Expand description
Orbit-determination kernels.
The one decision that governs the whole layer: least squares by orthogonal factorisation, never
by the normal equations. Forming HᵀWH squares the condition number, and orbit-determination
Jacobians are ill-conditioned enough that the difference is between eight correct digits and
one. Everything here is built on the square-root information filter Srif, and the nonlinear
solver GaussNewton wraps it.
This is the linear-algebra core. Wiring it to a measurement model (range, range-rate, angles) and
the state transition matrix — which maps an observation’s partials back to the epoch state — is
the next layer; the STM it needs is built (spacelab_propagation::variational).
Re-exports§
pub use conjunction::ConjunctionError;pub use conjunction::ConjunctionGeometry;pub use conjunction::EncounterPlane;pub use conjunction_domain::CurvilinearEncounter;pub use conjunction_domain::DomainError;pub use conjunction_domain::RectilinearEncounter;pub use conjunction_domain::RelativeSample;pub use conjunction_domain::UsageViolations;pub use conjunction_domain::usage_violations;pub use consider::consider_covariance;pub use consider::consider_sensitivity;pub use consistency::chi_square_mean_interval;pub use consistency::chi_square_scale_factor;pub use consistency::is_consistent;pub use consistency::normalized_squared;pub use consistency::scalar_nis;pub use converter::ConverterError;pub use converter::ConverterFit;pub use converter::DifferentiablePropagator;pub use converter::PropagatorBuilder;pub use converter::PropagatorConverter;pub use converter::SampledPropagator;pub use converter::TrajectoryResiduals;pub use converter::TrajectorySample;pub use converter::residual_statistics;pub use ensemble::EnsembleError;pub use ensemble::EnsembleKalman;pub use ephemeris::Component;pub use ephemeris::StateComponent;pub use fixed_lag::FixedLagSmoother;pub use gauss_newton::GaussNewton;pub use gauss_newton::GaussNewtonReport;pub use gauss_newton::ResidualModel;pub use hinf::HInfinityError;pub use hinf::HInfinityFilter;pub use iekf::InvariantEkf;pub use imm::ImmError;pub use imm::InteractingMultipleModel;pub use kalman::SquareRootKalman;pub use lie::UnitQuaternion;pub use light_time::SPEED_OF_LIGHT;pub use light_time::Transit;pub use light_time::down_leg;pub use lincov::LinearCovariance;pub use measurement::Station;pub use measurement::Topocentric;pub use measurement::azimuth;pub use measurement::declination;pub use measurement::elevation;pub use measurement::range;pub use measurement::range_rate;pub use measurement::right_ascension;pub use mekf::MultiplicativeEkf;pub use mixture::GaussianComponent;pub use mixture::GaussianMixture;pub use modifier::AtmosphericDelay;pub use modifier::Azimuth;pub use modifier::Bias;pub use modifier::Declination;pub use modifier::Elevation;pub use modifier::EstimatedBias;pub use modifier::EstimatedDrift;pub use modifier::EstimatedShapiroDelay;pub use modifier::Modified;pub use modifier::Modifier;pub use modifier::ModifierError;pub use modifier::Observable;pub use modifier::Partials;pub use modifier::Range;pub use modifier::RangeRate;pub use modifier::RightAscension;pub use modifier::ShapiroDelay;pub use modifier::SlantDelay;pub use modifier::SolarCoronaDelay;pub use modifier::partials;pub use modifier::shapiro_delay;pub use modifier::shapiro_delay_ppn;pub use multistatic::BistaticGeometry;pub use multistatic::BistaticRange;pub use multistatic::BistaticRangeRate;pub use multistatic::Fdoa;pub use multistatic::InterSatelliteRange;pub use multistatic::InterSatelliteRangeRate;pub use multistatic::MultistaticError;pub use multistatic::PairObservable;pub use multistatic::PairPartials;pub use multistatic::Tdoa;pub use multistatic::TdoaGeometry;pub use multistatic::Terminal;pub use multistatic::bistatic_range;pub use multistatic::bistatic_range_rate;pub use multistatic::fdoa;pub use multistatic::inter_satellite_range;pub use multistatic::inter_satellite_range_rate;pub use multistatic::pair_partials;pub use multistatic::tdoa;pub use observed::Observed;pub use observed::Signal;pub use observed::partials_at_emission;pub use observed::reduce;pub use particle::ParticleError;pub use particle::ParticleFilter;pub use particle::Resampling;pub use process_noise::DiscreteStep;pub use process_noise::ProcessNoiseError;pub use process_noise::dynamic_model_compensation;pub use process_noise::matrix_exp;pub use process_noise::process_noise_sqrt;pub use process_noise::rotate_position_velocity;pub use process_noise::state_noise_compensation;pub use process_noise::van_loan;pub use radiometric::AKSIM_PAVLOV_2022;pub use radiometric::Band;pub use radiometric::CASSINI_KA_TURNAROUND_RATIO;pub use radiometric::CoronaSeries;pub use radiometric::CoronaTerm;pub use radiometric::DopplerLink;pub use radiometric::DopplerReference;pub use radiometric::DopplerSignature;pub use radiometric::PLASMA_DELAY_COEFFICIENT;pub use radiometric::PlasmaSignature;pub use radiometric::Quasar;pub use radiometric::RadiometricError;pub use radiometric::RampSegment;pub use radiometric::RampTable;pub use radiometric::SOLAR_RADIUS;pub use radiometric::averaged_doppler;pub use radiometric::delta_dor;pub use radiometric::measured_frequency;pub use radiometric::range_modulus;pub use radiometric::sequential_range;pub use radiometric::solar_corona_delay;pub use richardson::richardson_extrapolate;pub use richardson::richardson2;pub use robust::RobustOutcome;pub use robust::RobustPolicy;pub use robust::RobustTally;pub use robust::chi_square_gate;pub use robust::robust_update;pub use rts::RtsError;pub use rts::RtsSmoother;pub use rts::SmoothingStep;pub use sampling::Rng;pub use screening::CloseApproach;pub use screening::MAX_SAMPLES;pub use screening::ScreeningError;pub use screening::closest_approaches;pub use screening::radially_separated;pub use se3::Se3;pub use se23::ImuNoise;pub use se23::Se23;pub use srif::Srif;pub use srif::SrifError;pub use srukf::SquareRootUnscentedKalman;pub use unscented::UnscentedParams;pub use unscented::unscented_transform;
Modules§
- conjunction
- Conjunction assessment: the probability of collision
Pcfor a short-term encounter, from a relative state and a combined covariance at the time of closest approach. - conjunction_
domain - Whether
conjunction’s 2-D model applies to the conjunction it is about to be handed — the four checks that decide, and the correction factor that comes out of them. - consider
- Consider covariance (Schmidt 1966; Tapley, Schutz & Born, Statistical Orbit Determination, §4.14) — the honest covariance of an estimate that deliberately does not solve for some uncertain parameters.
- consistency
- Filter consistency diagnostics — NEES, NIS, and the χ²-per-degree-of-freedom covariance-realism test (Bar-Shalom, Estimation with Applications to Tracking and Navigation, §5.4).
- converter
- Fitting a propagator’s own parameter vector to a sampled reference trajectory.
- ensemble
- The ensemble Kalman filter: the covariance is carried by a sample, and never formed.
- ephemeris
- Position and PV as measurements: fitting an orbit to a state vector somebody else published.
- fixed_
lag - Fixed-lag smoothing: most of what the backward pass buys, at a bounded latency.
- gauss_
newton - Nonlinear least squares by Gauss–Newton with Levenberg–Marquardt damping, on top of the SRIF.
- hinf
- The H∞ (minimax) filter: a state estimator that assumes nothing about the noise except that it has finite energy, and bounds the worst case rather than the average.
- iekf
- Invariant Extended Kalman Filter (IEKF) for attitude on
SO(3)(Barrau & Bonnabel 2017). - imm
- The interacting-multiple-model filter: several dynamics hypotheses run in parallel, mixed by a Markov transition matrix, with the posterior model probability read out as the answer.
- kalman
- The sequential square-root Kalman filter: covariance carried as its square root, updated by orthogonal array transformations, never formed and never symmetrised.
- lie
- Minimal Lie-group primitives for on-manifold estimation: unit quaternions and the group
SO(3). - light_
time - Which epoch a measurement’s geometry belongs to.
- lincov
- Linear covariance analysis: what a navigation filter will actually achieve, as opposed to what it will report, on a scenario that has not flown yet.
- measurement
- Measurement models, written once over any
Scalarso their partials come from automatic differentiation — exact, and from the same source that computes the predicted value. - mekf
- Multiplicative Extended Kalman Filter (MEKF) for spacecraft attitude determination.
- mixture
- A Gaussian-mixture state estimate: a weighted set of Gaussian components, each carried in square-root form and each unscented-propagated, for uncertainty that a single Gaussian cannot hold.
- modifier
- Measurement corrections as a composable layer: one correction is a decorator over an observable, and a decorated observable is itself an observable.
- multistatic
- Observables whose signal path touches more than one terminal: bistatic range and range rate, TDOA, FDOA, and the inter-satellite pair.
- observed
- Which epoch an observable’s geometry is evaluated at, and what that does to its partials.
- particle
- The bootstrap particle filter: a weighted sample, with no assumption that the posterior is Gaussian in any coordinate system.
- process_
noise - Discrete process noise: turning the continuous spectral density physics gives you into the
Qevery filter in this crate demands. - radiometric
- Deep-space radiometric observables: DSN sequential range in range units, n-way averaged Doppler and the measured frequency that complements it, ΔDOR against a quasar, and the solar-corona delay all of them carry.
- richardson
- Richardson extrapolation as sequence acceleration and estimator debiasing.
- robust
- Outlier-robust measurement updates: χ² gating, Huber M-estimation, and a Student-t variational update.
- rts
- The Rauch–Tung–Striebel smoother: the backward pass the filter does not have.
- sampling
- Deterministic pseudo-random sampling for the Monte-Carlo estimators.
- screening
- Finding the conjunctions that
conjunctionthen puts a probability on. - se3
- The rigid-motion Lie group
SE(3)and its algebrase(3), for pose/extended-pose estimation. - se23
- The extended-pose group
SE₂(3)and a bias-augmented invariant navigation filter on it. - srif
- The square-root information filter: least squares by orthogonal triangularisation, never by the normal equations.
- srukf
- The square-root unscented Kalman filter (Van der Merwe & Wan 2001): a sigma-point filter that
carries the Cholesky factor
Sof the covariance, neverP, and never re-forms it. - unscented
- The scaled unscented (sigma-point) transform: a small, deterministic set of samples that carries a Gaussian through an arbitrary nonlinear map and recovers the pushed-forward mean and covariance.