Admiral's Edition General Thread

[Erik Rutins]

ORIGINAL: pad152
No one is assigned to the AI?

The AI touches everything, so basically everyone has some role in how it pans out - you don't see anyone assigned specifically to PBEM either... [;)]

[witpqs]

ORIGINAL: witpqs

ORIGINAL: Erik Rutins

ORIGINAL: thinz2
My machine has an Intel Core 2 CPU 6400 @ 2.13 Ghz. I noticed in MS Flight Simulator X that the software in configuring itself tended to underrate the processing capacity of dual CPUs because of their lower clock speeds.
So would my system be considered optimal for AE?

Absolutely - the "GHz" rating is really somewhat misleading, but there are few better methods out there. I think it's probably fair to say that when we say 2GHz, we mean "2GHz back from when the first processors hit 2GHz". That means a new dual core like the 6400 is probably more equivalent to 4-5GHz in those terms.

Regards,

- Erik

Does this mean that AE will be compiled to take advantage of multiple cores?

In case you guys missed this query...

[Brady]

CC, but wouldent Just Climb and spead be suficient to express the diferance in manuaverabality? to say compare a Oscar and a P-38, Clearly the Oscare is a far more nimble plane, and the P-38 much faster and depending on the alt could out climb it, it would seam at a glance that to state that the P-38 was more manuaverable as well would be over statating the isue of it's comparative uberness, but I never fully understood how this all worked out in WiTP, I know when Nickademus worked his magic with the AtoA model he seamed to come a long way given the limited variables he could tweak, considering the team at hand I suspect good things will come from this in the end.

[Erik Rutins]

ORIGINAL: witpqs
Does this mean that AE will be compiled to take advantage of multiple cores?

In case you guys missed this query...

Not that I'm aware of, but the fact remains that a 6400 at 2GHz is quite a bit faster than an old Athlon XP at 2GHz.

Regards,

- Erik

[el cid again]

ORIGINAL: Brady

CC, but wouldent Just Climb and spead be suficient to express the diferance in manuaverabality? to say compare a Oscar and a P-38, Clearly the Oscare is a far more nimble plane, and the P-38 much faster and depending on the alt could out climb it, it would seam at a glance that to state that the P-38 was more manuaverable as well would be over statating the isue of it's comparative uberness, but I never fully understood how this all worked out in WiTP, I know when Nickademus worked his magic with the AtoA model he seamed to come a long way given the limited variables he could tweak, considering the team at hand I suspect good things will come from this in the end.

Well - the RHS team managed to achieve some success in the air model. A lot of effort went into figuring out how to have relative performance between different plane types properly rated. One problem is that you need to consider the combined set of fields defining an aircraft to assess its quality: Oscar is weak in protection (there isn't any for the I) and firepower (the Ia had two .30s, the Ib had 1 .30 and 1 .50, and the Ic had two .50s - not a lot of firepower in an age when standard fighters had four .50s or eight .30s). Yet "the Oscar was almost as great a technical suprise as the Zero" - and the surprise was mostly maneuverability - also tactics (the IJN turning in maneuver was adoped by JAAF).

It would be far better if horizontal maneuverability was separated from vertical. And altitude matters a big deal. So the ability to rate at 5000 feet, 15000 feet and 25000 feet (say) might be really useful.

Most planes are awful above 30,000 feet, many awful much lower. Just because the plane can get there does not mean it can maneuver, that its guns won't freeze up, etc. There is room for considerable more sophistication in the system addressing these matters. But ANY improvements are welcome - and we should look at what we get in that light. WITP II will no doubt do even better than AE.

[jwilkerson]

ORIGINAL: el cid again

ORIGINAL: Brady

CC, but wouldent Just Climb and spead be suficient to express the diferance in manuaverabality? to say compare a Oscar and a P-38, Clearly the Oscare is a far more nimble plane, and the P-38 much faster and depending on the alt could out climb it, it would seam at a glance that to state that the P-38 was more manuaverable as well would be over statating the isue of it's comparative uberness, but I never fully understood how this all worked out in WiTP, I know when Nickademus worked his magic with the AtoA model he seamed to come a long way given the limited variables he could tweak, considering the team at hand I suspect good things will come from this in the end.

Well - the RHS team managed to achieve some success in the air model. A lot of effort went into figuring out how to have relative performance between different plane types properly rated. One problem is that you need to consider the combined set of fields defining an aircraft to assess its quality: Oscar is weak in protection (there isn't any for the I) and firepower (the Ia had two .30s, the Ib had 1 .30 and 1 .50, and the Ic had two .50s - not a lot of firepower in an age when standard fighters had four .50s or eight .30s). Yet "the Oscar was almost as great a technical suprise as the Zero" - and the surprise was mostly maneuverability - also tactics (the IJN turning in maneuver was adoped by JAAF).

It would be far better if horizontal maneuverability was separated from vertical. And altitude matters a big deal. So the ability to rate at 5000 feet, 15000 feet and 25000 feet (say) might be really useful.

Most planes are awful above 30,000 feet, many awful much lower. Just because the plane can get there does not mean it can maneuver, that its guns won't freeze up, etc. There is room for considerable more sophistication in the system addressing these matters. But ANY improvements are welcome - and we should look at what we get in that light. WITP II will no doubt do even better than AE.

Manueverability has been substantially reworked - and altitude bands added - check over on the air thread for more details!

[donkey_roxor]

This is not really a question, but more of a comment -

Sardaukar asked a question on page 3 of this thread asking if the AI will be user-modifiable in any way. I realize that the answer to that is "work on AI will happen after most of the other stuff is done, and we're trying our best," but I'd like to cast my vote in favor of user-moddable AI scripting. Look at all the neat things modders have done with the original WiTP - imagine if the same modders could also script AI behavior. How cool would it be to have the Japanese AI launch an attack on Pearl Harbor or northern Australia? Now think about how cool it would be if AI scripts could be transferred between players, maybe with some type of encryption - like PBEM, except not quite...

Anyway, just some thoughts.

[jwilkerson]

ORIGINAL: donkey_roxor

This is not really a question, but more of a comment -

Sardaukar asked a question on page 3 of this thread asking if the AI will be user-modifiable in any way. I realize that the answer to that is "work on AI will happen after most of the other stuff is done, and we're trying our best," but I'd like to cast my vote in favor of user-moddable AI scripting. Look at all the neat things modders have done with the original WiTP - imagine if the same modders could also script AI behavior. How cool would it be to have the Japanese AI launch an attack on Pearl Harbor or northern Australia? Now think about how cool it would be if AI scripts could be transferred between players, maybe with some type of encryption - like PBEM, except not quite...

Anyway, just some thoughts.

Adding a true scripting engine - remotely like TAOW - is definitely a WITP_II list item and not an "AE" item. But agree it would certainly be cool! [:)]

[Shark7]

Didn't see this addressed anywhere else so I'll ask here.

The stock game left out Thailand when historically they did fight on the side of Japan early in the war. Will AE add the Thai OOB as a Japanese ally?

[jwilkerson]

ORIGINAL: Shark7

Didn't see this addressed anywhere else so I'll ask here.

The stock game left out Thailand when historically they did fight on the side of Japan early in the war. Will AE add the Thai OOB as a Japanese ally?

Kereguelen can give more details over on the land thread, but yes I think there will be a Thai OOB, it has certainly been discussed a number of times.

[Kereguelen]

ORIGINAL: jwilkerson

ORIGINAL: Shark7

Didn't see this addressed anywhere else so I'll ask here.

The stock game left out Thailand when historically they did fight on the side of Japan early in the war. Will AE add the Thai OOB as a Japanese ally?

Kereguelen can give more details over on the land thread, but yes I think there will be a Thai OOB, it has certainly been discussed a number of times.

Yup, Royal Thai Army is in.

[Nikademus]

will we be able to bring onto thy opponents.....a shrubbery?

[witpqs]

ORIGINAL: Nikademus

will we be able to bring onto thy opponents.....a shrubbery?

There is no cannibalism in the Royal Navy! Maybe in WITP-2?

[Buck Beach]

I'm not sure I truly understand the full concept of AI scripting? What does it include? I am only looking for a very brief explanation.

[DuckofTindalos]

ORIGINAL: Buck Beach

I'm not sure I truly understand the full concept of AI scripting? What does it include? I am only looking for a very brief explanation.

Basically, it's about letting the scenario creator define in great detail what the AI must do, i.e. the script it has to follow. TOAW is famous for having a very fine scripting engine; WitP not so much...

[jwilkerson]

ORIGINAL: Buck Beach

I'm not sure I truly understand the full concept of AI scripting? What does it include? I am only looking for a very brief explanation.

First a couple of examples.

(1) TAOW - in this game (and I'll admit I haven't coded up a full blown "event script" for TOAW in some years .. but way back when I spent far more hours "building" TAOW games than I did playing them!!!) the scenario builder has a reasonably useful "event scripting" language with which to build in some AI. Triggers can be set and event chains launched from such triggers. Supply levels can be changed, reinforcements can be triggered, weather can be changed, intel levels can be changed. And then from a campaign perspective, multiple objective paths can be created and units/organizations can be assigned to these objective paths. If you build the scenario, you will be able to beat it (if it can be beaten) because you know the options, but it can provide an interesting a variable puzzle for others.

(2) EU2/HOI2/Victoria - I've only played around extensively with the HOI(1) capabilities. Here you can define whole nations and all the related attributes. In these "paradox" games, the game creators themselves use the scripting engine to define the games. The scripting engine is not a "bolt on". So you have the same powers that the game designers have. The "scale" of HOI is a bit too macro for my tastes, hence I don't have nearly as many hours investeed there as I do with the TAOW engine, but both serve as examples of "AI Scripting" at least at some level.

What are some of the attributes of AI scripting?
(actually I will answer more closely to what would I like to see in an AI Scripting engine)

(a) Well an AI scripting engine may or may not be "exposed" i.e. available to modders. This is more of a business decision than a technical decision.

(b) One successful type of architecture would have the scripting engine be event based. In this case a full DES (Discrete Event Simulation) model would be preferred. Events in a DES system can create any number and type of other events when they fire. And they can destroy any number and type of other events. DES can be time based, so an event can create a number of other events and add them to the event list in the future. When the time reaches that time scheduled for an event to fire, then event fires. It make take actions, or it may CRUD other events.

(c) Another useful component would be a Petri Net based relationship. Most DES systems I've built in the past were "linear". A PN based system would be "graphical" (in the sense of graph theory not in the sense of pictures). A PN system allows multiple simultaneous asynchronus processes to be acted upon. That is we are now non-linear. Stocastic Petri Nets can add random variables to Petri Net elements to randomize event firings even further. I have used Petri Nets very successfully to simulate operations of semi-conductor manufacturing equipment and thus became aware of the power of this (mathematical) tool for simulating almost any phenomena.

So these would be among the basic elements I would like to see in a "go forward" AI scripting engine. And a further topic would be "extensibility". That being, what does it take to add a new event type? Can this be done without changing the even processing code? And there would certainly have to be some limits here, but at the same time event type extensibility would be another "nice to have" attribute.

So is this in scope for "AE"? Definitely not. But if Joe ever gets to work on "WITP_II" or equivalent, it would be something I would be advocating for.

--- BTW there are some decent books on AI for games --- I haven't yet seen one that mentions Petri Nets though!

[Buck Beach]

Actually more detail than I needed, but, I somewhat understand whats not going to be done. Now a follow-up question and then I'll shut up. Will the AI scripting in AE have a scripting that triggers basic events like refueling and replenishment when endurance/fuel & supplies are exhausted. If WITP has such scripting, it does not work as you undoubtedly are painfully aware. Sure was hoping AE or the Patch to WITP we were awaiting would have it.

[DuckofTindalos]

That's all part of the stuff we're working on; making sure that the AI doesn't sail around until its tanks are empty and it's a thousand miles away from base...

[Brady]

An Oldie but a goodie....
 
 Early War Allied Suply: Early on in the pacific war for the Allies suply was a bit of a problem, all maner of item, from fighter planes to profolactics were not in great suply, the tean pregancy rate in New Zeland Skyrocketd in early 42 puting a huge burdon on Allied Transports in their atempt to move desperatly neaded maternatly materials to their from the West coast. All punning aside the Allies cup runith over a tad in this area of the game, their is NO suply problem for the Alies early on, their is enough local suply at all the many allied bases to run things just fine and after the first ships arive even as far out as New Zeland and Nouma their set for the rest of the war, with a ton of materail. This large right from the start Allied suply bonious gives the allies to much of an edge, it alows them to be far to agreasive early on. IMO suply for the allies should be scaled up, the west coast and Indian ports should start off low and ramp up to present levals by the end of 42.

[Woos]

ORIGINAL: Buck Beach

I'm not sure I truly understand the full concept of AI scripting? What does it include? I am only looking for a very brief explanation.

To give a more simple explanation than Joe: In general AI scripting is if the designers of a game (or of the AI scripts) fix certain behaviour of the AI. The alternative is a free-planning AI which only decides based on its general algorithms and on what it sees the opponent do. A well-known example for AI-scripting (and "human-scripting" BTW) are the opening libraries of chess. For several moves there is a huge 'script' telling the AI how to react on each move of the opponent. Once out of the book, the AI has to switch to free planning.

Basic problem of AI scripting is that if you know what the script prescribes (either because you wrote it or due to several games against it) you can normally easily beat the AI since you can either make moves throwing it out of script or use the script against the AI (just expand quickly as IJ in WitP and watch the Allies AI try to attack base it is not scripted to give up already with insufficient forces; I think in TOAW it was landing some unit in the back of the enemy and see all units assigned to that city suddenly leave the front leaving gaps).


BTW I'm not an expert in AIs at all but is event scripting still state of the art for AIs? It seems to lack quite some flexibility. E.g. a script for the invasion of the Philipines would have to recognize and act upon several possible moves of the opponent (defend the beaches, fall back immediatly, sent reinforcements by ship (OK not that probable for the philipines) or plane, ....). Doing that all in an event driven script seems to be quite complicated to me.
Maybe for WitP-II one should look at the Robotics people (which I also am not an expert in) which seem to use task- and goal-oriented approaches to solve AI issues in 'games' like Robocup and Robocup@home.