Batch Least Squares¶

Iterative batch least squares estimator for offline orbit determination.

BatchLeastSquares ¶

BatchLeastSquares(epoch: Any, initial_state: Any, initial_covariance: Any, propagation_config: Any, force_config: Any, measurement_models: Any, config: Any = None, params: Any = None, additional_dynamics: Any = None, control_input: Any = None)

Batch Least Squares estimator for orbit determination.

Processes all observations simultaneously through an iterative Gauss-Newton algorithm. Supports both normal equations and stacked observation matrix solver formulations.

Example

import brahe as bh
import numpy as np

bh.initialize_eop()

epoch = bh.Epoch(2024, 1, 1, 0, 0, 0.0)
state = np.array([6878e3, 0.0, 0.0, 0.0, 7612.0, 0.0])
p0 = np.diag([1e6, 1e6, 1e6, 1e2, 1e2, 1e2])

bls = bh.BatchLeastSquares(
    epoch, state, p0,
    propagation_config=bh.NumericalPropagationConfig.default(),
    force_config=bh.ForceModelConfig.two_body_gravity(),
    measurement_models=[bh.InertialPositionMeasurementModel(10.0)],
)

Initialize instance.

consider_covariance `method descriptor` ¶

consider_covariance() -> ndarray

Consider covariance contribution.

Returns:

Type	Description
`ndarray`	numpy.ndarray or None: Consider covariance matrix, or None if consider parameters are not configured.

converged `method descriptor` ¶

converged() -> bool

Whether the solver has converged.

Returns:

Name	Type	Description
`bool`	`bool`	True if converged.

current_covariance `method descriptor` ¶

current_covariance() -> ndarray

Get current total covariance (formal + consider contribution).

Returns:

Type	Description
`ndarray`	numpy.ndarray: Total covariance matrix (n x n).

current_epoch `method descriptor` ¶

current_epoch() -> Epoch

Get current epoch (reference epoch for the batch).

Returns:

Name	Type	Description
`Epoch`	`Epoch`	Current epoch.

current_state `method descriptor` ¶

current_state() -> ndarray

Get current state estimate at the reference epoch.

Returns:

Type	Description
`ndarray`	numpy.ndarray: Current state vector.

final_cost `method descriptor` ¶

final_cost() -> float

Final cost function value J.

Returns:

Name	Type	Description
`float`	`float`	Final cost.

formal_covariance `method descriptor` ¶

formal_covariance() -> ndarray

Formal covariance (solve-for partition only).

Returns:

Type	Description
`ndarray`	numpy.ndarray: Formal covariance matrix.

iteration_records `method descriptor` ¶

iteration_records() -> list[BLSIterationRecord]

Per-iteration diagnostic records.

Only populated when config.store_iteration_records is True.

Returns:

Type	Description
`list[BLSIterationRecord]`	list[BLSIterationRecord]: List of iteration records.

iterations_completed `method descriptor` ¶

iterations_completed() -> int

Number of Gauss-Newton iterations completed.

Returns:

Name	Type	Description
`int`	`int`	Number of iterations.

observation_residuals `method descriptor` ¶

observation_residuals() -> list[list[BLSObservationResidual]]

Per-observation residuals for each iteration.

Only populated when config.store_observation_residuals is True. Outer list is indexed by iteration, inner list by observation.

Returns:

Type	Description
`list[list[BLSObservationResidual]]`	list[list[BLSObservationResidual]]: Nested list of observation residuals.

solve `method descriptor` ¶

solve(observations: list[Observation]) -> Any

Solve the batch least squares problem.

Iteratively processes all observations to find the state that minimizes the weighted sum of squared residuals.

Parameters:

Name	Type	Description	Default
`observations`	`list[Observation]`	List of observations to process.	required

Example

bls.solve(observations)
print(f"Converged: {bls.converged()}")
print(f"Iterations: {bls.iterations_completed()}")

BLSConfig ¶

BLSConfig(solver_method: Any = 0, max_iterations: Any = 10, state_correction_threshold: Any = Ellipsis, cost_convergence_threshold: Any = None, consider_params: Any = None, store_iteration_records: Any = True, store_observation_residuals: Any = True)

Configuration for the Batch Least Squares estimator.

Example

import brahe as bh

config = bh.BLSConfig()  # All defaults
config = bh.BLSConfig(max_iterations=20, solver_method=bh.BLSSolverMethod.STACKED_OBSERVATION_MATRIX)

Initialize instance.

cost_convergence_threshold `property` ¶

cost_convergence_threshold: float

Cost convergence threshold.

Returns:

Type	Description
`float`	float or None: Threshold value.

max_iterations `property` ¶

max_iterations: int

Maximum number of iterations.

Returns:

Name	Type	Description
`int`	`int`	Maximum iterations.

solver_method `property` ¶

solver_method: int

Solver method (0 = NormalEquations, 1 = StackedObservationMatrix).

Returns:

Name	Type	Description
`int`	`int`	Solver method identifier.

state_correction_threshold `property` ¶

state_correction_threshold: float

State correction convergence threshold.

Returns:

Type	Description
`float`	float or None: Threshold value.

store_iteration_records `property` ¶

store_iteration_records: bool

Whether iteration records are stored.

Returns:

Name	Type	Description
`bool`	`bool`	True if records are stored.

store_observation_residuals `property` ¶

store_observation_residuals: bool

Whether observation residuals are stored.

Returns:

Name	Type	Description
`bool`	`bool`	True if residuals are stored.

default `staticmethod` ¶

default() -> BLSConfig

Create a default BLS configuration.

Returns:

Name	Type	Description
`BLSConfig`	`BLSConfig`	Default configuration (normal equations, 10 iterations, threshold 1e-8).

BLSSolverMethod ¶

BLSSolverMethod()

Solver formulation for Batch Least Squares.

Available methods

BLSSolverMethod.NORMAL_EQUATIONS (0): Accumulate information matrix, solve via Cholesky. Memory-efficient O(n^2).
BLSSolverMethod.STACKED_OBSERVATION_MATRIX (1): Build full H matrix, better numerical conditioning. Memory O(m*n).

Example

import brahe as bh
method = bh.BLSSolverMethod.NORMAL_EQUATIONS
config = bh.BLSConfig(solver_method=method)

Initialize instance.

NORMAL_EQUATIONS `class-attribute` ¶

NORMAL_EQUATIONS: Any = 0

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

int('0b100', base=0) 4

STACKED_OBSERVATION_MATRIX `class-attribute` ¶

STACKED_OBSERVATION_MATRIX: Any = 1

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating-point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

int('0b100', base=0) 4

ConsiderParameterConfig ¶

ConsiderParameterConfig(n_solve: Any, consider_covariance: Any)

Configuration for consider parameters in batch estimation.

Partitions the state into solve-for (first n_solve elements) and consider (remaining elements) parameters. The consider parameters are not estimated but their uncertainty is accounted for in the covariance.

Example

import brahe as bh
import numpy as np

# 6D state: first 6 are solve-for, next 1 is consider (e.g., Cd)
consider_cov = np.array([[0.01]])
config = bh.ConsiderParameterConfig(n_solve=6, consider_covariance=consider_cov)

Initialize instance.

consider_covariance `property` ¶

consider_covariance: ndarray

A priori covariance for consider parameters.

Returns:

Type	Description
`ndarray`	numpy.ndarray: Consider covariance matrix (n_c x n_c).

n_solve `property` ¶

n_solve: int

Number of solve-for parameters.

Returns:

Name	Type	Description
`int`	`int`	Number of solve-for parameters.

BLSIterationRecord ¶

BLSIterationRecord()

Record of a single BLS iteration.

Contains the state estimate, covariance, state correction, cost, and residual statistics at each Gauss-Newton iteration.

Initialize instance.

cost `property` ¶

cost: float

Cost function value J at this iteration.

Returns:

Name	Type	Description
`float`	`float`	Cost function value.

covariance `property` ¶

covariance: ndarray

Covariance at this iteration (formal, solve-for only).

Returns:

Type	Description
`ndarray`	numpy.ndarray: Covariance matrix.

epoch `property` ¶

epoch: Epoch

Reference epoch for this iteration.

Returns:

Name	Type	Description
`Epoch`	`Epoch`	Iteration epoch.

iteration `property` ¶

iteration: int

Iteration number (0-indexed).

Returns:

Name	Type	Description
`int`	`int`	Iteration number.

rms_postfit_residual `property` ¶

rms_postfit_residual: float

RMS of all post-fit residuals at this iteration.

Returns:

Name	Type	Description
`float`	`float`	Post-fit RMS.

rms_prefit_residual `property` ¶

rms_prefit_residual: float

RMS of all pre-fit residuals at this iteration.

Returns:

Name	Type	Description
`float`	`float`	Pre-fit RMS.

state `property` ¶

state: ndarray

State estimate at this iteration.

Returns:

Type	Description
`ndarray`	numpy.ndarray: State vector.

state_correction `property` ¶

state_correction: ndarray

State correction applied at this iteration.

Returns:

Type	Description
`ndarray`	numpy.ndarray: State correction vector delta_x.

state_correction_norm `property` ¶

state_correction_norm: float

Norm of the state correction ||delta_x||.

Returns:

Name	Type	Description
`float`	`float`	State correction norm.

BLSObservationResidual ¶

BLSObservationResidual()

Per-observation residual from a BLS iteration.

Contains pre-fit and post-fit residuals for a single observation.

Initialize instance.

epoch `property` ¶

epoch: Epoch

Epoch of the observation.

Returns:

Name	Type	Description
`Epoch`	`Epoch`	Observation epoch.

model_name `property` ¶

model_name: str

Name of the measurement model used.

Returns:

Name	Type	Description
`str`	`str`	Model name.

postfit_residual `property` ¶

postfit_residual: ndarray

Post-fit residual: y - h(x_{k+1}, t) after state correction.

Returns:

Type	Description
`ndarray`	numpy.ndarray: Post-fit residual vector.

prefit_residual `property` ¶

prefit_residual: ndarray

Pre-fit residual: y - h(x_k, t) before state correction.

Returns:

Type	Description
`ndarray`	numpy.ndarray: Pre-fit residual vector.

Batch Least Squares¶