|
Drake
|
|
Drake
|
#include <drake/geometry/optimization/graph_of_convex_sets.h>
Public Member Functions | |
| template<typename Archive > | |
| void | Serialize (Archive *a) |
| Passes this object to an Archive. More... | |
Public Attributes | |
| std::optional< bool > | convex_relaxation {std::nullopt} |
| Flag to solve the relaxed version of the problem. More... | |
| std::optional< int > | max_rounded_paths {std::nullopt} |
| Maximum number of distinct paths to compare during random rounding; only the lowest cost path is returned. More... | |
| std::optional< bool > | preprocessing {std::nullopt} |
| Performs a preprocessing step to remove edges that cannot lie on the path from source to target. More... | |
| int | max_rounding_trials {100} |
| Maximum number of trials to find a novel path during random rounding. More... | |
| double | flow_tolerance {1e-5} |
| Tolerance for ignoring flow along a given edge during random rounding. More... | |
| int | rounding_seed {0} |
| Random seed to use for random rounding. More... | |
| const solvers::SolverInterface * | solver {nullptr} |
| Optimizer to be used to solve the MIP, the relaxation of the shortest path optimization problem and the convex restriction if no restriction_solver is provided. More... | |
| const solvers::SolverInterface * | restriction_solver {nullptr} |
| Optimizer to be used in SolveConvexRestriction(), which is also called during the rounding stage of SolveShortestPath() given the relaxation. More... | |
| const solvers::SolverInterface * | preprocessing_solver {nullptr} |
Optimizer to be used in the preprocessing stage of GCS, which is performed when SolveShortestPath is called when the preprocessing setting has been set to true. More... | |
| solvers::SolverOptions | solver_options {} |
| Options passed to the solver when solving the generated problem. More... | |
| std::optional< solvers::SolverOptions > | restriction_solver_options |
| Optional solver options to be used in SolveConvexRestriction(), which is also used during the rounding stage of SolveShortestPath() given the relaxation. More... | |
| std::optional< solvers::SolverOptions > | preprocessing_solver_options |
| Optional solver options to be used by preprocessing_solver in the preprocessing stage of GCS, which is used in SolveShortestPath. More... | |
| Parallelism | parallelism {Parallelism::Max()} |
| Some steps in GCS can be parallelized. More... | |
| void Serialize | ( | Archive * | a | ) |
Passes this object to an Archive.
Refer to YAML Serialization for background. Note: This only serializes options that are YAML built-in types.
| std::optional<bool> convex_relaxation {std::nullopt} |
Flag to solve the relaxed version of the problem.
As discussed in the paper, we know that this relaxation cannot solve the original NP-hard problem for all instances, but there are also many instances for which the convex relaxation is tight. If convex_relaxation=nullopt, then each GCS method is free to choose an appropriate default.
| double flow_tolerance {1e-5} |
Tolerance for ignoring flow along a given edge during random rounding.
If convex_relaxation is false or max_rounded_paths is less than or equal to zero, this option is ignored.
| std::optional<int> max_rounded_paths {std::nullopt} |
Maximum number of distinct paths to compare during random rounding; only the lowest cost path is returned.
If convex_relaxation is false or this is less than or equal to zero, rounding is not performed. If max_rounded_paths=nullopt, then each GCS method is free to choose an appropriate default.
| int max_rounding_trials {100} |
Maximum number of trials to find a novel path during random rounding.
If convex_relaxation is false or max_rounded_paths is less than or equal to zero, this option is ignored.
| Parallelism parallelism {Parallelism::Max()} |
Some steps in GCS can be parallelized.
This is the maximum number of threads used in all places in the algorithm.
| std::optional<bool> preprocessing {std::nullopt} |
Performs a preprocessing step to remove edges that cannot lie on the path from source to target.
In most cases, preprocessing causes a net reduction in computation by reducing the size of the optimization solved. Note that this preprocessing is not exact. There may be edges that cannot lie on the path from source to target that this does not detect. If preprocessing=nullopt, then each GCS method is free to choose an appropriate default.
| const solvers::SolverInterface* preprocessing_solver {nullptr} |
Optimizer to be used in the preprocessing stage of GCS, which is performed when SolveShortestPath is called when the preprocessing setting has been set to true.
If not set, the interface at .solver will be used, if provided, otherwise the best solver for the given problem is selected. Note that if the solver cannot handle the type of optimization problem generated, then calling the solvers::SolverInterface::Solve() method will throw.
| std::optional<solvers::SolverOptions> preprocessing_solver_options |
Optional solver options to be used by preprocessing_solver in the preprocessing stage of GCS, which is used in SolveShortestPath.
If preprocessing_solver is set but this parameter is not then solver_options is used. For instance, one might want to print solver logs for the main optimization, but not from the many smaller preprocessing optimizations.
| const solvers::SolverInterface* restriction_solver {nullptr} |
Optimizer to be used in SolveConvexRestriction(), which is also called during the rounding stage of SolveShortestPath() given the relaxation.
If not set, the interface at .solver will be used, if provided, otherwise the best solver for the given problem is selected. Note that if the solver cannot handle the type of optimization problem generated, then calling the solvers::SolverInterface::Solve() method will throw.
| std::optional<solvers::SolverOptions> restriction_solver_options |
Optional solver options to be used in SolveConvexRestriction(), which is also used during the rounding stage of SolveShortestPath() given the relaxation.
If not set, solver_options is used. For instance, one might want to set tighter (i.e., lower) tolerances for running the relaxed problem and looser (i.e., higher) tolerances for final solves during rounding.
| int rounding_seed {0} |
Random seed to use for random rounding.
If convex_relaxation is false or max_rounded_paths is less than or equal to zero, this option is ignored.
| const solvers::SolverInterface* solver {nullptr} |
Optimizer to be used to solve the MIP, the relaxation of the shortest path optimization problem and the convex restriction if no restriction_solver is provided.
If not set, the best solver for the given problem is selected. Note that if the solver cannot handle the type of optimization problem generated, the calling solvers::SolverInterface::Solve() method will throw.
| solvers::SolverOptions solver_options {} |
Options passed to the solver when solving the generated problem.