LSB#
- class LSB(propagator: Propagator | MissionSequence, measurements: MeasurementSpec, settings: FilterSettings, traj_name: str | None = None, traj_dir: str | None = None, overwrite_traj: bool = True, verbose: bool = True)#
Bases:
FilterODWeighted Batch Least-Squares (LSB) estimator for spacecraft orbit determination.
The LSB accumulates all measurements simultaneously (“batch”) rather than processing them one by one as in a sequential filter. All epochs contribute to a single linear system that is solved once to obtain the state deviation at the reference epoch \(t_0\).
Normal equations
Define \(A_k = H_k\,\Phi_{k,0}\), the measurement Jacobian at epoch \(t_k\) mapped back to \(t_0\) via the STM \(\Phi_{k,0}\), and \(y_k\) the prefit residual. The weighted normal matrix and right-hand side are accumulated over all \(K\) epochs:
\[M = P_0^{-1} + \sum_{k=1}^{K} A_k^T R_k^{-1} A_k, \qquad b = \sum_{k=1}^{K} A_k^T R_k^{-1} y_k\]where \(P_0^{-1}\) is the a-priori information matrix. The solution and its covariance are:
\[\hat{x}_0 = M^{-1} b, \qquad \hat{P}_0 = M^{-1}.\]\(M\) is the information matrix \(\Lambda\) of the full batch, so the LSB solution is the maximum-likelihood estimate under Gaussian noise.
Consider parameters (Schmidt extension)
When consider parameters are present, the cross-term \(M_{xc}\) is accumulated alongside \(M_{xx}\) and the posterior covariance is inflated by the consider contribution:
\[\hat{P}_0 = M_{xx}^{-1} + S_{xc}\,P_{cc}\,S_{xc}^T\]where \(S_{xc} = -M_{xx}^{-1} M_{xc}\) is the consider sensitivity matrix (Tapley et al., Chapter 7).
Numerical considerations
The LSB solves the normal equations directly via matrix inversion, which squares the condition number of the problem. For ill-conditioned measurement geometries,
SRIFBprovides the same solution through an orthogonal QR decomposition that avoids explicit normal-matrix inversion and is numerically preferable.Process noise and stochastic parameters
Sequential process noise models such as SNC and DMC require intermediate time updates between measurement epochs and are therefore not supported in the purely batch formulation. Use
SRIForLKFwhen unmodelled accelerations must be compensated sequentially epoch by epoch.A batch analogue can still be constructed through so-called “stochastics” or pseudo-epoch states, where empirical accelerations are represented as estimated solve-for parameters within the batch state vector. In this approach, first-order Gauss–Markov or piecewise-constant stochastic accelerations are parameterized over predefined intervals and estimated simultaneously with the orbital state.
Each stochastic parameter contributes additional columns to the design matrix
\[A_k = H_k \Phi_{k,0},\]allowing the batch filter to recover time-varying empirical accelerations without performing intermediate sequential covariance updates. This formulation is commonly referred to as a pseudo-epoch-state filter and is preferred when the stochastic signal can be adequately represented by deterministic parameters over the data arc.
- Parameters:
- propagatorPropagator or MissionSequence
Defines the reference trajectory and STMs.
- measurementsMeasurementSpec
Measurement specification object.
- settingsFilterSettings
Filter configuration.
settings.process_noisemust beNone.- traj_namestr, optional
Name for the trajectory BSP file.
- traj_dirstr, optional
Directory for the trajectory BSP file.
- overwrite_trajbool, optional
If
True, overwrite any existing trajectory file.- verbosebool, optional
If
False, silence the printout of the propagation run at construction. Defaults toTrue.
- Raises:
See also
scarabaeus.SRIFBNumerically superior batch estimator in the square-root information domain; produces the same solution with better conditioning.
scarabaeus.LKFSequential covariance-form filter; supports process noise.
scarabaeus.FilterODBase class providing trajectory and measurement handling.
Notes
The LSB inverts the normal matrix directly, squaring the condition number. For ill-conditioned measurement geometries prefer
SRIFB, which produces the same solution via an orthogonal QR decomposition. Process noise (SNC/DMC) is not supported; useLKForSRIFwhen unmodelled accelerations must be compensated.References
[1]Tapley, B. D., Schutz, B. E., & Born, G. H. (2004). Statistical Orbit Determination. Elsevier Academic Press. ISBN 978-0-12-683630-1.