The problem with pathfinding

[CyclonatorZ]

Just a suggestion I have to offer - if you guys aren't already working on making the pathfinding in Warzone 2100 better, I'd urge you to make it a priority. Coming from a relatively casual player, the often hilariously stupid AI of the tanks is still probably the biggest problem with the game, and it often leads to huge traffic jams and units dying without reason. Is there a reason why there hasn't been any improvement in this area?

Thanks again.

[Zarel]

http://developer.wz2100.net/browser/branches/

We've been working on pathing for ages.

Are you, by any chance, using trunk? I'd recommend switching to 2.3 for the moment - 2.3 has slightly better pathing.

[CyclonatorZ]

http://developer.wz2100.net/browser/branches/

We've been working on pathing for ages.

Are you, by any chance, using trunk? I'd recommend switching to 2.3 for the moment - 2.3 has slightly better pathing.

Yes, and unfortunately the pathfinding really doesn't seem any better than it was when I played the first open source release five years ago. I suppose this is understandable, considering that Warzone 2100 was practically the first 3D RTS ever, but it still can make gameplay in the singleplayer mode a bit of a hassle. Glad to see that you guys have been at least trying to improve it though - I just wish more progress had been made since the original open source release.

[Rman Virgil]

.

Interesting. I find the path-finding noticeable improved. Been observing it closely now as I balance test one of the most complex maps ever made and the minor PF glitches that have arose I've been able to easily fix through map crafting. But then again, what is asked of PF is to be able to handle effectively poor map crafting in legacy maps. However, even in that respect, I know the developers have implemented a number of specific measures in recent times to counter precipitating cluster fracking craft like units being able to destroy poorly placed map features. Still, optimizing PF through map crafting is little understood and less practiced.
.

[Per]

Path-finding speed has improved massively since the original, and several bugs have been fixed. However, the basic (poor) design remains. Making a good path-walking/steering design is a very non-trivial problem. We are working on it, though.

[JDW]

We are working on it, though.

I can certainly vouch for that. I have been following revisions of the trunk version every now and then, and noticed varying performances in path-finding. Sometimes the changes have failed and may have made path-finding worse, but they were rectified soon enough. There is active experimentation going on to find the best implementation for sure, or else I would have experienced the same performance in every revision.

[CyclonatorZ]

Well, it seems that I was confused as to what you guys meant by "trunk build," as it turns out that I've actually just been using the official release on the main page. That probably would explain why I didn't notice any improvements - but I am still curious as to why you haven't implemented the trunk pathfinding improvements in the main build.

[JDW]

- but I am still curious as to why you haven't implemented the trunk pathfinding improvements in the main build.

No one said anything about the trunk having a better path-finding implementation than 2.3. Trunk is still experimental, and sometimes worse than 2.3 (in some situations), but they're working on it.

EDIT: What I'm saying is that trunk is not necessarily better than 2.3. But they are trying to find ways to improve it by using trunk as a testbed.

[Safety0ff]

EDIT: What I'm saying is that trunk is not necessarily better than 2.3. But they are trying to find ways to improve it by using trunk as a testbed.

Not true, we prefer having trunk's path'ing in a bearable state and using the path'ing branch as a testbed.

[JDW]

I see, thank you for correcting me there.

[Zarel]

Not true, we prefer having trunk's path'ing in a bearable state and using the path'ing branch as a testbed.

Well, we only created the pathing branch recently. Before that, we made a bunch of pathing patches and then argued about which one was better, and it never went very far.