Jump to content

Who wrote this "fuzzy logic?"


Recommended Posts

I had to laugh when I read the AI uses "fuzzy logic" in the manual. Game AI has come a long way, and "fuzzy logic" is used so much now a days, haphazardly, that people don't even know what fuzzy logic is.

In fact, we all know AI is impossible (at least today). What has developed however is what is called "expert systems" which takes a narrow domain, and works only within that realm. Baysian Networks has proven very productive over the last 10 years in expert systems.

So, how can I find out more about this "fuzzy logic" used in CM? Who can I discuss the programming methods used, and what type of Game Theory was applied?

Thanks in advance!

------------------

Doc

Link to comment
Share on other sites

  • Replies 56
  • Created
  • Last Reply

Top Posters In This Topic

ROTFLMAO!

We can have an 800+ post on hamsters, a dozen 50+ posts on Brinnel numbers and mantels, but AI is boring.

I am now going to go to alt.binary.ai and begin a discussion on the proper August 3rd, 1945 camo pattern for British Airborne squads to see the other side of this case.

Link to comment
Share on other sites

I hate to "mitou" as they say here in France but me too, I'd find this really interesting if anyone knows some details about the AI. I'm not asking for any trade secrets, just a general overview of the "thought" processes used by the computer. Thanks.

DeanCo--

Link to comment
Share on other sites

try this:

http://www.battlefront.com/discuss/Forum1/HTML/000400.html

here it is

"Thomas Davie

Member posted 07-08-1999 09:20 PM            

------------------------------------------------------------------------

Slowly but surely I think I'm understanding this game as it is developing. Some points I'll support almost unto death, whilst others come pretty close to having my fist plunged through the monitor screen in frustration.

But, from the very beginning I've been puzzled by the distinction of AI between tactical and strategic...and now operational. I had always thought that 'there is the AI', and that is that. I can understand a distinction between reactive (how your or the computer's units behave once the turn initiates) and proactive (how the computer plans out the turn; deciding what is or is not a threat and where to send it's units,ranking or threat assessment and so forth)but this naming of different levels of AI seems, well...here is how I understand what will happen. Could Steve or Charles please comment?

1) Tom plots a turn and presses an end turn button or 'whatever' when he is finished.

2) Either simultaneous with Tom, or subsequent to Tom the computer is plotting it's turn. In computer chess, it usually happens ot be the former, which explains why I lose so much, frequently taking long turns and thus allowing the AI more time to plan/plot. Although, I suspect that the 'planning' AI in Combat Mission will not be making multiple passes at a solution....or will it?

3) Prior to playback, the ENTIRE turn is resolved; ALL of it. Sightings, combat resolution, morale checks, what have you. Everything. This may or may not be perceived as a delay by the player of the game. Here, the tactical AI or what I call the responsive/reactive AI governs both the computer and the human.

4) Humans can view the playback (eh, of course the computer has no need to do so). This is what will be perceived by a human player as the processing of the game turn, since of course the internal resolution of the turn by the computer is imperceptible to a human player.

5) Once a human player has viewed/reviewed a turn's occurence they then can plot another turn.

Now, here is where Tom starts to have a problem. I equate the AI of a game with what *I* call a proactive AI. I'm willing to believe that it can be done, but that it will take time. And yet reports that I've heard indicate that the operational/strategic AI has not yet been implemented into the game. How realistic is it that a good 'proactive' AI be implemented and the game still gets released by summer? Given of course that definition of the word 'good' is higly subjective. Sigh.

Of course, if the 'proactive' AI has been implemented already, but you're just, ah not talking about it, please ignore this whole post

Thanks for even reading this.

Tom

IP: Logged

Big Time Software

Moderator posted 07-08-1999 10:59 PM            

------------------------------------------------------------------------

Hi Tom,

Your points 1-5 are correct. So you are all set there Just to make sure that you, and everybody else understands the difference between the 3 levels of AI...

Tactical - governs the lowest level reactions of an individual unit. There is no coordination, just very selfish behavior of who to shoot, when to hit the dirt, when to panic, etc. Actions are all momentary.

Operational AI - gets the unit from A to B in some coordinated fashion. Say, moving a Platoon to a good flanking spot using leaps and bounds for example. But there is no concept of WHY it is a good idea to establish the flanking position; only that this is what it is supposed to do. Actions last until out of orders.

Strategic AI - figures out if attack is better than defense, where to do each, to what degree, with what units, etc. This is what gives the whole battlefield its sense of purpose. Actions last until the Strategic AI itself determines that things need to change.

To tie this altogether, the Strategic level figures that an attack should be made on the left flank using x units. The Operational AI then takes these "orders" and figures out how to move each unit to achieve this attack. The Tactical AI then reacts as things go along and target, speed up, hit the dirt, etc.

The order in which the AIs are being done is not accidental either. Each builds on the other. There is no real way to tell if the Operational AI is any good until there is a solid reactive AI underneath. Likewise, the Strategic AI can't do anything until there is logic to move around the units.

quote:

------------------------------------------------------------------------

How realistic is it that a good 'proactive' AI be implemented and the game still gets released by summer? Given of course that definition of the word 'good' is highly subjective. Sigh.

------------------------------------------------------------------------

Considering that most wargames spend about 3-4 weeks on AI TOTAL... a whole lot can be done in the time remaining. The Tactical AI (which we honestly feel is 1000 better than anything else anybody has ever done) took about 3 weeks total to do. This is the more complicated and important of the three AIs. The Operational AI has been started, but it is not finished at this point. The Strategic AI will be the easiest as there is far less involved (programming wise).

Already CM's Tactical AI, with zero coordination, can kick your ass on defense. If we even put in rudimentary coordination (i.e. 1 week's work) the AI would probably be 10x better than the AIs found in contemporary games. Partly because the foundation is so good already, but also because CM's AI doesn't use trigger points or other crutches (Fuzzy Logic rules!!). And as most people seem to agree, Charles's previous AIs in BTS' flight games are a definite cut above the rest.

We have no fear that CM's AI will be the best out there, and will soundly kick ass many a time. We expect that even Grogs will get a really painful lesson from the combination of realistic game system and superior AI. Will it never do anything stupid? Will it always lead an inspired battle? No, as no AI ever will. But will it be better than anything else out there, and will it be capable of kicking even a hardcore gamer's butt? You bet.

Steve

[This message has been edited by Big Time Software (edited 07-08-99).] "

I have never ever BLASTED anyone for not using the Search function BUT...

Am I the only one here who knows how to click "search"

and enter in " AI 3 Levels" ????

Seriously...

there are discusions of this nature here if you care to look for them.

-tom w

P.S. I am not Thomas and that was not my orignal post I just did a search to see what "AI" would turn up.

------------------

<BLOCKQUOTE>quote:</font><HR> "Remember that no dumb bastard ever won a war by dying for his country. He won it by making the other poor dumb bastard die for his country."

G. S. Patton

<HR></BLOCKQUOTE>

[This message has been edited by aka_tom_w (edited 08-24-2000).]

Link to comment
Share on other sites

Dr. Brian wrote:

> I'm actually hoping that by discussing it, the "AI" can improve, that the game actually becomes competitive if playing just the computer.

So I assume you're an expert on AI yourself? (I'm not being sarcastic.)

> We kicked those German butts in WWII!!!

Who is we? This is an international forum. Such comments are unwelcome.

David

------------------

There's a splinter in your eye, and it reads REACT

Link to comment
Share on other sites

<BLOCKQUOTE>quote:</font><HR>Originally posted by David Aitken:

Dr. Brian wrote:

> I'm actually hoping that by discussing it, the "AI" can improve, that the game actually becomes competitive if playing just the computer.

So I assume you're an expert on AI yourself? (I'm not being sarcastic.)

> We kicked those German butts in WWII!!!

Who is we? This is an international forum. Such comments are unwelcome.

David

<HR></BLOCKQUOTE>

Well Dr. Brian says this about himself in his profile:

Occupation: Engineer and University Professor

Interests:

"Teaching the world that Nazi Germany and those that fought for it were evil. "

I guess that about sums up his opinion.

Everyone is free to state their opinions here even if they offend other people are they not?

We have sort of agreed that flame wars and excessive profanity and personal attacks and insults are unwelcome here, but the stating unpopular opinions or those that may offend some here is not exactly banned by the user agreement or am I mistaken?

At least we know where he stands.

-tom w

------------------

<BLOCKQUOTE>quote:</font><HR> "Remember that no dumb bastard ever won a war by dying for his country. He won it by making the other poor dumb bastard die for his country."

G. S. Patton

<HR></BLOCKQUOTE>

Link to comment
Share on other sites

Guest Germanboy

<BLOCKQUOTE>quote:</font><HR>Originally posted by Dr.Brian:

Interests:

"Teaching the world that Nazi Germany and those that fought for it were evil. "<HR></BLOCKQUOTE>

Wow, I learn something new about my grandfather every day...

------------------

Andreas

Link to comment
Share on other sites

I'm almost inclined to take offence at that 'teaching.. ' statement. Firstly it's not beffiting a uni professor, secondly I had relatives on both sides of the war!

My Grandad was there in the occupation and uncles of my Grandmother were on the west and eastern fronts. (One was in an MG post on d-day on a beach after being invalided there, saw the horizon totally black with ships, walked down the beach and 'make give up'. Spent some nice years in Canada I believe).

They weren't evil.. ah well, we all know it so i'm not going to bother going on about it. Anyway. silly!

PeterNZ

------------------

---------------------

__COMBAT____VISION__

*Film from the CM Front*

http://combatvision.panzershark.com

Link to comment
Share on other sites

I might be studying this next year.

From Britannica.com:

>b/>fuzzy logic

<i/>in mathematics, a form of logic based on the concept of the fuzzy set. Membership in fuzzy sets is expressed in probabilities or degrees of truth, i.e., as a continuum of values ranging from 0 to 1.

Fuzzy logic and fuzzy sets were first described in the mid-1960s by Lotfi Zadeh. According to fuzzy-logic theorists, classical logic oversimplifies the concept of set membership by flatly including or excluding an individual, whereas fuzzy logic expresses the extent to which an individual pertains to a set. For example, under classical logic, theoretical tree x is a member of the set of tall trees; in contrast, under fuzzy logic, x pertains partly to the set of tall trees and can be described as fairly tall.

As a form of data processing, fuzzy logic is employed by advanced electronic computer systems. In less complex information processors, the possibility that a particular event will occur is expressed as a certainty (either false or true) represented by the binary digits 0 or 1. Fuzzy-logic systems, in contrast, break down the chance of the occurrence into varying degrees of truthfulness or falsehood (e.g., will occur, probably will occur, might occur, might not occur, etc.). This allows the outcome of an event to be expressed as a probability. Moreover, as additional data is gathered, many fuzzy-logic systems are able to adjust continually the values assigned to different probabilities. Because some fuzzy-logic systems appear able to learn from their mistakes and mimic human thought processes, they are often considered a crude form of artificial intelligence.

Fuzzy-logic systems were considered an experimental technology during the 1980s, but they achieved commercial application in the early 1990s. Advanced clothes-washing machines, for example, use fuzzy-logic systems to detect and adapt to patterns of water movement during a wash cycle. Other applications for fuzzy logic include expert systems, self-regulating industrial controls, and computerized speech and handwriting recognition programs.

Link to comment
Share on other sites

<BLOCKQUOTE>quote:</font><HR>Originally posted by aka_tom_w:

At least we know where he stands.

<HR></BLOCKQUOTE>

Yes, I proudly accept the title of Anti-Nazi, and the evil they represent. Are you implying that you are pro-Nazi?

------------------

Doc

We kicked those German butts in WWII!!!

Link to comment
Share on other sites

<BLOCKQUOTE>quote:</font><HR>Originally posted by David Aitken:

Dr. Brian wrote:

>

So I assume you're an expert on AI yourself? (I'm not being sarcastic.)

> We kicked those German butts in WWII!!!

Who is we? This is an international forum. Such comments are unwelcome.

David

<HR></BLOCKQUOTE>

David, yes, but not to the degree as a true AI professional. I did some major research in Baysian Networks and Expert Systems.

You have to understand that "AI" really doesn't exist. What does exist are expert systems, which is what the "AI" in CM really is. I'd like to find out if they truly used "fuzzy logic" which Lafti developed.

I'm not a fan of "fuzzy logic" as I found it limiting in versatility. However, after soul searching, I became a "fan" of Bayesian Networks, which is based on Bayes Theorem.

Regarding my "we defeated Germany comment," I am referring to all those people that took up arms against Germany. I do not understand why it is "unwelcome," as it is a historical fact. however, I will change it to accommodate others. I hope that helps.

------------------

Doc

We kicked those German butts in WWII!!!

Link to comment
Share on other sites

Dr. Brian wrote:

> Yes, I proudly accept the title of Anti-Nazi

I think your opinions about Nazi Germany are less bothering than your views about "those that fought for it". Nazis may be evil, but a vast proportion of German soldiers were answering a call to arms and fighting for their country. Many of them hated the Nazis as much as you do, but they were doing their duty.

About the signature... if "we" is referring to you and other people, that would infer that you fought in the war, which I doubt. If "we" refers to the people on the forum, it is inappropriate for obvious reasons. Anyway, as Tom says, you're free to your personal opinions - I just think you'll find such comments don't go down too well here.

David

------------------

There's a splinter in your eye, and it reads REACT

Link to comment
Share on other sites

Regarding the topic of the CM AI, I am interested in learning more. I am intrigued by how the game can get smarter when you give it bonuses right before setup. I seem to lose with or without the computer having a bonus. Hmmm...

------------------

-Work Hard

-Type Fast

-Save Often

Link to comment
Share on other sites

You have to understand that "AI" really doesn't exist.

no **** doctor brain. so, first you enlighten us with the breaking news that the allies (or you personally? or who is we?) won the war, now you're telling us "AI" doesn't really exist - what's next? "the sky is blue" or something?

you must be a very educated man. I envy you.

M.Hofbauer

outta here, gotta go tell my grampa that he was evil. you know, the coward, sorry not coward but evil, he is he didn't take his assault engineer flamethrower to walk up to Hitler and blast that asshole, instead he spent some vacation in Leningrad, getting his back massaged by kind russian artillery.

Link to comment
Share on other sites

Thanks, holdit, for reposting that. So in game terms (if I have understood correctly, I'm sure I will be corrected if I haven't) could this be expressed as another 'layer' of a Combat Results Table? Like you would roll to see what state the unit was in, (no enemy spotted, under heavy fire, etc.) then that result would be used to determine the roll for combat resoultion? The article says that after all 'states' have been taken into account, the outcome will still be expressed as a probability, which for me is a dice roll. Is this right?

PS : If we could get this thread back on topic it would really be great. Thanks in advance.

DeanCo--

Link to comment
Share on other sites


×
×
  • Create New...