Trajectory estimation¶

The estimation of the trajectory, and all auxilliary parameters, are performed through the TrajectoryEstimator class.

Behind the scenes, the estimator builds a Ceres-Solver optimization problem that is then solved.

class kontiki.TrajectoryEstimator¶

Class that performs trajectory estimation.

solve(progress=True, max_iterations=50)¶

Solve the estimation problem by minimizing the measurement residuals. Returns a Summary instance.

If progress is True, then a progress report is written to stdout while the optimization in progressing.

max_iterations specify the number of steps the optimizer is allowed to take.

add_measurement(m : Measurement)¶: Add a measurement to the optimization problem.

Warning

Make sure the measurement m is alive until solve() has been called, e.g. by putting it in a list.

add_callback(func, update_state=False)¶

Adds a callback function which is called after every iteration of the optimizer.

The callback function should accepts a single IterationSummary instance as argument, and return either None, or a CallbackReturnType value. None is the same as CallbackReturnType.Continue.

If your callbacks need to access the objects currently being optimized (e.g. the trajectory), you must set update_state=True.

Estimator helper classes¶

The following classes are wrappers around the Ceres-Solver classes of the same name.

class kontiki.CallbackReturnType¶

Return types for iteration callbacks

Members:

Abort : Abort the optimization

Continue : Continue the optimization

TerminateSuccessfully : Stop the optimization, marking it as success

class kontiki.Summary¶

BriefReport(self: kontiki._ceres.Summary) → str¶

FullReport(self: kontiki._ceres.Summary) → str¶

IsSolutionUsable(self: kontiki._ceres.Summary) → bool¶

final_cost¶

fixed_cost¶

initial_cost¶

inner_iteration_time_in_seconds¶

is_constrained¶

iterations¶

jacobian_evaluation_time_in_seconds¶

line_search_cost_evaluation_time_in_seconds¶

line_search_gradient_evaluation_time_in_seconds¶

line_search_polynomial_minimization_time_in_seconds¶

line_search_total_time_in_seconds¶

linear_solver_time_in_seconds¶

message¶

minimizer_time_in_seconds¶

num_effective_parameters¶

num_effective_parameters_reduced¶

num_inner_iteration_steps¶

num_linear_solver_threads_given¶

num_linear_solver_threads_used¶

num_parameter_blocks¶

num_parameter_blocks_reduced¶

num_parameters¶

num_parameters_reduced¶

num_residual_blocks¶

num_residual_blocks_reduced¶

num_residuals¶

num_residuals_reduced¶

num_successful_steps¶

num_threads_given¶

num_threads_used¶

num_unsuccessful_steps¶

postprocessor_time_in_seconds¶

preprocessor_time_in_seconds¶

residual_evaluation_time_in_seconds¶

termination_type¶

total_time_in_seconds¶

class kontiki.IterationSummary¶

cost¶

cost_change¶

cumulative_time_in_seconds¶

eta¶

gradient_max_norm¶

gradient_norm¶

iteration¶: Current iteration

iteration_time_in_seconds¶

line_search_function_evaluations¶

line_search_gradient_evaluations¶

line_search_iterations¶

linear_solver_iterations¶

relative_decrease¶

step_is_nonmonotonic¶

step_is_successful¶

step_is_valid¶: Step was numerically valid

step_norm¶

step_size¶

step_solver_time_in_seconds¶

trust_region_radius¶