Struct Correlation 

pub struct Correlation {
    pub gnss_position: Option<Seconds>,
    pub gnss_height: Option<Seconds>,
    pub gnss_velocity: Option<Seconds>,
    pub baro_altitude: Option<Seconds>,
    pub mag_heading: Option<Seconds>,
    pub gnss_heading: Option<Seconds>,
    pub course: Option<Seconds>,
}

How long each source’s measurement error persists: the time constant τ of equation (24′), per source. None fuses that source as white, which is (24) as written.

Equation (24) treats each measurement’s error as independent of the last, and almost no source in the corpus is: fused as white, the lag-1 autocorrelation of its innovations (the acf1_ keys of the replay summary line) is positive on nearly every real log. A filter that fuses such measurements as white averages down an error it cannot observe, and its covariance claims the averaging worked. (24′) fuses each at the variance that makes a run of them carry what they actually carry, which is less the shorter the interval is against τ.

Inflating R rather than flooring P, and in the gain rather than the gate, is measured; the decision records what each alternative read.

Per source, like Recovery and Gates, because a source is what the gate judges and what (24′) times. Configured rather than derived, as ImuNoise is: it is a property of the sensor, and the filter cannot measure it in flight without retuning itself. The replay harness’s --derive reads it offline, τ = −T / ln ρ at a log’s sample interval T and its lag-one innovation autocorrelation ρ, and that reading is a lower bound: an innovation is whiter than the error behind it, because the filter follows part of that error. It falls well short of a simulated sensor’s own τ, so --derive prints it only above the default (measured); an estimator the filter does not bias is #195.

Fields

gnss_position: Option<Seconds>

GNSS horizontal position.

gnss_height: Option<Seconds>

GNSS height.

gnss_velocity: Option<Seconds>

GNSS velocity, all three axes.

baro_altitude: Option<Seconds>

Barometric altitude. The reference’s drift is (30′)’s, not this: this is the noise about it.

mag_heading: Option<Seconds>

Magnetic heading, with the leveling variance of (36′) it carries.

gnss_heading: Option<Seconds>

Dual-antenna GNSS heading.

course: Option<Seconds>

Course constraint: the sideslip the constraint allows, which persists as long as the wind and the vehicle’s trim do.

Implementations

impl Correlation

pub const WHITE: Self

Every measurement independent of the last: equation (24) with nothing added.

Trait Implementations

impl Clone for Correlation

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for Correlation

impl Debug for Correlation

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Correlation

fn default() -> Self

The corpus’s: τ = −T / ln ρ from each real log’s acf1_ for the source, read with every measurement fused as white, at that log’s own sample interval, and the median over the real logs whose autocorrelation is positive. GNSS heading and the course constraint rest on one log each. The table, and the logs left out, are the decision’s.

impl PartialEq for Correlation

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for Correlation