Module estimation 

Estimator ports: process model and observation.

An estimator fuses observations into a NavigationState without knowing their source. A filter with update_gnss, update_gyro, update_log methods must change for every new sensor and cannot be tested without faking one; instead each observation describes itself — measurement, prediction, Jacobian, noise — through the Observation port. The ProcessModel port describes motion between observations.

Neither port exposes the state layout: an observation names StateComponent::Heading and the crate maps names to columns. Standard observations and process models live in kinavis; an adapter with an unusual sensor implements Observation without touching the estimator.

Structs

GatingPolicy

Innovation gating policy.

JacobianRow

One Jacobian row: derivatives of one element with respect to each state component.

ObservationJacobian

Observation Jacobian: one row per element, in ObservationVector order.

ObservationNoise

Observation noise covariance R.

ObservationVector

Observation elements, measured or predicted.

ProcessNoise

Process noise Q: unmodelled state change over a step.

StateJacobian

Process Jacobian F = ∂f/∂x over a step.

Constants

MAX_OBSERVATION_DIM

Maximum observation dimension.

Traits

Observation

Observation from the estimator’s view: z, h(x), H, R, no source details.

ProcessModel

State propagation between observations; injected into the estimator.