Well, since this is becoming a more general AI discussion thread...
I did a few trials to see how well the pathfinding algorithms of the AI did.
The results were quite interesting. Allowing for the ai's inability to negotiate small gaps between buildings, I made a small maze of buildings to see just where he could reach and where not.
In each instance, I restarted the trial with AI at the centr, and issued a move order.
There was a cutoff point that seemed to extend to the left at right angles (that I marked with a black line) to the direction the AI was facing; as you can see he could find his way out of the relatively complex maze to his left, but had an inablility to travel a few metres more further back, to a point which was quite easily accessible from the first point.
On the right, the point at which the AI got confused was further up relative to the left hand side one. The maze jutted out a bit more here than in the left, and this seemed to cause problems for the AI.
Also, note the way the path that the AI took was not the shortest distance, but had a tendency to hug the sides of the buildings. In the few pathfinding algorithms that I've written (not for ofp), the paths that I end up with look similar to this, with similar limitations. The cutoffs are also similar; any path that requires 'backtracking' (well, not backtracking but travelling in the reverse direction to any other direction in the path, provided that the end is not in reach) is unable to be found.
Id like to know exactly what the process that flashpoint AI uses to achieve this; keeping in mind BIS has one up on games that require a pre-set bunch of co-ordinates for their AI to follow- this process is dynamic, ie I could re-arrange the 'maze' in any order I liked and the AI would perform equally well.
Sorry for the rambling post, just thought I'd introduce another aspect of the AI for discussion