Any AI Improvements?

[Dogrobber]

This is really good news. So good in fact that after years of lurking, I actually joined the forum. Just curious about a couple of things:
1) Will there be any AI improvements? The AI in the original BG games and successor Campaign HPS is well terrible, completely scripted & a battery by golly will continue to move towards its designated deployment hex even if a division of enemy infantry is in its way.
2) Will some sort of C&C be added? The other major flaw of the series IMHO was that unless troops routed you had perfect command & control, ie you could move any troops anywhere regardless of their quality or the quality of the commander or their distance from their leaders. This was certainly not an accurate reflection of the communications or leaders of that era. This is one of the things Frank Hunter really got right in ACW and is reflected in some CW boardgames through use of Command points.
3) I realize that both of these issues are not "oh yeah I'll get that done this afternoon" type issues. That's why I'm just curious...

[rhondabrwn]

I'd have to agree... AI improvements would be quite welcome, but that's a hard nut to crack. The AI in the newer HPS generation games (Campaign Gettysburg, for example)is just useless in the campaign game. As the games introduce expanded options, larger maps, alternative orders of battle and arrival, the older AI approaches seem really challenged.

The scripting approach has severe limitations. I would like to see an AI engine developed that would be database driven (accessible to users and modders)and which would use a chain of command architecture in which army strategy would be determined by the AI based upon (1) scenario objectives and victory conditions (2) an analysis of known enemy positions and strength. (Hey, if we can do it for chess playing computers....)

Once the over-all strategy is set at the army level, general directions would be communicated to the AI routines for corps level and independent formations (if in communication on the game map). Then an AI routine would determine a corps level strategy in accordance with the army directives received (or an independent plan if out of communication). The AI would then drop down to sucessive lower levels, evaluating terrain, objectives, and enemy positions and developing operational tactical orders for each division, Brigade, etc.

This would be programming the thought processes of a skilled commander.

Would this take time to process? Absolutely... I could see the computer grinding away for hours to produce the game plan for the turn. Can I accept playing one turn a day against an outstanding computer opponent? Definitely!

Of course, the "advanced AI" would have to be an option for people who just want a fast game (and a sure victory against the AI).

[SurrenderMonkey]

There is a solution, of sorts - albeit an unwieldy one.

Someone could program a set of 8 or 10 AI's under the relatively simple older system, and then distribute these AI's via zip files. When you go to play a battle, you load one of the AI files 'blind' and then play away. Tried and true strategies that take advantage of AI scripting go out the window. It works, but it's a lot of work.

[lancerunolfsson]

Can I accept playing one turn a day against an outstanding computer opponent? Definitely!

Easier to PBEM that way your computer is not locked up;^) But I think that the path to better AIs definately leads in the direction of the AI recognizing colections of units (eg Brigades, divisions) as intregral wholes that follow specific tactical doctrines relative to thier hierarchy in the chain of comand. (which is maybe exactly what you are saying on some level.)