more on multiplayer since HI is always better than AI.


Most, but not all. The games produced by Panther Games and SSG have a wickedly clever AI

Greetz,

Eddy Sterckx


Panther Games games seem to be the only ones I played so far in which I saw an AI that would have some semi-reasonable organization.




Adraeth Montecuccoli -> RE: So... AI in wargames (4/1/2011 12:01:07 PM)

Even Crown of Glory Emperor's edition has a good AI on a strategic level, on tactical level it is a good opponent [:)]

Anyway just focusing on tactical (or grand tactical) aspects i presume are the main goal of the post, it is true that AI is often bad.

Sometimes, if a scenario editor is present, editing scenarios in some way can give the feel that AI is a lot better.

For example in Close Combat Modern Tactics i managed to make scenarios in which the AI is near objective points, and the same has some different deploy zones so the player can replay the same scenario watching different approach from the AI.

I like modern combat.... and i really liked the idea of Combat Mission Shock Force, however even if i can understand the need to make scenarios with a scripted AI ... in the end there is not so much space of decision making by the AI.

In my opinion and experience the AI just need tons of different strategies ... then "she" have to choose one of them and, maybe, changing the strategy after a given time (or time cycles) or a given "if":

A Sort of If Objective X is AI controlled then Defend in Line/Column/Echelon/Ambush

or

If Objective X is Player controlled then call artillery/Assault... and so on

The more the lines of code the best the AI...

indeed can be a proving and boring task




Yogi the Great -> RE: So... AI in wargames (4/1/2011 12:21:56 PM)

One factor for at least better AI is just remove the stupid factor.

Artillary that moves right up to enemy lines
Leaving units in the open to be bombarded or shot at until shattered/routed before moving away
Ignoring Victory point hexes
Attacking strong enemy lines with one or two units time after time
Failure to form reasonable defensive lines and/or take advantage of good defensive positions
Failure to withdraw apparently believing that one unit can withstand massive assaults all alone.
Or to stay simple, just remove AI tendency to be suicidal and stupid

I understand that good AI that can properly respond to any situation is a difficult task. It does seem though that some of the aspects of "stupid/suicidal AI" is just not caring enough when developing the game. Considering that a large part of the gaming community mainly plays solo against the AI, a good AI is a feature that could sell a large number of games.




Jeffrey H. -> RE: So... AI in wargames (4/1/2011 5:05:39 PM)


quote:

ORIGINAL: jomni

Most game developers have given up the search for the AI holy grail and instead focuses more on multiplayer since HI is always better than AI.


And this is not a good development IMO. I sense the topic is simply being abandoned by indy devs as too much to manage in the face of everything else.





pzgndr -> RE: So... AI in wargames (4/1/2011 5:25:00 PM)

Brit, the Empires of Steel developer over at Battlefront.com, recently posted news of his AI development plans:

quote:

At this point, I'm playing around with the idea of creating a generalized strategy-game AI. I've looked around for this kind of thing on the internet, but haven't had any luck. The closest I've seen to a strategy AI is that there's a handful of pathfinding AIs, but nothing that handles the larger picture. It seems like there might be a demand for it. If there's some demand to license the technology for other games, then there's some additional reasons to work on an expanded AI system.

If I'm going to make a generalized, licensable AI system, I'll have to add a scripting mechanism. I don't know what I'll use. I think LUA has been pretty popular as a scripting language, but haven't looked into enough to make a decision about it yet.


Optimistically, a project like this that can be improved and enhanced over time should produce a very good generic AI, at least something decent enough for most indy developers to start with. Short of new 3D graphics effects and ever increasing complexity in wargame details, it would be nice to see some serious development work done on AI at all levels - tactical, operational and strategic. [8D]




Jeffrey H. -> RE: So... AI in wargames (4/1/2011 7:57:18 PM)

That's a smart approach IMO, do something big and license it. That way the costs can possibly be recouped. Kudos to Brit.




simovitch -> RE: So... AI in wargames (4/1/2011 10:00:08 PM)

I found in many wargames including Panther Games that scenario design has significant influence on how the AI responds. Even the best designed AI's will fall on their face if the scenario designer lacks a basic understanding of how the AI responds.

The AI in John Tiller's Civil War and Napoleonic series (Talonsoft and HPS) could be a lot better if the designers would concentrate less on producing massive numbers of 'what if' scenarios and more time on scripting and timing and playtesting a few historical ones. I've seen whole divisions (3 brigades and artillery) just sit there on the entry hex through the entire game because there was no script written for it.




Perturabo -> RE: So... AI in wargames (4/1/2011 11:15:21 PM)

Well my post wasn't really about the amount of work put into the AI.

From what I've seen, a lot of stuff in FMs is basically algorithms written in natural language and there is a lot of stuff that is supposed to assist planning and execution of plans. From what I've seen, almost none of these is used in wargames.




Lieste -> RE: So... AI in wargames (4/7/2011 12:43:35 AM)

Actually the closest is probably 360's Patriot. It was of limited appeal due to the one-sidedness, but had 'natural' presentation of the control measures, and force organisation...

...although the implementation left a lot to be desired (reserves and HQ would usually lead an assault unless manually told to defend, dismounted attacks would never 'remount' after they had begun (and mounted attacks by infantry would always behave like dismounted ones - meaning that the armour would romp ahead of the mech infantry, there were a few really silly data errors that needed to be fixed before the mission would play even vaguely sensibly (enemy 152mm howitzers that were non-data and behaved like A10s were the worst), equipment held by HQ wasn't available for surrender or fire casualties, and would fire at support elements and air routinely).

It did have "scenario editing" tools, but they were partially disabled, poorly documented and awkward to use. The one that I remember most vividly was being unable to edit a scenario - you instead had to make a 'saved game' and edit that - and then remember not to play the scenario but to 'find' your edited version... bonkers.




Perturabo -> RE: So... AI in wargames (4/11/2011 10:33:02 PM)

Is Patriot worth buying? Are there any fixes to these bugs?




Lieste -> RE: So... AI in wargames (4/12/2011 9:28:58 PM)

I doubt you can find it to buy still... It was mid 90s and floppy only... My floppy copy is deceased, but the installation is still good on the HD - works fine in XP on my desktop, and on my older laptop too.
Most of the bugs can be 'worked around' with mild editing of the save game (or major surgery for the HQ equipment - this I haven't attempted yet, as I can have as much 'fun' doing similar things to more modern, well rounded systems), or with 'aware' settings of posture during play - (ie intermittently set reserves/HQ from their march/attack postures needed to stay in touch with front line troops, to a defend posture to avoid an unintended passage of lines - this is a PITA, but it does 'work'). Good (simplistic) presentation of control measures (mostly just an alternative presentation of movement arrows AFAICT, but it looks the part) - the OOB/attachment UI is clean and easy to use... just overall it needs to be more robust and simulate a balanced war - maybe the Marines (or French/Arab) fighting is better than the Breach/Medina/73 Easting scenarios, but I haven't run every one still.
It is a fairly one sided war - when a US Tank Battalion pushes to 'close range' of an Iraqi Battalion it is common to see 30-60% casualties within a single 15 minute pulse, usually for no losses... Ok for a couple of play-throughs, but ultimately perhaps not the most exciting and tense experience...

I'm having *a lot* more fun with a modified BFTB... Still, I wish I could constrain attack plans to unit sectors and have phase lines and a defined 'ours' and 'theirs' side of the nominal FLOT (even better if it could be very out-of-date, in a horribly surprising way [:D])




Perturabo -> RE: So... AI in wargames (6/30/2011 9:50:57 PM)

I bought Patriot. I'm reading the manual now.




Lieste -> RE: So... AI in wargames (7/1/2011 2:44:20 AM)

I'll look out the scenario and unit info that was suspect (IIRC it was D30 batteries in the Medina/73 Easting battle - but probably rife throughout the OOB where present).

I can send a save-game from the scenario start with the corrected gun data, &/or describe how to fix it yourself to get rid of it as you see it.

It results in A10 strikes from his 152mm howitzers, and 'perfect' recon for his side (assuming you play as Coalition of course ;)

A few weird 'bugs' in the UI. Reserves/HQ/Artillery will overrun infantry based 'dismounted' attacks. They must be set to halted most of the time, and allowed to resume their operations when they are dropping a bit behind - a pain, but manageable. I found it hard to get a dismounted attack to remount and move at normal rate, and for infantry to stay mounted. This isn't often a huge problem, but does mean a slower pace of advance. You might also want to pause briefly as you approach the enemy positions to give enough time to destroy them with a mounted armoured attack - other wise you will find his force density is sufficient survive the overrun, and take casualties from the close range fight as you pass through.

Try not to pass units through each other when under artillery fires, or when in contact with the enemy - it takes longer to close/break contact than if they don't overlap, and casualties may be higher.

If you enjoy the concepts, but find the implementation a bit iffy, then look at the Airbourne Assault/Command Ops series (there are demos of both RDOA and BfTB). They aren't exactly the same, but the simulation of limited direct control & OODA loops are good.




Perturabo -> RE: So... AI in wargames (7/2/2011 4:48:01 AM)

quote:

ORIGINAL: Lieste

I'll look out the scenario and unit info that was suspect (IIRC it was D30 batteries in the Medina/73 Easting battle - but probably rife throughout the OOB where present).

I can send a save-game from the scenario start with the corrected gun data, &/or describe how to fix it yourself to get rid of it as you see it.

It results in A10 strikes from his 152mm howitzers, and 'perfect' recon for his side (assuming you play as Coalition of course ;)

A description would be nice.

I have noticed that this game is very moddable. I already replaced Iraqi forces with line infantry and cavalry with naval forces (4 land dreadnoughts and 12 land destroyers instead of cavalry troops + replacement with most of former cavalry troops with naval infantry).

It's amazing that a game with such possibilities of customization is practically unknown.

quote:

ORIGINAL: Lieste

A few weird 'bugs' in the UI. Reserves/HQ/Artillery will overrun infantry based 'dismounted' attacks. They must be set to halted most of the time, and allowed to resume their operations when they are dropping a bit behind - a pain, but manageable. I found it hard to get a dismounted attack to remount and move at normal rate, and for infantry to stay mounted. This isn't often a huge problem, but does mean a slower pace of advance. You might also want to pause briefly as you approach the enemy positions to give enough time to destroy them with a mounted armoured attack - other wise you will find his force density is sufficient survive the overrun, and take casualties from the close range fight as you pass through.

Try not to pass units through each other when under artillery fires, or when in contact with the enemy - it takes longer to close/break contact than if they don't overlap, and casualties may be higher.

Thanks for the tips.

quote:

ORIGINAL: Lieste

If you enjoy the concepts, but find the implementation a bit iffy, then look at the Airbourne Assault/Command Ops series (there are demos of both RDOA and BfTB). They aren't exactly the same, but the simulation of limited direct control & OODA loops are good.

I already have a full version of RDOA (and Fire-Brigade) but I can't really get into it for long. I'd like to play it if it would be set in France 1940. But Arnhem? BftB is way too expensive for me.




furryCuds1z -> RE: So... AI in wargames (7/7/2011 5:17:09 AM)

I think it's very hard to do AI. because the computer knows everything about you the AI could probably shoot you in the head everytime. This would not be good AI. The AI has to appear to be natural as if it is a human. That is what is so challenging.
_____________________________
Check out the Top Toys For Christmas 2011 like Lets Rock Elmo. Rock Star Mickey and Lets Rock Elmo should be in a band.




JW -> RE: So... AI in wargames (7/7/2011 3:57:03 PM)

The simpler the game, the better the AI.  The less the AI has to figure out the better, and the less opportunity for 'gamey' situations the better.  Chess AIs can be designed for example that are virtually unbeatable and are available for computers and phones and such.  




E -> RE: So... AI in wargames (7/7/2011 10:49:59 PM)


quote:

ORIGINAL: Perturabo

It's amazing that a game with such possibilities of customization is practically unknown.



It was well known in the 90's. But mostly for how disliked it was. As M Evans Brooks called it in his wargame reviews site...

"This simulation was to be the land version of HARPOON. A simulation of the Gulf War, it was bug-ridden and unplayable; as a military task force planner, it had merit and could serve as the Harvard Graphics of Wargaming. Thus, it is recommended for military professionals because there are concepts buried in this "game" that merit serious consideration (e.g., utilization of unit frontages and boundaries, cross-attachment of units), but as an entertainment program or one that teaches valid lessons of warfare, this was the biggest American disaster since Bataan."




walsh_Fever -> RE: So... AI in wargames (7/9/2011 11:53:45 PM)

AI is perhaps one of the hardest aspects to program. It's very easy to think of ideas but very difficult to implement in practice. Of course, that doesn't mean that it can't be done. It takes some of the best minds to collaborate to push the boundaries of AI in gaming.




E -> RE: So... AI in wargames (7/10/2011 12:15:36 AM)

I find it interesting that the Operational Art of War series (TOAW) stopped calling it an "AI" and called it what it really is... a "Programmed Opponent" (PO).

Of course, if any game had a real artificial intelligence, you'd hear more screams of "Colussus" or "Skynet!" or "Pre-Borg!" or "Replicators!" or even "Saturn 5!, where's Farrah?!" (...and I would be the first to run screaming into the mountains!)




Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.0234375