Timucin Ozger

Hi everyone, I’m running into a weird behavior with an crowd agent i am trying to rig.
I’m swapping between two geometry inputs into a Switch node before my Agent/Clip setup

Input 0 (USD): Asset imported via USD (red network box). (This is the properly assetized geometry so its preferred)
Input 1 (BGEO): The same asset cached out as a BGEO.
When I switch to the USD input, the agentclip test node causes the agent to snap back to its A-pose (no deformation). When I use the BGEO input, the clip deforms perfectly.

Point counts, attribute names, and values appear identical between both inputs.
Both files share the same source (the USD was published from the BGEO).

Any idea why the USD-sourced geo would fail to deform despite having matching attributes? Is there a hidden USD intrinsic or metadata requirement for the Agent definition that I'm missing?

Yep, that must be it, i was thinking to solve this from wrong direction as how do i affect the next guy rather than should be infected, let me give that a spin

Hello,
I am trying to solve a problem, but kind of stuck

Basically I am trying to do a propagation, but only based on a vector available per point on the attribute itself, so if the underlying point says go right, spread should be only to the right, and next frame we query again, check the vector and only spread along that direction.

So I played around with points in a solver sop, pcopen check the nearest attribute on the surface @P+=v@dir; and points move as expected.

But how do I do that for an attribute on the geo itself. Well I mean not the attribute but the value of the next point that should inherit the value on the neighboring point based on the direction vector. Almost like a 2d gas solver I suppose. But the surface is not a plane, and UVs are not very straight forward to work on UV space on a grid.
Open to ideas, Thanks