Wow, having seen A* pathingfinding quite a bit, that is pretty good. However, I would like to see an image or two of how the pathfinding is when going through a maze. It might lead to making the pathfinding in mazes better if we can see how they will react.
Well, Proper A* will find the optimal path... Our A* does of course have a node limit in built. If this is exceeded, all bets are off. At the moment I handle this by simply picking the node closest the goal, and then 'backing up' a bit, in case that happens to be in a cul-de-sac. When this 'temporary destination' is reached, a new path is calculated.
The node limit with mazes probably wont work so well, it will almost certainly expand many nodes trying to go around... if you limited such opportunities (with a very 'packed' map) then it would come down to whether the node limit was hit or not... if so it will pick the closest node and backoff a few steps... unlikely to be very effective in a maze.
eg. I had to increase the node limit to 4096 (probably too high for regular in game use) to get through this...
![](http://i687.photobucket.com/albums/vv231/silnarm/glest/astar_on_maze.jpg)
Of course, once we go hierarchical, even with the node limit, it will handle this with ease
![Smile :)](https://forum.megaglest.org/Smileys/SoLoSMiLeYS1/happy.png)
As well, I am curious how the units will move exactly in crowds of units. Would you be able to upload a build (win32) of your modified engine for the pathfinding, which shows the pathfinding like that? I really would like to test some things!
Group movement is still an issue... I have been looking at options to make this more intelligent. "Windowed Hierarchical Cooperative A*" (WHCA*) looked promising, but there are serious issues that would need to be solved to get it working 'in the wild'. Also if I do run with that, I have to figure out how to make it work with Hierarchical Pathfinding A* (HPA*) [Actually, we'll be getting a variant, HAA*]
I have some ideas I'm going to play with after we get a new release out, to 'manage' resource areas and choke points on a per team basis... If that works well enough, maybe a 'proper' cooperative movement algorithm will not be needed.
I'd prefer a prebuild, and besides, I can't figure out which I am looking for. I want the one that is used for the pathfinding shown above...
I'll add options to switch algorithms and turn on/off timing and debug textures... binaries soon...