Game AI thoughts
Moderator: Oberlus
Game AI thoughts
I have been thinking of what game engine should be able to do. I think that most players are interested in having game AI that would offer them a tough challenge. Thus I think it would be good that most if not all aspects of AI should be modifiable.
At the same time I am quite interested in pondering what AI should be able to simulate. I think that the game would be best served by having AI function as a hierarchical entity. The idea is that AI selects strategies, each that will then affect the strategy chosen below this level.
A] Overall Strategy AI
Selects one victory condition to be followed: for example it could be "find Orion and defeat its guardian", "exterminate all other races", "win crown of entire space in council".
B] Strategy AI
This selects the state of the imperium as one of (to name few): "strengthen reseach", "strengthen economy", "expand boldly". In essence this would be the overall selection between guns (military spending) and butter (research, economy) in budget.
These two AIs would allow state to select between military, commercial and research oriented approach to interstellar politics.
At the same time I am quite interested in pondering what AI should be able to simulate. I think that the game would be best served by having AI function as a hierarchical entity. The idea is that AI selects strategies, each that will then affect the strategy chosen below this level.
A] Overall Strategy AI
Selects one victory condition to be followed: for example it could be "find Orion and defeat its guardian", "exterminate all other races", "win crown of entire space in council".
B] Strategy AI
This selects the state of the imperium as one of (to name few): "strengthen reseach", "strengthen economy", "expand boldly". In essence this would be the overall selection between guns (military spending) and butter (research, economy) in budget.
These two AIs would allow state to select between military, commercial and research oriented approach to interstellar politics.
- eleazar
- Design & Graphics Lead Emeritus
- Posts: 3858
- Joined: Sat Sep 23, 2006 7:09 pm
- Location: USA — midwest
Re: Game AI thoughts
I don't have the AI expertise to comment on your ideas for an AI structure... but here are some of the things about the AI that have been discussed, and IIRC generally agreed upon.
1) There should be no special abilities that effect only the AIs and not human players equally. I.E. MoO had "diplomacy" which made AI empires less likely to go to war with you, but obviously didn't effect the attitude of human players.
2) The part of the "AI" that contains the species personality is not the overall strategic empire AI -- what you're talking about -- but in the citizenry which react to both human and AI imperial actions according to the nature of the species.
The general idea is to make a fair game where the AI and human players are on an equal footing, and thus to hopefully make it easier to build a powerful AI that doesn't need to cheat to be competitive.
EDIT: here's a short thread you might find interesting.
1) There should be no special abilities that effect only the AIs and not human players equally. I.E. MoO had "diplomacy" which made AI empires less likely to go to war with you, but obviously didn't effect the attitude of human players.
2) The part of the "AI" that contains the species personality is not the overall strategic empire AI -- what you're talking about -- but in the citizenry which react to both human and AI imperial actions according to the nature of the species.
The general idea is to make a fair game where the AI and human players are on an equal footing, and thus to hopefully make it easier to build a powerful AI that doesn't need to cheat to be competitive.
EDIT: here's a short thread you might find interesting.
Re: Game AI thoughts
I agree that game rules should be same to vis a vis humans and AI competitors because it makes the coding (and rules creation) a lot simpler.
I have pondered the AI issue a bit in my free time. Most of this is based on two books about geopolitics and how great power could come to be. Triggers are my ideas rather than professionals.
My idea is that each state has a unique set of victory conditions based on their "nature". Basicly a very aggressive species would want to annihilate others, while more peaceful species would want to become crowned by galactic council. This is ofcourse gross simplification of species "culture" but would serve as basis to game in first phase at least.
Basicly the idea is that empire is run as a state machine where it selects one of four modes as its grand strategy and if it fails to fullfil it it would return back to A (back to basics) and then after some soul searching gets back to business. The species "culture" would describe priorities of methods used to achieve these goals.
The four grand strategies are:
A] Formulation of Ideological Basis
-development of new ideology
-collecting elites behind the ideology
**Essentially at this phase the state would strive to make its leaders unified behind the new concept. The idea here is that "nature" could change from aggressive empire to peaceful based on their leaders, population and ideas.
**Trigger for success would be referendum within elite (over half of leaders behind new ideology) showing new ideology is accepted as basis of future actions.
B] Internal Development
-development of science and technology
-economic expansion
-infrastructure
**Essentially this is peaceful race with other nations to create economical foundation to successful empire.
**Trigger for success would be when elites state publicly to other states they demand their place in a universe too (some meters of success like technological, military power or economic power within top 25% of all states)
C] Cultural Unification
-Assimilation of different cultures
-Creation of nationally important cultural achievements
**In this pahse the culture of elites is moved to citizens at increasing pace. The state would work in cultural aspects.
**Trigger for success would be over 50% of all planets having similar culture and creation of at least one cultural wonder.
D] External Expansion
-Assimilation of other states and systems
**Peacefully if possible, by war if necessary.
**Trigger for failure would be a lost major war (losses 10-20% of military power).
Does any of this make any sense?
I have pondered the AI issue a bit in my free time. Most of this is based on two books about geopolitics and how great power could come to be. Triggers are my ideas rather than professionals.
My idea is that each state has a unique set of victory conditions based on their "nature". Basicly a very aggressive species would want to annihilate others, while more peaceful species would want to become crowned by galactic council. This is ofcourse gross simplification of species "culture" but would serve as basis to game in first phase at least.
Basicly the idea is that empire is run as a state machine where it selects one of four modes as its grand strategy and if it fails to fullfil it it would return back to A (back to basics) and then after some soul searching gets back to business. The species "culture" would describe priorities of methods used to achieve these goals.
The four grand strategies are:
A] Formulation of Ideological Basis
-development of new ideology
-collecting elites behind the ideology
**Essentially at this phase the state would strive to make its leaders unified behind the new concept. The idea here is that "nature" could change from aggressive empire to peaceful based on their leaders, population and ideas.
**Trigger for success would be referendum within elite (over half of leaders behind new ideology) showing new ideology is accepted as basis of future actions.
B] Internal Development
-development of science and technology
-economic expansion
-infrastructure
**Essentially this is peaceful race with other nations to create economical foundation to successful empire.
**Trigger for success would be when elites state publicly to other states they demand their place in a universe too (some meters of success like technological, military power or economic power within top 25% of all states)
C] Cultural Unification
-Assimilation of different cultures
-Creation of nationally important cultural achievements
**In this pahse the culture of elites is moved to citizens at increasing pace. The state would work in cultural aspects.
**Trigger for success would be over 50% of all planets having similar culture and creation of at least one cultural wonder.
D] External Expansion
-Assimilation of other states and systems
**Peacefully if possible, by war if necessary.
**Trigger for failure would be a lost major war (losses 10-20% of military power).
Does any of this make any sense?
- Geoff the Medio
- Programming, Design, Admin
- Posts: 13603
- Joined: Wed Oct 08, 2003 1:33 am
- Location: Munich
Re: Game AI thoughts
You should reread eleazar's post... Point point 2 in particular.
- Krikkitone
- Creative Contributor
- Posts: 1559
- Joined: Sat Sep 13, 2003 6:52 pm
Re: Game AI thoughts
Those sound like good ideas for things that Players should be doing in the game rather than how they decide what to do in the game.
ie those should be hurdles the AI+human player have to overcome/tools they can use, rather than the decision process.
The main way 'culture' would shape an AIs overall game goal is what game goal it can best use the 'culture' for.
IE I plan to whipe everyone out not because I "unified my elites" for that goal... but because I think the race I'm playing with will be most efficient/easiest/most certain to win that way.
Now an AI Should probably be able to 'soul search' ... but that should be a consideration of "from where I am now, winning with the goal I have now seems unlikely.. check and see if another goal looks easier/more certain to achieve than the on I am currently going for"
It should Not be a 'what are my people behind' but 'what can I make my people most useful at'.
ie those should be hurdles the AI+human player have to overcome/tools they can use, rather than the decision process.
The main way 'culture' would shape an AIs overall game goal is what game goal it can best use the 'culture' for.
IE I plan to whipe everyone out not because I "unified my elites" for that goal... but because I think the race I'm playing with will be most efficient/easiest/most certain to win that way.
Now an AI Should probably be able to 'soul search' ... but that should be a consideration of "from where I am now, winning with the goal I have now seems unlikely.. check and see if another goal looks easier/more certain to achieve than the on I am currently going for"
It should Not be a 'what are my people behind' but 'what can I make my people most useful at'.
- eleazar
- Design & Graphics Lead Emeritus
- Posts: 3858
- Joined: Sat Sep 23, 2006 7:09 pm
- Location: USA — midwest
Re: Game AI thoughts
There are different opinions about weather FO should have malleable cultures roughly as you are describing. You can find some of that discussion by searching for "culture" and/or "ethos".
However you should be aware that that is a area scheduled for a later stage of development.
1) Discussing a good AI is immediately relevant, because it's something needed in this stage of development.
2) However discussing how the AI interacts with a species culture is much more theoretical, since what kind (if any) culture FO will have is undecided.
You are welcome to discuss either in the brainstorming section. But you should know that something might come of #1), while nothing for a long time will come from discussing #2).
However you should be aware that that is a area scheduled for a later stage of development.
1) Discussing a good AI is immediately relevant, because it's something needed in this stage of development.
2) However discussing how the AI interacts with a species culture is much more theoretical, since what kind (if any) culture FO will have is undecided.
You are welcome to discuss either in the brainstorming section. But you should know that something might come of #1), while nothing for a long time will come from discussing #2).
Re: Game AI thoughts
Right now it is important in AI development to update AI to somehow build ships and buildings, research techs, and (especially important) to be aware of fuel limits and move its ships accordingly. After this will be done and some combat AI will be also complete, then FreeOrion would be somehow playable and we may implement tougher things in AI development.
I added to linux compile wiki howto link to Eric Python IDE(it is open source multiplatform IDE) which has refactoring ability and developing AI with it is a lot of easier than with normal text editor.
I added to linux compile wiki howto link to Eric Python IDE(it is open source multiplatform IDE) which has refactoring ability and developing AI with it is a lot of easier than with normal text editor.
-
- Space Floater
- Posts: 43
- Joined: Sun Jan 09, 2005 4:00 am
- Location: Canada
Re: Game AI thoughts
First, I would like to stipulate that I am a game designer, but in no way shape or form am I a programmer. Take the following with the appropriate grain of I-don't-know-how-hard-this-is salt.
The way that the AI should act should be related to the race and/or government that they are playing as. We could create AI templates that are military focused, research focused, expansion focused, etc. If a race is described as 'Xenophobic' for example, they would use the 'military' template, use one of the military- or slave-focus governments, and would choose to enslave or exterminate alien populations.
If the game had 'modular' AI templates, users could create 'new' AIs on the fly: For example, if the Creative AI template focuses on research, and the Xenophobic template focuses on nasty NBC weapon builds, the AI could do a lot of research on NBC technology and improvements.
Here is the big idea: mutually exclusive AI templates for overall racial tendencies mixed with smaller non-exclusive AI templates tied to the optional racial abilities.
The way that the AI should act should be related to the race and/or government that they are playing as. We could create AI templates that are military focused, research focused, expansion focused, etc. If a race is described as 'Xenophobic' for example, they would use the 'military' template, use one of the military- or slave-focus governments, and would choose to enslave or exterminate alien populations.
If the game had 'modular' AI templates, users could create 'new' AIs on the fly: For example, if the Creative AI template focuses on research, and the Xenophobic template focuses on nasty NBC weapon builds, the AI could do a lot of research on NBC technology and improvements.
Here is the big idea: mutually exclusive AI templates for overall racial tendencies mixed with smaller non-exclusive AI templates tied to the optional racial abilities.
Professor Hernandez, Human ambassador to Silica:
"Hey, rocks are people too!"
Black Dawn
"Hey, rocks are people too!"
Black Dawn
Re: Game AI thoughts
Bear in mind that the AI will act as a player, not a character, and will be playing to win, taking into account racial ideologies only inasmuch as they affect gameplay.
Warning: Antarans in dimensional portal are closer than they appear.
- Krikkitone
- Creative Contributor
- Posts: 1559
- Joined: Sat Sep 13, 2003 6:52 pm
Re: Game AI thoughts
The AI is not playing AS a race/government, it is playing WITH a race/government.Black_Dawn wrote:The way that the AI should act should be related to the race and/or government that they are playing as. .
An AI should recognize the strengths and weaknesses of its tools (race/government), but it should not try to BE its tools.