nx_dyna_bind_skin.nx

nx_dyna_bind_skin.nx -- BIND THE XPBD SOFT-TISSUE CAGE TO SKIN VERTICES, AND MEASURE OUR OWN PLANT IN THE SUBJECT'S UNITS. WHY THIS EXISTS. `dyna_bind_skin` is OPEN on the charsim and graphics boards. DYNA already DECLARES per-bone soft-tissue contracts in the asset (nx_nxa_dyna: anchor, axis, k, c, travel clamp, influence radius, falloff, all derived FROM THE MESH) and nx_softtissue already SOLVES tissue as XPBD over a tetrahedral cage with per-layer compliance. Nothing joined them: the declared physics never reached a skin vertex, so the shipped path was still a shader band. WHY XPBD, NOT A MASS-SPRING RIG -- the model choice, justified against the alternatives, because the operator's bar is that the motion be best-of-breed and that previous attempts were awful: * XPBD (Macklin, Muller, Chentanez, Kim -- "XPBD: Position-Based Simulation of Compliant Constrained Dynamics", ACM SIGGRAPH Motion in Games 2016) integrates by POSITION PROJECTION, so it cannot detonate the way an explicit force solver does at stiff settings. Its compliance alpha is a genuine material parameter (inverse stiffness) and alpha_tilde = alpha/dt^2 makes effective stiffness STEP-SIZE INDEPENDENT. That property is precisely the cure for the buzzing/framerate-dependence failure mode, and plain PBD does not have it. * The estate ALREADY MEASURED why the previous attempt was awful. nx_softtissue's header records that the rigmesh viewer drove breasts with a 2-mass EMA spring, and that "an oscillator has no VOLUME, no LAYERS and no relationship to gravity DIRECTION" -- so it physically cannot reproduce supine flattening. A damped oscillator is a POINT; tissue is a VOLUME. Rejected on measured evidence, not taste. * Shape-matching lattice deformers (Muller 2005) are the other cheap production option: unconditionally stable, but they preserve no volume and their stiffness is a blend factor with no material meaning -- so their parameters CANNOT be sourced from tissue literature, which this estate requires of every number. Rejected on sourcing. * FEM production stacks (Chaos Flesh, Ziva, DMM) have better constitutive fidelity at far higher cost and none is reachable as a sovereign integer solver. XPBD is the tier that is both principled and ours. WHAT "BOUND TO SKIN VERTICES" MEANS HERE. The cage is a physics proxy; the skin mesh is the render surface. Binding is an EMBEDDING: a skin vertex inside a DYNA region takes its displacement from the cage, weighted by that region's OWN declared falloff over its OWN declared influence radius. A vertex outside every region is SKIN-LOCKED -- it moves with the rig and nothing else. That split IS the capability, and it is what the gate discriminates: soft moves, rigid does not. Both signals must be present at once or the tooth proves nothing. FRAMES DIFFER AND THAT HAS BURNED THIS ESTATE BEFORE -- SO IT IS STATED, NOT ASSUMED. NXA asset frame : +z is the STATURE axis, y<0 is the FRONT hemisphere (nx_nxa_dyna relies on this and says so; the floor gate and the shader agree).