My AI Design Q&A

[JLS]

I like the idea of copying the entire heart of the build queue multiple times, doubling production for each copy. This better allows the AI to build up even force compositions in both the zero bonus and higher bonus states. I've seen this build up limited to attack ships in most TDM type AI's. IMHO, why not extend the idea and build up the whole construction queue.

This would result in a more intimidating AI earlier, my concern would be maintenance for the None and Low Bonus Games, the AI may not have enough resources, this would need to be tested and calabrated, but the result may be even a less intimidating AI in the Mid to end game.

Cybersol, what are the thoughts of the new way the AI is hard coded to use Population Transports for the se4 minister change?

[cybersol]

Originally posted by JLS:
This would result in a more intimidating AI earlier, my concern would be maintenance for the None and Low Bonus Games, the AI may not have enough resources, this would need to be tested and calabrated, but the result may be even a less intimidating AI in the Mid to end game.

Cybersol, what are the thoughts of the new way the AI is hard coded to use Population Transports for the se4 minister change?

If you have good ratios like 1 support ship for every 4 attack ships, and maintain that through all the doublings of the queue, then I don't see how you sacrifice late game performance. In fact where most empires just stop building quickly at higher bonuses in the late game, this should keep the AI always churning out more ships. And like a human player, you then end up utilizing most or all of your shipyards right up to the point maintenance becomes a problem. As maintenance becomes a problem the Maximum Maintenance Percent of Revenue kicks in and slows the building down. And because the queues are repeated evenly, you end up with a nice consistent force composition no matter when maintenance becomes a problem.

I am not aware of minister change you are refering to. Please tell me more about it. I played before 1.84 but I never got into modding the AI or using the ministers before then. I tried to search for more information, but all I found was a vague note in history.txt for 1.82.

[JLS]

Originally posted by cybersol:

quote:

---

Originally posted by JLS:
This would result in a more intimidating AI earlier, my concern would be maintenance for the None and Low Bonus Games, the AI may not have enough resources, this would need to be tested and calabrated, but the result may be even a less intimidating AI in the Mid to end game.

Cybersol, what are the thoughts of the new way the AI is hard coded to use Population Transports for the se4 minister change?

If you have good ratios like 1 support ship for every 4 attack ships, and maintain that through all the doublings of the queue, then I don't see how you sacrifice late game performance. In fact where most empires just stop building quickly at higher bonuses in the late game, this should keep the AI always churning out more ships. And like a human player, you then end up utilizing most or all of your shipyards right up to the point maintenance becomes a problem. As maintenance becomes a problem the Maximum Maintenance Percent of Revenue kicks in and slows the building down. And because the queues are repeated evenly, you end up with a nice consistent force composition no matter when maintenance becomes a problem.

I am not aware of minister change you are refering to. Please tell me more about it. I played before 1.84 but I never got into modding the AI or using the ministers before then. I tried to search for more information, but all I found was a vague note in history.txt for 1.82.

---

Agreed, the Basic maintenance considerations, would probably not be a factor in Medium and especially High AI computer Bonus games as was stated in my Post, just its concern in None to Low bonus games.
---
In regards to the loss of AI intimidation in mid to Late AI game play , overall Ship and Base Strength as it relates to score that is directly related to AI Diplomacy and the quality of the AI ship and base in itself.

If what has been posted is true, in regards to BUG 3, 3a and 3b.
And what you may be suggesting that; If the AI designer were to issue orders for the AI to build massive amounts of Ships earlier then in past designs:

Originally posted by cybersol:
What I have not seen as much is using these same capabilities to better handle the various resource bonus games available. I like the idea of copying the entire heart of the build queue multiple times, doubling production for each copy . This better allows the AI to build up even force compositions in both the zero bonus and higher bonus states. I've seen this build up limited to attack ships in most TDM type AI's. IMHO, why not extend the idea and build up the whole construction queue .

As opposed to the gradual build up in the designers Current non-Moded and original se4 designs; then wouldn’t this exasperate the 3rd Bug you mentioned?

Originally posted by cybersol: Problem/Bug 3b)
A similar bug where ships that hit the 50% retrofit limit just sit there useless trying to retrofit but failing over and over again. This happens for a long time, sometimes they switch and try to retrofit at a different ship yard with the same result.

=================================================

Follow up reference:

3) Is there anyway to stop the AI from trying to retrofit to a “newer” Version of a ship that is in fact the same as the old Version? This results in ships being wasted just sitting there and trying to be retrofitted but failing because the ships are identical.

3)As I recall, the AI ship hangs around for a short time if cannot Retrofit, plus the AI gets a neat little message it can't read

3) I have seen them try for a while, over and over.

3) Is there anyway to stop the AI from trying to retrofit to a “newer” Version of a ship that is in fact the same as the old Version? This results in ships being wasted just sitting there and trying to be retrofitted but failing because the ships are identical.

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

Nope, I don't know how to do it... Think Aaron "fixed" it in several patches, but the behavior is still there.

Well, for me it tries to retrofit for 10 or more turns until finally the tech changes so the new design is different and then the ships can retrofit properly. Since posting this problem, I now run into a similar one where ships that hit the 50% retrofit limit just sit there useless trying to retrofit but failing over and over again for a long while.

In my case 10 to 50 turns depending. Longer for the 50% limit problem than for the same design problem.

Problem/Bug 3a)
Is there anyway to stop the AI from trying to retrofit to a “newer” Version of a ship that is in fact the same as the old Version? This results in ships being wasted just sitting there and trying to be retrofitted but failing because the ships are identical.

Problem/Bug 3b)
A similar bug where ships that hit the 50% retrofit limit just sit there useless trying to retrofit but failing over and over again. This happens for a long time, sometimes they switch and try to retrofit at a different ship yard with the same result.

Bug/Problem 3).
Think doesn't exist a way to avoid this problem.

[ June 18, 2003, 18:40: Message edited by: JLS ]

[cybersol]

Originally posted by cybersol:
What I have not seen as much is using these same capabilities to better handle the various resource bonus games available. I like the idea of copying the entire heart of the build queue multiple times, doubling production for each copy. This better allows the AI to build up even force compositions in both the zero bonus and higher bonus states. I've seen this build up limited to attack ships in most TDM type AI's. IMHO, why not extend the idea and build up the whole construction queue.

Originally posted by JLS:
Agreed, the Basic maintenance considerations, would probably not be a factor in Medium and especially High AI computer Bonus games as was stated in my Post, just its concern in None to Low bonus games.

In regards to the loss of AI intimidation in mid to Late AI game play, overall Ship and Base Strength as it relates to score that is directly related to AI Diplomacy and the quality of the AI ship and base in itself.

And what you may be suggesting that; If the AI designer were to issue orders for the AI to build massive amounts of Ships earlier then in past designs: As opposed to the gradual build up in the designers Current non-Moded and original se4 designs; then wouldn’t this exasperate the 3rd Bug you mentioned?

Ah, I think I see why we seem to be miscommunicating now. Obviously I was not clear. First of all, I am mainly advocating an EVEN build up of forces. Consider the AI in the midgame for now. Assume its in a state like defend (short term). Assume there are just three ship designs for the whole empire: colonizer, attack, and layer for example. I am saying the defend (short term) queue should look something like this:

name ppi mhal
attack 120 2
colonizer 360 1
layer 720 1
attack 60 4
colonizer 180 2
layer 360 1
attack 30 8
colonizer 90 4
layer 180 2
attack 15 16
colonizer 45 8
layer 90 4

Thus it starts out building small numbers of everything in the "right" relative ratios. Then it goes back and build more of everything in the same ratios. The top repeat is designed to handle the no bonus early game case while the bottom repeat is designed to handle the high bonus late game case.

Most people do something like this for their most important ships, say attack ships, but not for the entire line of attack, support, infrastructure, and colonizer ships.

Now in the exploration and not connected queues at the begging you have the normal build one attack ship, build one colonizer, etc. But after the initial preparations, you again have escalating copies of the heart of the build queue.

Does that explain it better? It thus builds very deliberately at first, then an even small number of all your ships, then an even medium number of all your ships, etc. Thus as long as you have resources all the shipyards will be producing full bLast. Only when the maintenance limit approaches will the AI slow down its building process. In a no bonus game it will produce a smaller number of ships and stop in the second doubled copy, but in a high bonus game it might get to the fourth or fifth doubled copy.

[JLS]

[quote]Originally posted by cybersol:

Originally posted by cybersol:
[qb]Ah, I think I see why we seem to be miscommunicating now. Obviously I was not clear. First of all, I am mainly advocating an EVEN build up of forces. Consider the AI in the midgame for now. Assume its in a state like defend (short term). Assume there are just three ship designs for the whole empire: colonizer, attack, and layer for example. I am saying the defend (short term) queue should look something like this:

name ppi mhal
attack 120 2
colonizer 360 1
layer 720 1
attack 60 4
colonizer 180 2
layer 360 1
attack 30 8
colonizer 90 4
layer 180 2
attack 15 16
colonizer 45 8
layer 90 4

Thus it starts out building small numbers of everything in the "right" relative ratios. Then it goes back and build more of everything in the same ratios. The top repeat is designed to handle the no bonus early game case while the bottom repeat is designed to handle the high bonus late game case.

Most people do something like this for their most important ships, say attack ships, but not for the entire line of attack, support, infrastructure, and colonizer ships.

Now in the exploration and not connected queues at the begging you have the normal build one attack ship, build one colonizer, etc. But after the initial preparations, you again have escalating copies of the heart of the build queue.

Does that explain it better? It thus builds very deliberately at first, then an even small number of all your ships, then an even medium number of all your ships, etc. Thus as long as you have resources all the shipyards will be producing full bLast. Only when the maintenance limit approaches will the AI slow down its building process. In a no bonus game it will produce a smaller number of ships and stop in the second doubled copy, but in a high bonus game it might get to the fourth or fifth doubled copy.

The results are the same in respect to a quicker build up of the Ships and Base amounts, from previous se4 Designs, in the opening to mid stages of the game. Possibly, the maximum depletion of the AIs Resource Reserves, well before the latest and strongest ship designs can be built in the Mid to Late game and leading to probable exasperated retrofit problems (“50% bug3, 3a,3b”) in the Mid to late game for the old existing designs the AI built and NOT lost thru combat from the early to mid game builds.

Possibly resulting in a loss of AI intimidation in mid to Late AI game play, and overall Ship and Base Strength as it relates to score that is directly related to AI Diplomacy and the quality of the AI Ship and Base designs built, in that of itself.
=
===
=
With this aside, you are not suggesting that a Designer would start his Short Term Defense State in the manner as you described.

The objective of the AI Player, in the Short term defensive State; will be to expel the other Player out of his territory as soon as possible and for that AI Player to consider Defensive and Counter Offensive, options; thru a Strategic State Change and/or Diplomacy measures.

Colony ships would not be a good early placement in this AI Strategic State.

Even, Mine and Sat Layers would not be a good early choice unless they followed Attack Ships with a few STALL entries to increase their survivability.
=================================================
Possible early entries for the AI, Vehicle Construction File Layout for the AI Players Short Term Defense:

A much higher Attack Ship MHAL should be suggested not to mention *Displacement Attack Ship Grouping to expedite the need for a quick preemptive measure. A few Mine Sweepers /(types) also must be introduced very early if your (minefield value is :=False) , as not to have your AI, suckered into uneeded and continued losses; thru other players Minefields

Another possible, less likely to be successful scenario, in the short scope, but safer for your AI. Would be, If your Minefield Value was :=True , then you could build ALL attack ships and CV's first, and ending a preconceived fleet with many Standard Mine Sweepers, at that point your AI Players Offensive Attack Ships will be in near Position to Counter attack or in defensive actions with in, and the AI Players, latest built; Mine Sweepers will attempt to clear the way for the INITIAL Counter Offensive in the enemies terrotory.

===
=
===
*Reference:
(EE, FF, DD etc..Displacement Attack Ship Grouping. Please refer to PvK’s Proportions, and PvK’s XiChung race Design File of an Excellent if not the best; Displacement Ship Design Group, example.)

Link
> Proportions for se4 <

[ June 18, 2003, 23:24: Message edited by: JLS ]

[cybersol]

Ah, I see more what you are saying now as well. I guess I'm not as worried about obsolete (old tech) ship designs on the battlefield. If you are at war, then there are two possibility for these obsolete ships. If they survive then your fleets must be strong enough to conquer the opposition without having the latest tech (so it doesn't matter). And if they die, then the resources are free for modern ship designs to start rolling off the line.

As for the example I gave, I was assuming for demostration purposes that only three type of ships existed (which would be a dumb thing to do, but makes the example easier to show).

[JLS]

Originally posted by cybersol:
Ah, I see more what you are saying now as well. I guess I'm not as worried about obsolete (old tech) ship designs on the battlefield. If you are at war, then there are two possibility for these obsolete ships. If they survive then your fleets must be strong enough to conquer the opposition without having the latest tech (so it doesn't matter). And if they die, then the resources are free for modern ship designs to start rolling off the line.

As for the example I gave, I was assuming for demostration purposes that only three type of ships existed (which would be a dumb thing to do, but makes the example easier to show).

I am not confirming what you say is true, in regards to Bug 3, 3a, and 3b.

From “your play testing trials”, and in all aspects of your Posts and the follow-up post of others, it gives the impression that your “ obsolete (old tech) ship designs ” may never make it to a " forward battlefield ". For this reason, that you have declared, only the Newest and Strongest designs in a fragmented fleet will do the forward fighting, as the “ obsolete (old tech) ship designs ” will remain in a Ship Yards "orbit for a very long while"?

Please Refer to the Follow up reference:
quotes 5 Posts down.
=
==
=

The AI Strategic State Planning and Layout is important, I posted the rebuttal only so others may understand that; your choices here, may decide the out come of that AI Players situation. A bad layout in the Short Term State (or many other AI States) may result in a prolonged and time-consuming Solid AI LT-Defense State; spelling the doom of this AI, with little hope for that AI Player changing the fortunes; of the chance for a Counter Offensive.

Also to point out the AI State Vehicle Build layout must have strong consideration as to the AI settings; you made for (Ships don't move through minefields := True or False) as well as all the other changes and/or additions you made thru the entire AI Race’s Folder
==============
Opinion only, for the need of a structured AI State plan, is outlined below:

The most forgiven AI State may be Not Connected.
The most flexible AI State may be Secure Holdings and/or Incursion.
The most, Strait forward AI State is Prepare to Attack and Attack
The most need for a meticulous planed AI State, may be explore and Infrastructure.
The most unforgiving AI State is Defend Short Term, Prepare for Defense and Defend Long Term.

[ June 19, 2003, 14:55: Message edited by: JLS ]

[cybersol]

Originally posted by JLS:
I am not confirming what you say is true, in regards to Bug 3, 3a, and 3b.

From “your play testing trials”, and in all aspects of your Posts and the follow-up post of others, it gives the impression that your “ obsolete (old tech) ship designs ” may never make it to a " forward battlefield ". For this reason, that you have declared, only the Newest and Strongest designs in a fragmented fleet will do the forward fighting, as the “ obsolete (old tech) ship designs ” will remain in a Ship Yards "orbit for a very long while"?

Are you saying I'm crazy and I like to make AI bugs up?
Luckily AI bugs 3, 3a), and 3b) do not occur all that frequently. I've only seen up to about 5 ships affected simultaneously so far.

Originally posted by JLS:
The AI Strategic State Planning and Layout is important, I posted the rebuttal only so others may understand that; your choices here, may decide the out come of that AI Players situation. A bad layout in the Short Term State (or many other AI States) may result in a prolonged and time-consuming Solid AI LT-Defense State; spelling the doom of this AI, with little hope for that AI Player changing the fortunes; of the chance for a Counter Offensive.

Also to point out the AI State Vehicle Build layout must have strong consideration as to the AI settings; you made for (Ships don't move through minefields := True or False) as well as all the other changes and/or additions you made thru the entire AI Race’s Folder
==============
Opinion only, for the need of a structured AI State plan, is outlined below:

The most forgiven AI State may be Not Connected.
The most flexible AI State may be Secure Holdings and/or Incursion.
The most, Strait forward AI State is Prepare to Attack and Attack
The most need for a meticulous planed AI State, may be explore and Infrastructure.
The most unforgiving AI State is Defend Short Term, Prepare for Defense and Defend Long Term.

Thanks for this great insight on AI State and its implications. I am still learning a lot here. For now, I am still using basically three build queues. One for exploration, one for not connected, and one for everything else (similar to Mephisto's AIs). I may expand on that later, but right now there are more critial areas under development.

In the meantime, why don't you explain a little more about AI state. You start out in exploration (assuming you are connected). Then encounter another empire. If they are far away, not a threat then you most likely go into infrastucture? If the are close by and maybe a threat then defend (short term)? What other state transitions are possible starting in the exploration state? The reason I ask is because it is important that my AI manage this transition from being alone in the universe to encountering another race, because right now it is very aggressive with colonization in its expansion phase (exploration state).

[JLS]

I also have used a few of your Posts to redo a few changes in AIC...

EDIT:
If I came off a little scruffy, I had the flue most of Last week

Thanks for this thread, Cybersol

[ June 21, 2003, 23:36: Message edited by: JLS ]

[JLS]

ALL quotes: Originally posted by cybersol:

You start out in exploration (assuming you are connected).

(Yes) In the very first turn of the game. If the AI Starting HS is void of a warp point and unable to Explore and not a Ancient Race, the AI will Change to the Not Connected State.

~

Then encounter another empire. If they are far away, not a threat then you most likely go into infrastucture?

(No) The AI will remain in the Exploration State, if there are more Systems nearby still unexplored

~

If the are close by and maybe a threat then defend (short term)?

There is not a (maybe a threat), for the AI.
Threat would be constituted, if the other Player was not in an (Right Of Passage) Agreement, that Player would be considered the (Enemy) for the sake of this discussion.

(Yes) There is an Enemy ship in your AIs Claimed territory… Your AI may Change to Defend Short Term State.

(No) If there is an Enemy Near but not in the Claimed Territory, then your AI may go into Infrastructure State
~

What other state transitions are possible starting in the exploration state?

Again, there are many possible Scenarios that can occur.

However, along with the AI Exploration State change in the Scenarios posted above. I would say the most Common would be if:

Your AI had an agreement with another Player and Your AI no longer had Systems, reasonably close to Explore, your AI would go into the Infrastructure State.

Definitions.

AI Exploration Strategic State:
This is where the AI; will explore new worlds and seek out new life, and boldly go where no other, AI has gone before.

On the serious side, your AI is expected to colonize more in this AI Exploration State and the Secure Holdings State then any other State.
However, it is recommended the Designers; AI Colonizer Production, stay within the Parameters of the MOD he, or she is designing for.
The finished product, in my opinion. Would be measured not how your AI competes against another AI, but how that AI deals and interacts with a Human Player
-
AI Infrastructure Strategic State:
This is where the AI, will build up its Infrastructure, to setup Passive defenses and to Start long term Projects. Example: Non-Offensive Stellar Manipulation devises and other Vehicles that would require a long and extensive build process.
-
AI Defend Short Term Strategic State:
This would be the first stage for the AI Player, to boot the other Enemy player out of his territory and for that AI to analyze its current Status for that current Situation.
-
AI Not Connected Strategic State:
This AI Player has colonized all of its available systems, and this AI player is currently disconnected from other systems in the Quadrant.
The basic goal for the AI here is to Research Techs that will Colonize new Planet types, open news Warps, Create new Planets and to stay competitive with its own Ship Designs, so when the Warp is opened; this AI has not opened Pandora’s Box

[ June 24, 2003, 04:52: Message edited by: JLS ]