Fit a boundary collar with affine hanging-control constraints against private PL observations. Each leaf is an axis-aligned UV rectangle, so clipping and local coordinates preserve the physical-area polynomial objective. Adaptive attempts split residual-marked leaves at their observed UV locations and close opposed interface intervals. Exhausted or incompatible attempts retain the previous field. Returned edge identities are coefficient relations, not source topology or G1 claims. Optional physical-gradient observations change the declared fitting objective. Their millimetre exchange length scales with physical units. Refinement retains the preceding candidate unless the combined objective improves without worsening its original physical-area position RMS. Every attempt reuses the initial coefficient enclosure; restricting a boundary cubic cannot silently shrink the feasible box. Point fallback supplies no gradient fit.
Fit a boundary collar with affine hanging-control constraints against private PL observations. Each leaf is an axis-aligned UV rectangle, so clipping and local coordinates preserve the physical-area polynomial objective. Adaptive attempts split residual-marked leaves at their observed UV locations and close opposed interface intervals. Exhausted or incompatible attempts retain the previous field. Returned edge identities are coefficient relations, not source topology or G1 claims. Optional physical-gradient observations change the declared fitting objective. Their millimetre exchange length scales with physical units. Refinement retains the preceding candidate unless the combined objective improves without worsening its original physical-area position RMS. Every attempt reuses the initial coefficient enclosure; restricting a boundary cubic cannot silently shrink the feasible box. Point fallback supplies no gradient fit.