AI cities (colonization and management)

[kyrub]

This AI thread takes the whole part of

- choosing the right place to settle
- managing the buildings & unit training right
- speed building
- leaving the right units in cities for defense
- building roads

(plus the whole AI economy, if you like to speak about it - but it was largely repaired already)

Feel free to discuss it here.

[kyrub]

Conditions for starting a colonization of other continents could be influenced by AI wizard's personnality...

Expansionist essentially set as possible after turn 30?
Militarist could use number of armies as a determining factor. Armies > X?
Perfectionist might want at least 2-3 villages built up with the caveat at expansion occurs if there's no terrain for new hamlets.
Pragmatist might want 2 villages built and a set number of armies to guard them prior to expansion.
Theurgist might consider expanding if there's no more 'readily' available mana source on their continent.

[SIZE="1"]moved from AI overland thread by kyrub[/SIZE]

[Catwalk]

Have you made any headway in teaching the AI to garrison cities from other cities, or do cities still have to garrison themselves? This holds it back quite a bit.

[b0rsuk]

There's something strange about the way AI garrisons cities. Many times units seem stuck in there and don't act on opportunities. Sometimes I walk around AI cities with relatively weak army, and the AI doesn't make the garrison chase me. Also when I have some cities of my own close by, AI could be able to conquer them easily if it bothered to make use of city garrison. This is especially aggravating on Myrror. Generally I can conquer a couple of AI cities with the same killer stack and don't have to worry that AI will recapture them instantly using enchanted roads. A single unit would be enough.

If this behavior has downsides, could it at least be used for Aggressive wizards ?

[kyrub]

Have you made any headway in teaching the AI to garrison cities from other cities

I did not make it, it was already present in the game. I even had to tone it down seriously. What I omitted till now was the defense of Outposts. At first I thought it would be a good idea, but tests showed it did not work.

Why? There is one general problem in AI we should speak about. MoM uses very slow movement (if you are not standing on a Myrror road). When a stack is selected to attack or to help to defend its own city, all other AI units are told to avoit the target... for THIS turn. Next turn, a new unit is chosen to attack / defend. And so on.

When it comes to attack, it may be not so bad. Concentrated attack is what human player fears. But you will see 6 or more stacks attacking a distant (non-human player) city and even when it is captured the others move during next 10 turns to arrive there.

The right disaster starts when this easily applies on defense (it did). In a big empire, AI sent many stacks to its distant city. When they finally arrived, the city had already a garrison on its own, but AI lost incredible amount of time. This was worse with higher difficulty because ideal garrison was much higher. I decided to put it quite low, and I stopped AI making defending decision every turn. This had major effect on AI concentration on attack, but combined with big-stacks-emerging-from-cities, you know see less defended AI cities.

But the problem is still present, if you ask me. And I don't want to stretch it to outposts (which would help early building of garrison and generally speed up the city development). Any thoughts?

There's something strange about the way AI garrisons cities. Many times units seem stuck in there and don't act on opportunities.

I can see your point but it is generally hard to create a totally new sort of AI behaviour. But I will think about it since it would make the AI defense much more interesting.

[Catwalk]

I did not make it, it was already present in the game. I even had to tone it down seriously. What I omitted till now was the defense of Outposts.
...
But the problem is still present, if you ask me. And I don't want to stretch it to outposts (which would help early building of garrison and generally speed up the city development). Any thoughts?

Ah, I misread you earlier then. As I said, I haven't been paying that much attention to Insecticide :P
This looks like a tough one, and a pretty important one as it ties into all aspects of troop movement. Which is one area where the AI fails in a spectacular way, rendering it harmless.

[SDragon]

Why? There is one general problem in AI we should speak about. MoM uses very slow movement (if you are not standing on a Myrror road). When a stack is selected to attack or to help to defend its own city, all other AI units are told to avoit the target... for THIS turn. Next turn, a new unit is chosen to attack / defend. And so on.

Hmm... Does this mean that there is only 1 target for attack or defense per turn or that there is 1 attack and 1 defense target per turn?

I guess the first question would be can there be more than 1 attack/defense priorities and if so, could this be modified based on personality as well? Militarist at war might have 1-2 attack priorities and only 1 defense priority. Other types might have 1 attack priority only but 2-3 defense priorities.

Then, based on the priorities, army distribution and movement can follow? Sounds complicated so I'm hoping the algorithm code might actually be more aligned.

[kyrub]

I guess the first question would be can there be more than 1 attack/defense priorities and if so, could this be modified based on personality

As many targets as there are idle stacks not going anywhere. Lowering number of targets artificially would probably lead to much less expansionist AIs. Some of the strongest AI results in Insecticide are due to its new expansionist nature.

[Catwalk]

How does the Insecticide AI go about choosing which buildings to construct? Vanilla seemed to make horrible decisions.

[kyrub]

Vanilla was very poor & heavily bugged. Insecticide is miles better and free of bugs.


AI uses a basic sequence when building a city. This includes first few defenders and a given set of buildings. This feature was not functional in Vanilla, so you could witness AI towns without barracks...

Then it uses a pre-set RNG weights for different kinds of buildings (economical, food, military), it modifies them with the objective weights. Then it throws the dice...

There are some extras, like incentive to build settlers and engineers early. I am quite content with current setting. All I am trying to do currently is to make that more "Wizard's objective specific", this will have massive impact on the quality of the game, we will see a definite difference. I hope. - Also, the decision whether to build a unit or a building needs a revision. I need to insert an emergency situation calculation. Not easy.