Creating a dynamic research pathern. |
Right now, i'm working on making my AI research things in a less predictible manner. Right now it just pick things on a list in a sequential order wich means that all AI gets the same things (at least until racial techs start to be researched). My new AI does quite well in all area (it does expend and build stuff quite well) but it's research pathern is lame. I could take hours to make a very elaborate research queue but i would prefer to let the AI choose what it need and pick appropriate tech to level up. I also don't like it to queue up 5-10 level at once as it waste a reseach slot for a long time and the other research fields are left destitute.
I tryied to use the random function, but the random function run each turns and when the roll fail, the item currently researched disapear and research points are wasted. Worst the AI don't fill the free slot so the AI stop researching (though there might be a work around)...
Bah, i'll try something else later today, for now i need some sleep...
Use the AI state to prioritize research
What you might want to do is use the AI's current state (explore, infrastructure, attack, defend, etc) to push certain techs ahead of others using conditional statements.
I plan to this in two ways for the Balance Mod AI. The first will be set percentages for each research type (theoretical, weapons, applied, custom ones etc.) and the second way is in the research script itself to push "needed" techs based on the current state.
-----
Grr...
I forgot to escape my greater signs, so I lost most of the rest of that post... crap...
Just in case you didn't know, there is a file in the SE5/Docs folder that explains most of the basic scripting. Even if you know about it I can understand how long getting a hold of the calls can be. I know I still haven't gotten many down...
I'm tired and running low on waking moments, so if you want, I could take some time to explain how to do what I proposed. It'll make more sense than my first ramblings, too. I'll probably think about the best way to do it during school, so I'll have a smoother way of doing it by tomorrow. Or I'll just make a giant topic in the mods forum about dynamic research, and give an example script.
-----
Watch for my mod - codename: Dimensions - coming by Summer, 2007!
Setting up research priorities...
It's not too hard.
First thing is you'll want to organize the techs in the research script individually or for stock-type level amounts, perhaps in jumps of 2 or 3 levels at a time. I sorted further into the groups that matched the different tech groups.
In the empire's main script file, there is a function "Set_Race_Research_Spending". Essentially create a case structure for the AI's state like this example:
set ai_state := Sys_Get_AI_State(sys_long_Player_ID)
case (ai_state)
AI_STATE_EXPLORE_AND_EXPAND:
call lst_Research_Spending_Group_Name.add("Cultural Advancement")
call lst_Research_Spending_For_Group.add(2)
call lst_Research_Spending_Group_Name.add("Theoretical Science")
call lst_Research_Spending_For_Group.add(20)
call lst_Research_Spending_Group_Name.add("Applied Science")
call lst_Research_Spending_For_Group.add(48)
call lst_Research_Spending_Group_Name.add("Weapon Technology")
call lst_Research_Spending_For_Group.add(30)
AI_STATE_INFRASTRUCTURE:
blah blah
AI_STATE_ATTACK:
blah blah
AI_STATE_DEFEND:
blah blah
AI_STATE_NOT_CONNECTED:
blah blah
endcase
-----
Ok...
In the empire's main script file, there is a function "Set_Race_Research_Spending". Essentially create a case structure for the AI's state like this example:
set ai_state := Sys_Get_AI_State(sys_long_Player_ID)
case (ai_state)
AI_STATE_EXPLORE_AND_EXPAND:
call lst_Research_Spending_Group_Name.add("Cultural Advancement")
call lst_Research_Spending_For_Group.add(2)
call lst_Research_Spending_Group_Name.add("Theoretical Science")
call lst_Research_Spending_For_Group.add(20)
call lst_Research_Spending_Group_Name.add("Applied Science")
call lst_Research_Spending_For_Group.add(48)
call lst_Research_Spending_Group_Name.add("Weapon Technology")
call lst_Research_Spending_For_Group.add(30)
AI_STATE_INFRASTRUCTURE:
blah blah
AI_STATE_ATTACK:
blah blah
AI_STATE_DEFEND:
blah blah
AI_STATE_NOT_CONNECTED:
blah blah
endcase
Not sure it will be of use to me as my research script is quite flexible already. Applied and Weapon Technology are splitted in half and can get things done quite fast with low level techs or with a research powerhouse. Theorical get tech on a need basis (like when a required tech show up), i don't set any theorical techs (i removed the part of the script that picked tech to research in alphabetical order). The Cultural value is so insignificant that i don't bother with it. Besides i don't put any tech in there before the second age so a required tech almost always take the slot.
So what i did is i set tech in 1 to 3 level queue. When in low priority, they have less level to queue and i don't set a second instance of it in the list so a free slot to pick it up. This increase the likelyness that a wanted tech get the slot and get researched. My testing show great promise for my scripts and i rooted out all the bugs i could find, it's bullet proof 
.
Here an exemple :
if (Sys_Get_AI_State(sys_long_Player_ID) = AI_STATE_ATTACK) then
call Add_Tech_Area("Point - Defense Weapons", 2)
call Add_Wanted_Tech_Areas_For_Designs(2)
if (lng_Research_Age_Max_Level = 40) then
if bool_Race_Uses_Cloaking_Devices then
call Add_Tech_Area("Cloaking", 2)
endif
endif
call Add_Wanted_Tech_Areas_For_Espionage(1)
if (lng_Research_Age_Max_Level = 20) then
call Add_Tech_Area("Ship Weapon Mounts", 2)
call Add_Tech_Area("Base Weapon Mounts", 2)
call Add_Tech_Area("Shields", 1)
call Add_Tech_Area("Organic Armor", 1)
call Add_Tech_Area("Crystalline Armor", 1)
endif
call Add_Tech_Area("Sensors", 2)
call Add_Tech_Area("Armor", 1)
if (lng_Research_Age_Max_Level = 20) then
call Add_Tech_Area("Psychic Sensors", 1)
call Add_Tech_Area("Temporal Sensors", 1)
call Add_Tech_Area("Scanners", 2)
endif
call Add_Tech_Area("Point - Defense Weapons", 3)
call Add_Wanted_Tech_Areas_For_Designs(3)
call Add_Tech_Area("Repair", 2)
if (bool_Race_Uses_Mines) then
call Add_Tech_Area("Mines", 1)
endif
if (bool_Race_Uses_Satellites) then
call Add_Tech_Area("Satellites", 2)
call Add_Tech_Area("Unit Weapon Mounts", 2)
endif
if (bool_Race_Uses_Fighters) then
call Add_Tech_Area("Fighters", 2)
call Add_Tech_Area("Fighter Weapons", 2)
endif
if (bool_Race_Uses_Drones) then
call Add_Tech_Area("Drones", 2)
endif
if (bool_Race_Uses_Weapon_Platforms) then
call Add_Tech_Area("Weapon Platforms", 1)
call Add_Tech_Area("Unit Weapon Mounts", 2)
endif
if (bool_Race_Uses_Troops) then
call Add_Tech_Area("Troops", 2)
call Add_Tech_Area("Troop Weapons", 2)
endif
call Add_Wanted_Tech_Areas_For_Designs(1)
call Add_Tech_Area("Point - Defense Weapons", 1)
call Add_Wanted_Tech_Areas_For_Hull_Upgrades(1)
endif
//------------------------------------------------------------------------
// Add_Wanted_Tech_Areas_For_Racial
//------------------------------------------------------------------------
function Add_Wanted_Tech_Areas_For_Racial returns boolean
params
max_level: long
vars
item_added: boolean
loop_index: long
tech_area: string
begin
return TRUE
// Add wanted types for Racial Techs.
set item_added := FALSE
for loop_index := 1 to 5 do
case (loop_index)
1:
set tech_area := "Psychic Technology"
2:
set tech_area := "Religious Technology"
3:
set tech_area := "Temporal Technology"
4:
set tech_area := "Crystalline Technology"
5:
set tech_area := "Organic Technology"
endcase
set item_added := Add_Tech_Area(tech_area, max_level)
endfor
end
set lng_Research_Age_Max_Level := 10
if (lng_Research_Age_Max_Level = 10) then
call Add_Wanted_Tech_Areas_For_Planets(1)
call Add_Tech_Area("Ion Engines", 3)
call Add_Tech_Area("Sensors", 3)
call Add_Tech_Area("Armor", 3)
call Add_Wanted_Tech_Areas_For_Designs(3)
call Add_Tech_Area("Cargo", 1)
call Add_Tech_Area("Storage", 1)
endif
That's how i currently coded my research. Of course i may find some more efficient way to do it but it seem to work flawlessly. My empire get research done quite fast...
I succeded.
My latest version of my mod does have a dynamic research. It was hard to make, a pain to test and required me to figure out how the game research actually work. For starter, the game always scan the whole research file each round, wich can become quite slow later on. Second the number in () is the level to reach. Once the level is reached, that line is "dead" and won't do anything but waste processing time. The game will run each lines of code in sequence until all the research slots are filled (in stock it's just 3, in my mod, it's 6). Also, if there is nothing to research in a particular field, the game will pick something from another (i believe the game scan the research tree several times). For exemple, it's possible to run out of Theorical techs quite early in the game, this mean the slot is free and the game will pick the next thing you haven't researched to max level(). If not it goes for stuff in the next age, and last will start researching things in sequential manner (alpha order from what is availiable). The game also have the possibility of researching requirments for a tech. How far it will research to get it is dictated by the current research age value. The flaw in that proceding is that it look for level, not how much you have to spend to do it. The AI can be stock a long time researching an expensive 1 level tech (like Astrophysics and the like) instead of doing something else more pressing...
My new research pathern work differently. Max levels are a value that increment in a loop until all research "slots" are filled and all research points are used. It's also divided in the 4 AI states so i can give different priorities (base level to research) for each fields. For thoses i don't want to be picked too early, i give them low or negative numbers so the AI will pick the other techs first. Since the AI clear research spendings every turns, when it switch states it just pick other technologies and continue spending as usual. It work quite well, but only actual field testing will confirm how effective my system works. I just hope to read about it in the following days when comments start comming...
I did not release my code yet, but i will once it showed it's efficiency and all bugs are rooted out. I don't want to repost it 10 times (already post new versions way too often ).
I've had the same idea
Research was the first section I attempted to do in my AI. Unfortunately, I realized that there was a huge amount of work to be done in this regard. I suggest you manually setup a list of the various technologies (a longlist of IDs) and a list of the type of research (longlist of constants you define elsewhere). Put all techs in this list, and later you can pick and choose which ones to exlcude to make more unique AIs.
When the AI gets around to choosing its research, make a longlist of weighting of the various kinds of technologies. If it's in a war, boost up the weighting on battle technologies. If its production is lagging behind, increase the facility research. Pick a random type from this list, and then choose a random technology of that type.
I suggest using something like this (in psuedo code):
goodtech = false
tries = 0
while (not goodtech) and (tries