[engine] AI/micro_ai idea - would anyone else find this useful?

[Spannerbag]

Hi,
Just curious if anyone would find this idea useful, plus any suggestions as to if it can be done with the existing AI tools.

What I'd like is to have a side whose units are mutually antagonistic, i.e. as well as attacking units on other sides they also attack each other (but possibly with a lower targeting value than units on other sides so given a choice they attack units on other sides).
In one scenario I'm testing this would have been handy but I've got the scenario working well enough with existing sides, just thought I'd ask here to see if anyone else would find this idea useful.
I was thinking of having a MAD micro_ai (Mutual Alliance Disabled).

Any comments very welcome.

Cheers!
-- Spannerbag

[Lord-Knightmare]

Call it maybe ai_type=frenzy? (Like in Skyrim?)

[white_haired_uncle]

I think instead of having this applied to a side, you'd have it applied to a unit (you could then, of course, apply it to all units on a side if you wanted). And it would be contagious. And probably be called 'rabid'.

[Spannerbag]

I think instead of having this applied to a side, you'd have it applied to a unit (you could then, of course, apply it to all units on a side if you wanted). And it would be contagious. And probably be called 'rabid'.

Yep, but the only way I can see of having a side with "regular" units plus some lunatics who attack everything including their own was via a micro_ai?
I don't think there's any way a unit can override it's side's team_name?
Your contagious idea could be huge fun in some situations - but I've no idea how to implement any of this or even if it's possible...

Cheers!
-- Spannerbag

[gnombat]

I can see how this might fit with a campaign storyline in some situations, but...

1. It seems like it would violate the KISS principle pretty bad.
2. It would require changes to more than just the AI, wouldn't it? I don't think it is even possible right now for a unit to attack its own side?
3. It would lead to some weird/pathological situations. For example, suppose a side has some units with leadership. If a unit with leadership were attacked by another unit from its own side, would the attacker gain the benefit of the leadership (while attacking its own leader)?!

An alternative approach might be to have some code which generates a new side for each and every unit, and assigns the unit to that side. I'm not sure this is possible, but it seems like it might work (using wesnoth.sides.create).

[Spannerbag]

I can see how this might fit with a campaign storyline in some situations, but...

1. It seems like it would violate the KISS principle pretty bad.
2. It would require changes to more than just the AI, wouldn't it? I don't think it is even possible right now for a unit to attack its own side?
3. It would lead to some weird/pathological situations. For example, suppose a side has some units with leadership. If a unit with leadership were attacked by another unit from its own side, would the attacker gain the benefit of the leadership (while attacking its own leader)?!

An alternative approach might be to have some code which generates a new side for each and every unit, and assigns the unit to that side. I'm not sure this is possible, but it seems like it might work (using wesnoth.sides.create).

Re. your points, all good.
Leadership (and other abilties/specials that involve interactions between units on the same side such as healing or distract) are irrelevant; these units are all enemies so none of these would be appropriate.
Also, it should be impossible to ally with such a side - which is something I didn't think of until after I posted my idea/query.

My understanding is that a side's team_name(s) determines friends and enemies, i.e. who is "attackable".
I wondered if it might be possible to craft a special team_name (something unlikely to be used such as "UniversalEnmity") that might allow this behaviour to be possible but as I say, no idea how to do this or even if it's possible.
Maybe a more secure way to do this would be to have a key within a [side] or maybe even a new side definition tag for this situation.

Yeah, did think of spawning a new side for each and every unit... but it seemed overkill and I've no idea if there is a limit to the number of concurrently active sides a scenario can accommodate (there are potentially a lot of these maniac units in my scenario).
But yes, that's an approach that could work if you can have ~50 sides in a scenario (maybe less because I could re-use an existing side once their only unit is killed).
Would need to set a few elements for the newly created side, I'm guessing wesnoth.sides.get could do that?
Might try this when I get chance.

Thanks for taking the trouble to reply, much appreciated.

Cheers!
-- Spannerbag

[white_haired_uncle]

Leadership (and other abilties/specials that involve interactions between units on the same side such as healing or distract) are irrelevant; these units are all enemies so none of these would be appropriate.

...

My understanding is that a side's team_name(s) determines friends and enemies, i.e. who is "attackable".

Hmm, you think leadership and healing are restricted to the same side? Not so much. Anyway, it's still a problem based on your original specification. If unit A on side S attacks unit B on side S and unit B has leadership (applied to same team), does A benefit from it? Technically, they're both on the same side, even though they are fighting each other. And that's just one example of weird stuff.

Not that it matters, I can't imagine you can make a unit attack another on the same side during that side's turn. While team does control friends and enemies, they are on the same team. And there's a lot of places that I've seen in code/docs that assume same side = friend, IIRC.

You might be able to force combat using something like [do_command][attack]. Might. Maybe.

What you could do pretty easily is to have a manic unit do damage to adjacent allies when fighting an enemy, though I think I'm veering off from ideas to UMC here.

[gnombat]

Yeah, did think of spawning a new side for each and every unit... but it seemed overkill and I've no idea if there is a limit to the number of concurrently active sides a scenario can accommodate (there are potentially a lot of these maniac units in my scenario).
But yes, that's an approach that could work if you can have ~50 sides in a scenario (maybe less because I could re-use an existing side once their only unit is killed).

The SideWML documentation says, "There is no hard limit on the number of sides a single scenario can have."

That having been said, I don't know how well ~50 sides would work in practice. Possibly there might be other issues (e.g., the AI might start getting confused when there are that many sides).

[Spannerbag]

Thanks for all the thoughts and comments, much appreciated.
I think for now I'll stick with what I have as it works but might revisit this once I have everything else done (if I'm not completely sick of the damn thing by then).

Thanks again.
Cheers!
-- Spannerbag