Doors that don't run NS or EW

[grayman]

While debugging #720 today, I noticed that my AI can't walk through a door oriented along the SW/NE axis (i.e. rotated 45 degrees). He opens the door, then circles at the door opening, repeatedly bumping into the door jamb.

 

At first I thought this had to do with the AI's bounding box never changing its orientation, and not fitting through the angled opening.

 

But when I turned on debug drawing, I saw that the angled wall into which my door was set hadn't factored into how the AAS areas were calculated. The angled wall, the door, and the floors on either side of the door were all part of the same rectangular AAS area. So pathfinding found a nice straight path to the far side of the door, but unfortunately it went through the door jamb. That ain't gonna work. (Pathfinding rule: you can walk in a straight line from any point inside an AAS area to any other point inside the same AAS area, barring any dynamic obstacles.)

 

So my question for mappers is: have you used doors in your maps that aren't oriented NS or EW? And did you have to do anything special to let your AI use them?

 

Thanks.

[Fidcal]

Angled doors are so rare I can't think of a single one.

 

So, are you saying AAS is faulty in this regard? It calculates a path that cannot be followed? And is this true for angled gaps of any kind, ie, no doors involved, two monsterclipped crates at an angle with a gap between?

[grayman]

So, are you saying AAS is faulty in this regard? It calculates a path that cannot be followed? And is this true for angled gaps of any kind, ie, no doors involved, two monsterclipped crates at an angle with a gap between?

 

I haven't tried that, but I'll bet it doesn't let the AI path correctly between the crates. I'll try a test case tomorrow where the only way from path_corner A to path_corner B is between the crates. I'll bet that even if there's a way around the crates, if the crates are on a direct line between A and B it'll probably fail.

 

I could easily be wrong; perhaps other conditions are present at dmap time that don't exist in the test map I'm using.

 

 

 

 

[grayman]

I created an area with two angled crates where the only path past the crates was between them.

 

dmap correctly carved up the floor into AAS rectangles, placing a few small ones between the crates.

 

So the AI walked between the crates correctly.

 

I then placed an angled wall farther along the path and cut a door-sized opening in it.

 

dmap again placed a few small AAS rectangles inside the wall opening and the guard had no problem walking through the opening and the crates.

 

However, I noticed that the small rectangles inside the opening were inside a larger rectangle that included the angled wall. That's the first time I've noticed AAS areas inside other AAS areas. I don't know if this is legal or not, but the guard had no problems.

 

When I added a door to the angled opening, the guard bumped through the opening when the door opened away from him, but got stuck on the way back, when the door opened toward him.

 

So there are issues with angled architecture and how AAS is laid out. Maptesting by the mapper should expose any of these issues, though, so we'll probably never see finished maps with these problems.