Waterloo AI Questions

[mitra76]

No Corps AI has been not touched, because there has been not be the physical time to change and test it properly before the release. The change at corps and army level are foreseen for the patches.

[Marching Thru Georgia]

That's too bad. Combined arms attacks involving an infantry division and cavalry division are not possible. Does the division AI know how to use infantry and cavalry brigades together in a combined arms attack or defense?

[mitra76]

That's too bad. Combined arms attacks involving an infantry division and cavalry division are not possible. Does the division AI know how to use infantry and cavalry brigades together in a combined arms attack or defense?

In the schemas cavalry and infantry spots are mixed so they can deploy and fight alongide if mixed in teh same division.

[Marching Thru Georgia]

Yes, but does the cavalry stay out of the fight until the enemy is weak? Will the division commander then order the cavalry forward? Or does the cavalry still behave like infantry with fast legs?

[mitra76]

The division AI in fight control brigades assigning spots and stances (and guard special stance for the cavalry), so cavalry remains out of fight if the schema selected place it out of range and with low or no stance. But it never order to attack specific targets at specific moments

[Marching Thru Georgia]

But it never order to attack specific targets at specific moments

It does not try to take advantage of an enemy weakness such as a hole in the enemy line?

[mitra76]

Hole is graphical thing, in the code you cannot evaluate this if not making crazy calculations, you have to leave to the brigade commander and to the regiment path checks to use the open spaces in the lines. Of course you can know that a enemy is in trouble and in the patch i'm working to use this for some specific reaction under specific condition but not linked specifically to the cavalry, because in any case the only divisions with cavalry inside were the prussian brigades and their role was mainly advance guard and exploration not combined attack also because they were mainly from the landwehr, so it is not a very necessary work on this level. At this level the only presence of cavalry nearby force the enemy to remains in squares creating possibility for the infantry.

The breaking impact of cavalry is something we need to manage at corps level and also at army level, but having clear how historically it worked doesn't mean have the easy design solution (is not only a question of AI but also to create interaction functions with the environment) and a lot of tests will be necessary

[Marching Thru Georgia]

A hole can be mathematically represented fairly easily by counting flags along the line. It's not something that has to be done every cycle but could be checked every few minutes. It can be used by both the attacker and defender so they can properly commit their reserves.

[mitra76]

And how you can know what is the line? the fixed POV of checking? where it starts? where it ends? the deep of checking respect the POV (the gap is closed 50 yards behind? what if is the ground cover the view of observer along part of line?

[Marching Thru Georgia]

OK. A division commander looks at the enemy line that he is facing. He uses a square 200 yd long on each side, (I assume a battalion is 100 yd wide). He places the square on the enemy line across from his left flank. If there is an enemy flag in the square there is no hole. He moves the square 200 yd. to the right and measures again. He continues to do this until he reaches his right flank. If he finds a place where there is no flag in the square, it indicates a hole in the enemy line.

This is the simple case of a straight line that is in complete view. But this is how we can begin to develop an algorithm for hole searching.

The next step that the SOW AI needs to take is learning tactics. Currently it only knows how to advance and keep pushing the enemy. It does not know how to take advantage of opportunities.