DoN's 3D-Pathfinding / Flying-AI system (with full source!)

[Ryan_Darcey]

Man, thanks for sticking through this with me, @. I feel like we’re getting close to figuring this out!..and then I can start trying to figure out how I’m ever going to be able to reward you for incredible generosity

@anonymous_user_e71e0d8aSZ;573632

One thing that’s not apparent from your screenshots is the location of the origin and destination volumes. Assuming DrawDebugVolumes is turned on the origin should show up as a white box and the destination as a green box. I’m curious to know where they’re placed when the query fails, especially the white origin box.

Here are the origins for the 2 pawns with the debug display you were interested in. The pawn sitting on the left obviously failed and the one that was to the right made it to the goal. As you can see, just empty space between them.

Google Drive: Sign-in

Also, here are the voxels surrounding the origins:

Google Drive: Sign-in

Here’s the goal. You can see the pawn that had a successful query is chillin’ and the green voxel is slightly sticking out.

Google Drive: Sign-in

Also, voxels around goal:

Google Drive: Sign-in

@anonymous_user_e71e0d8aSZ;573632

Here’s a theory - maybe your tweak magnitudes are somehow bumping the origin volume somewhere unexpected so even though we see an apparent sea of green voxels from actual origin to actual destination, the effective origin (created by applying the tweak magnitude on your origin) may be somewhere else.

I tried reducing my tweak magnitudes all down to 100, still same result.

@anonymous_user_e71e0d8aSZ;573632

At this stage, a screenshot capturing only the origin, the origin volume and the first 2-3 set of voxel neighbors alone will really help us understand what’s going on. You can achieve this by using the existing SolverIterationCount counter in TickNavigationSolver to limit all your drawdebug activity if the SolverIterationCount has exceeded say 5 (for example). That should provide a simple visualization of just origin and the initial set(s) of neighbors with which we should finally be able to solve this puzzle!

Here are the results. Dude on the left just doesn’t want to leave his home!

Google Drive: Sign-in

@anonymous_user_e71e0d8aSZ;573632

Edit: More ideas - Turn collision off on your pawn and then on your entire landscape and see if either makes a difference. Anything to help narrow down which collision entity is involved here may also help us narrow down the issue.

Turning off collision on the landscape results in both pawns succeeding with a direct path sweep. Turning off collision on the pawns has no impact on the result.

OK…

…with this last test I just did I am 99% convinced there’s something about the distance between the origin and goal that’s the culprit here. I slightly nudged the goal towards both the origins, and now they both succeed!

Google Drive: Sign-in

This seems to tell me that there’s nothing inherently wrong with the origins or goals. We have examples where both pawns are able to leave their origin to pursue the goal, and the goal is always able to be reach by at least one of the pawns. I’ve been looking but haven’t found anything…are there any other numbers that would be sensitive to distances other than tweak magnitudes? Like, is the scale of my map breaking some assumption in code about the distances the pawns might want to travel?