Unit formations?

[John.d.h]

I agree with the point that the AI will probably have to be updated as we go, in order to keep up with new features, but I think currently it doesn't even properly use the features we already have, so maybe an AI update is in order.

[Little Helper]

for unit formations you probally have to group up the units up to four in order to make the formations I guess..

[jda]

I agree with the point that the AI will probably have to be updated as we go, in order to keep up with new features, but I think currently it doesn't even properly use the features we already have, so maybe an AI update is in order.

I agree. If the AI is outdated (regarding other features already implemented for humans), then updating the AI should be done before implementing a new feature such as this one.
And yes, the AI does have a lot to learn:
- if an unit has two different attacks that hit the same field, the AI will always use only one. AI should use the best attack it has in every situation (e.g. ranged but weaker attack when enemy is farther away, switching to stronger melle attack when enemy is right in front)
- when there a faction has two different kinds of resource-gathering units, the AI will produce and use only one - shoudl be able to determine the best one to produce/use in each case and do it.
- etc
BTW I think there are open tickets for these former two.

@ Little Helper: I don't think that would help. You can already group units up to 16. So I guess it would be just as easy as to make formations with 16 as with 4.

[silnarm]

<OT>
The AI definitely needs work, and it's gonna get it. But a good AI that can 'think for itself' is going to be difficult to build and take a long time to get right (Game AI is not easy, RTS game AI is not even remotely easy).

In the mean time, the current (C++) AI interface I believe can and should be exposed to Lua, which I'm confident will be fast enough (with some slight changes to the way the AI currently works)

Then the more traditional scripted AIs (for specific factions) can be developed, and hopefully by a wider audience than the current C++ interface appeals to
</OT>

I've nothing on-topic to say, sorry... just doesn't seem feasible with Glest's mechanics.

[jda]

The OT part sounds good to me.