[Lvlug] idea for AI algorithm

[Christopher Hever]

2008/6/17 Mark Stanley <mstanley at technologist.com>:

> For an improvement, look up "closest point of approach" (CPA) to see how
> sailors do a similar set of calculations.  CPA will not only tell you if
> someone is headed in your direction but how close you will be and when and
> where that will happen.  Add to that your knowledge of the terrain and it
> could be a real one-two punch in your favor.

I looked that up and it is kind of similar. The assumption I make here
is that friendly units are stationary defenders. The goal of the
algorithm is to gather data for assessing threats; if I know, for
example, that five human horsemen are approaching my keep, I'd better
re-up with the archers that will take them down easily.

Also, a game like Wesnoth has discrete time and space, but if I
applied this algorithm to a real-life situation—which I won't have an
opportunity for until I work for some kind of defense contractor—I'd
of course be maybe more interested in estimating the slope of the
distance function for each enemy unit and seeing whether it were
decreasing.

BTW—I'm taking a philosophy course online atm and the average grade is
really low because people argue with their midbrains instead of
thinking. It's crazy. Sometimes it seems like plain old unbiased
machine intelligence on its own is a good enough opponent, without
having to worry about any really special optimizations.

-- 
In autumn when wind and the sea
Rejoice to live and laugh to be
And scarce the blast that curbs the tree
And bids before it quail and flee...