I am using vellum to simulate clothing on crowd agents. I am simulating different combinations of clothing, and I am using the Agent Vellum Unpack node to unpack my vellum geometry, which is going quite well.

However, because some of the different combinations of clothing, some are penetrating each other. Due to this, the interleaved_detangle and final_detangle nodes inside the vellum solver are tripling the simulation time trying to resolve some over the overlap and inner penetrations. This is mostly do to both clothing layers overlapping with each other, along with itself. The meshes are not that bad, I used a remesh node to get consistent topology, and there are not many egregious areas, but there must be enough.

I have tried bypassing the interleaved_detangle node, which helps, but I was wondering if I could prepare my geometry better before I simulate it. Not necessarily the topology or point positions, but more in terms of attributes. Is there a way I could identify the problematic points the detangle nodes will look to detangle before the sim so I can take before the sim? Can I assign those points some attributes? This is for crowds so the accuracy does not have to be perfect, so I could exclude those points from collisions if that meant speeding up the simulation.