Read on Twitter
more Monte Carlo Geometry - 3D vector field interpolated from surface vector field, which is defined by projecting a fixed direction onto surface normals. Tracing streamlines to visualize. There is no "solve" or precompute here, it's all pointwise, which is interesting... (1/3)
A 2D slice gives a better sense of what the vector field looks like. Computing a slice is pretty quick, looks like this vector field might be useful for moving things around objects. And since there is no precomputation the fixed 'target' vector can be arbitrarily changed...(2/3)
Pointwise evaluation + arbitrary direction means it's plausible to use for local *3D* pathfinding at runtime, to make an object follow me while going around *and over top of* scene geometry. Neat! Scene geometry is static here but it could be moving around too (3/5)
Makes me think this might be the the real power of this MC Geometry stuff. The paper was all about comparison to FEM but I think the most interesting usage is in situations where you have dynamic boundary conditions and just need a few samples (4/5).
In that kind of situation FEM would be completely implausible even if you had a domain mesh (which of course you wouldn't)! I would never even try to solve a global vector field to move a sphere around things in a game, but I can point-sample it! @keenanisalive @rohansawhney1
