AI current state of affairs...
Posted: Fri Sep 12, 2014 12:59 pm
First and forth most I want to thank everybody that have put work in producing Freeciv. I've been coming to the project every now and then over the years, and I can certainly assert that it's current state is probably the most well done version there is.
I've been following the mailing list for a couple of month now, and been getting an idea of how the project exist and who are the players. I think the whole Bazzar model is in full sail and most probably works the best for most of the issues that arise. It is my understanding that the Bazzar model does work the best for problems that need to be handled in simple mechanical terms and more agile for groups of people that do not work with a managerial control.
AI seem to have a lot problems from my point of view, every now and then a new problems appears because of the new elements that are put in the whole structure. There very good to fix the problem because of their whole knowledge of the code, but it seems that putting out the fires all the time will not solve the problem in the long run. It seems that the whole AI architecture needs to be thought out from the beginning, but in this case the Cathedral model makes more sense; the AI should be the central pivot of the code in how its handled. All other parts like GUI, network, sounds, etc. are to be respondent from the AI perspective, sort of the main column in the structure.
There are certain problems that need to be taking care of AI shouldn't have to deal with the values that are raised in every turn as the sole decision point, AI should have strategies regarding in the way that wants to win the game, the only way to do that is by keeping a trend of decisions throughout the game. For that nations need to have identity on the information that will produce from turn to turn. Make diplomatic decisions in order to ally with another nation if the other nation intelligence shows that their goals are compatible with one another.
It seem to me that the only way to tackle this is that nations need to keep a history of the information that it gathers, so that it can draw a timeline and compare it's previous decisions to assert that it's present complies to it's goals of development. Various axis of decision values are to be kept resources, science, growth, expansion, happiness, etc. decisions are to have different values on the different axis, the sum of all decisions point to a certain axis. Certain axis make more sense to have a bigger incidence on a certain point at the game, depending on the strategy that want's to take. It makes sense that AI behavior could be designed by mod creators from this point of view.
Well this is my point of view of the problem for now; any comments are appreciated.
I've been following the mailing list for a couple of month now, and been getting an idea of how the project exist and who are the players. I think the whole Bazzar model is in full sail and most probably works the best for most of the issues that arise. It is my understanding that the Bazzar model does work the best for problems that need to be handled in simple mechanical terms and more agile for groups of people that do not work with a managerial control.
AI seem to have a lot problems from my point of view, every now and then a new problems appears because of the new elements that are put in the whole structure. There very good to fix the problem because of their whole knowledge of the code, but it seems that putting out the fires all the time will not solve the problem in the long run. It seems that the whole AI architecture needs to be thought out from the beginning, but in this case the Cathedral model makes more sense; the AI should be the central pivot of the code in how its handled. All other parts like GUI, network, sounds, etc. are to be respondent from the AI perspective, sort of the main column in the structure.
There are certain problems that need to be taking care of AI shouldn't have to deal with the values that are raised in every turn as the sole decision point, AI should have strategies regarding in the way that wants to win the game, the only way to do that is by keeping a trend of decisions throughout the game. For that nations need to have identity on the information that will produce from turn to turn. Make diplomatic decisions in order to ally with another nation if the other nation intelligence shows that their goals are compatible with one another.
It seem to me that the only way to tackle this is that nations need to keep a history of the information that it gathers, so that it can draw a timeline and compare it's previous decisions to assert that it's present complies to it's goals of development. Various axis of decision values are to be kept resources, science, growth, expansion, happiness, etc. decisions are to have different values on the different axis, the sum of all decisions point to a certain axis. Certain axis make more sense to have a bigger incidence on a certain point at the game, depending on the strategy that want's to take. It makes sense that AI behavior could be designed by mod creators from this point of view.
Well this is my point of view of the problem for now; any comments are appreciated.