Player designed scenarios and the AI

[lucero1148]

When it comes to designing scenario's especially if it is to be played against the AI would it be possible to give the AI whether for the attack or defense mission orders. Such that if it is a static defense the defending AI will hold it's position instead of making kamikaze runs?

Or if there are reinforcements within the scenario that these units have a set of preset orders in which to follow. AS well there could be a subset of orders for it to follow in the case the attacker has succeeded or failed certain objectives?

Would this be difficult to do or out of the question?

[Wicky]

I think there is a reference from BFC to your query in this bone thread

[lucero1148]

Nope that isn't what I was suggesting. That explains the conceptof the unit Order of battle and how they are to be integrated in the game.

what I am suggesting is for the AI to be given specific strategic/tactical orders bt the designer so that the AI will be "smarter" in play. With the possibility of subset orders for it to choose the best course of action if the human player is past point A on the map. that way say a AI reinforcement when it arrives will actually have a couple of choices, 1. attack if the human player has not reached point A or defend if the player has reached or passed point b. This way the AI doesn't blindly advance and getting itself massacred.

[Wicky]

Somewhere in there is a post from Steve - quick search reveals

One other bone...

The new AI allows for pretty powerful AI "hints". These are things that the designer puts down on the map and instructs the AI how to go about using them. For example, identifying the primary defensive points or the best avenues for advance. These are things that the AI had to figure out on its own. While I think it did a better job for a game of its type than any other generic AI ever has, it certainly wasn't as good as any of us would have liked. Now the designer has the ability to help the AI figure out stuff and therefore sculpt the AI's behavior to a fairly significant extent.

Note that this is not the same as scripting the AI. The AI will still be able to use its judgement to do or not do something at a given time or later. The difference is that it doesn't have to figure out as many of the basics that are relevant for evaluation. For example, if the designer specifies 3 possible routes of advance and 4 strong points and 2 important objectives, the AI can weigh all of these things to determine which route/s to take, which strong points to focus on, which objectives to bust butt for and what NOT to do. It also has the ability then to better evaluate unexpected stuff along the way, such as "I've been told this is a VERY important thing to take, but some enemy is peppering me with light fire. I'll ignore it" vs. "I've been told this is a minor objective and I wasn't all that interested in taking it, but I am very interested in shutting down that annoying fire I'm getting on my flank!". That sort of stuff.

This ties into the new Editor interface because without the new UI I'm not really sure how easy it would be to do this sort of stuff. With the UI in place, it's a piece of cake. Charles can then focus on the "hints" instead of how the user makes them known to the AI.

[Peterk]

Hmmm...I would argue that the number one AI problem in CM is that when defending and a flag is lost, even if victory is still atainable by the AI depsite the loss, it will compromise its defenses on the other flags it holds to try and regain the one it lost.

Have a feeling that is more what lucero had in mind.

[lucero1148]

Hi

thanks Wicky guess I should learn how to read. That addresses some of the issues and Peterk confirms my other concern.

The AI defense could be coded to stay in place and defend flags until opponent losses are at a certain percentage before it initiates a counterattack. that would really mix it up for single play.