Recent work by Overmars et al. on corridor maps is well-suited for pathing many units with circular footprints of varying radii.

I'm not sure if this is your case but if you calculate the path at each frames, perhaps you could save the path in an appropriate variable and calculate the path once every 10 frames or so. Things don't change that much in a single frame to affect path finding. Just a thought.

Thank you for the replies.

Quote: Original post by Sneftel
Recent work by Overmars et al. on corridor maps is well-suited for pathing many units with circular footprints of varying radii.

That seems like a very interesting paper, however 3D pathing is not covered in it it seems. Also I've just fully implemented my solution which seems to be doing a great job :D.

Quote: Original post by Hexmind
I'm not sure if this is your case but if you calculate the path at each frames, perhaps you could save the path in an appropriate variable and calculate the path once every 10 frames or so. Things don't change that much in a single frame to affect path finding. Just a thought.

I limit path calculations to like 3 paths per frame using a queue system. The game is action based so the response time of the units is a factor here.


Seems like this is a difficult topic, maybe I should write an article in the near future on my fairly simple to implement approach.

Quote: Original post by Sneftel
Recent work by Overmars et al. on corridor maps is well-suited for pathing many units with circular footprints of varying radii.

Looks a lot like Voronoi roadmaps. The idea of smoothly interpolating between max-clearance and min-time paths is interesting though...

Quote: Original post by Emergent
Looks a lot like Voronoi roadmaps.

That's its basis, certainly. Actually, the thing I find most intersting about corridor maps is how well they dovetail with dynamic behaviors like collision avoidance. Rather than rubber-banding to the exact, inscrutable polyline one gets out of most path planners, agents can use conventional separation behaviors, and still maintain a good-looking trajectory.

Quote: That's its basis, certainly. Actually, the thing I find most intersting about corridor maps is how well they dovetail with dynamic behaviors like collision avoidance. Rather than rubber-banding to the exact, inscrutable polyline one gets out of most path planners, agents can use conventional separation behaviors, and still maintain a good-looking trajectory.

Correct me if I'm wrong but collisions can still and probably will occur though. It might be so that rubber-banding is not needed for this particular technique, but that goes for most path planners as well. Is there something I missed?