navlie.batch.estimator.BatchEstimator

class navlie.batch.estimator.BatchEstimator(solver_type: str = 'GN', max_iters: int = 100, step_tol: float = 1e-07, ftol: float | None = None, gradient_tol: float | None = None, tau: float = 1e-11, verbose: bool = True)

Bases: object

Main class for the batch estimator.

Instantiate BatchEstiamtor.

Parameters:

• solver (str, optional) – Solver type, either “GN” or “LM”, by default “GN”.

• max_iters (int, optional) – Maximum number of optimization iterations, by default 100.

• step_tol (float, optional) –

  Convergence step tolerance, by default 1e-7. The solver exits when

  \[||\Delta x||_2 < \text{step_tol}\]

  where \(\Delta x\) is the change in the state estimate for successive steps.

• ftol (float, optional) –

  Convergence relative cost decrease tolerance, by default None (not used). The solver exits when

  \[|\Delta C /C| < \text{ftol}\]

  where \(\Delta C\) is change in the cost function for successive accepted steps.

• gradient_tol (float, optional) –

  Convergence gradient infinity norm tolerance, by default None (not used). The solver exits when

  \[\max_i |\nabla J|_i = \max_i |\mathbf{e}^T \mathbf{H}|_i < \text{gradient_tol}\]

• tau (float, optional) – tau parameter in LM, by default 1e-11.

• verbose (bool, optional) – Print convergence during runtime, by default True.

solve(x0: State, P0: ndarray, input_data: List[Input], meas_data: List[Measurement], process_model: ProcessModel, return_opt_results: bool = False) → List[StateWithCovariance]

Creates and solves a batch problem.

The input data is used to propagate the initial state x0 forward in time using the process model, to generate an initial estimate of the state at estimate timestep.

The batch problem created involves a PriorResidual, a ProcessResidual for each input used to connect subsequent states through the process model, and a MeasurementResiduals for each measurement.

Parameters:

• x0 (State) – x0: Initial state.

• P0 (np.ndarray) – Initial covariance

• input_data (List[Input]) – List of input data.

• meas_data (List[Measurement]) – List of measurements.

• process_model (ProcessModel) – Process model used to propagate the initial estimate and form ProcessResiduals.

• return_opt_results (bool, optional) – Flag to optionally return the results dictionary from the batch problem, by default False

Returns:

List of estimates with covariance.

Return type:

List[StateWithCovariance]

previous

navlie.batch.estimator

next

navlie.batch.gaussian_mixtures