TA3D

Really cool pathfinder, maybe we can ask them for this?

http://gamesrenter.com/hd-streaming-tra ... r-trailer/
maybe you can watch by that
and this ///
http://bbs.taclub.net/taclub/newta/thre ... 3-1-1.html

This is really impressive, unfortunately we can't push the particle flow model that far in TA3D because it requires too much changes in data structures which are required to keep OTA compatibility

Yeah, looks cool But TA pathfinding is not bad at all

I know, trying to change OTA's pathfinder atm....
So, maybe will used TA3D's pathfinder in (very very little possable) for TA.....

I've been involved in some discussions over at the official SupCom 2 forums regarding pathfinding that you might be interested in:

- http://forums.gaspowered.com/viewtopic.php?f=50&t=49916
- http://forums.gaspowered.com/viewtopic.php?f=50&t=49870

The second link contains quite a bit of flaming, but eventually employees of GPG participated in the discussion and divulged some information, otherwise unpublished, on SupCom 2.

The pathfinding in SupCom 2 is not that much of an improvement over what was used before (i.e., SupCom 1 + FA). The SupCom 2 engine still relies on A* pathfinding. You will see from the discussions at the above links that GPG significantly changed the gamestyle in SupCom 2 compared to SupCom 1 (i.e., dumbed it down) because even with Flow Field the engine still bottlenecks. In other words, had they not changed the gameplay and left it as sophisticated as SupCom 1 and TA, then they would have produced another "system killer" game. I propose a solution to the pathfinding routines (see first link) that might solve the engine's limitations and that might be of use to the TA3D project.

That's interesting. TA3D uses a global and a local approach. It extracts local information in a way that seems similar to what you described : a kind of energy map is derived from map data (height, obstacles, ...) then blurred to allow units to detect obstacles before reaching them and knowing more or less where free space is (it computes the gradient to decide where to go). The local approach is used when a unit wants to move through a group of units, the group moves away to allow the former unit to move freely (it works because units emit "energy" too). The global approach is used to compute full paths at map scale (but using the local energy map to help avoid obstacles). Pathfinding is so critical it has its own thread with its own task scheduling code. A fully local approach would be great for performance but I fear it may not find good enough paths.