Receiver error and accumulated IMU noise are separate contributions

All 28 focused software checks pass. The new component calculates how a declared linear inertial-noise model accumulates into position uncertainty, including correlation between earlier and later observations. It can combine that covariance with the existing receiver-error family.

The tests use a separate matrix-exponential and numerical-integration reference, irregular sampling, cross-axis noise, a four-minute horizon and interval subdivision. They also check that taking a later subset does not reset the noise-integration origin. Invalid inputs are rejected rather than repaired.

No aircraft observations, instrument noise estimates or model fits enter these checks. Acceleration and gyro noise are independent white Gaussian drives in this supplied linear approximation; receiver noise is assumed independent of them. Actual recordings may violate those assumptions. Quantization, changing aircraft orientation, nonlinear propagation and unknown onboard processing need their own assessment. No numerical default is inferred from survey-grade status.

This closes a missing mathematical component, not V01's adequate-error-law requirement. It does not turn the existing receiver-only uncertainty into an established full-aircraft error model or permit new prediction fits.

Assumptions and finite scope · Mathematics · Overall error assumptions.