Plane pathfinding

[ginekolog]

From my little research viewtopic.php?f=11&t=13828 we concluded that ATM MAJOR FPS blow is air units sea distance above 500.

On my old barton 3200+ i get 30 fps with 500 armfig with sight distance 500 and only 1 FPS with distance 710 (BA's radius).

Perhaps this is what Tobi said that colision detection check is sight distance + 6 - bigger radius, more hungry check?

IMO if unit has "nocolision = 1" tag than those hungry cheks should be disabled. Are they?

[eriatarka]

Perhaps this is what Tobi said that colision detection check is sight distance + 6 - bigger radius, more hungry check?

I think it's hitsphere radius + 6, not LOS radius + 6.

[eriatarka]

It's a pathfinding problem. The pathfinder is aware of the positions of all objects already, it just doesn't use that information at all when calculating paths. Simply put, aircraft just shouldn't be assigned paths that are currently occupied, or in use, by other aircraft. It's possible to make the pathfinder intellegent enough to do so, especially with units as simple in their pathing as aircraft are. Fixing this on any level higher than the pather is asking to rape CPU usage.

I'm not sure it's really that easy. After all we are talking about moving objects here, so just taking the location of objects at time t into account won't be enough. Either the pather is reinvoked regularly to correct the path (is it?), or it has to take probable future positions of all relevant objects into account. Both sound expensive to me; more so than local collision avoidance, but I could easily be wrong.

[KDR_11k]

Aircraft don't use pathing, they turn straight toward their target. If anything is in the way they either bump into it and then try to correct their course or just fly through it (if collide=0).

[Scratch]

collide=0 in the new BA, definetly noticed performance increase.