leap_c.diff_mpc.utils.create_solver

Utilities for creating an AcadosOcpBatchSolver from an AcadosOcp object.

Functions

create_batch_solver(→ acados_template.AcadosOcpBatchSolver)

Create an AcadosOcpBatchSolver from an AcadosOcp object.

create_forward_backward_batch_solvers(...)

Create a batch solver for solving the MPC problems (forward solver).

make_ocp_sensitivity_compatible(sensitivity_ocp)

Make the given ocp compatible with sensitivity computation.

Module Contents

leap_c.diff_mpc.utils.create_solver.create_batch_solver(ocp: acados_template.AcadosOcp, export_directory: str | pathlib.Path | None = None, discount_factor: float | None = None, n_batch_init: int = 256, num_threads: int = 4, verbose: bool = True) → acados_template.AcadosOcpBatchSolver[source]

Create an AcadosOcpBatchSolver from an AcadosOcp object.

Parameters:

  • ocp – Acados optimal control problem formulation.

  • export_directory – Directory to export the generated code. If None, a temporary directory is created and the directory is cleaned afterwards.

  • discount_factor – Discount factor. If None, acados default cost scaling is used, i.e. dt for intermediate stages, 1 for terminal stage.

  • n_batch_init – Initially supported batch size of the batch OCP solver. Using larger batches will trigger a delay for the creation of more solvers.

  • num_threads – Number of threads used in the batch solver.

  • verbose – Whether to print the code generation output.

leap_c.diff_mpc.utils.create_solver.create_forward_backward_batch_solvers(ocp: acados_template.AcadosOcp, sensitivity_ocp: acados_template.AcadosOcp | None = None, export_directory: str | pathlib.Path | None = None, discount_factor: float | None = None, n_batch_init: int = 256, num_threads: int = 4, verbose: bool = True) → tuple[acados_template.AcadosOcpBatchSolver, acados_template.AcadosOcpBatchSolver][source]

Create a batch solver for solving the MPC problems (forward solver).

If this solver is suitable for computing sensitivities, it will also be returned as backward solver (the solver for computing sensitivities). Otherwise, a second batch solver will be created, which is suitable for computing sensitivities.

Parameters:

  • ocp – Acados optimal control problem formulation for the forward solver.

  • sensitivity_ocp – Acados optimal control problem formulation for the backward solver. If None, this will be derived from the given ocp.

  • export_directory – Directory to export generated code. If none, a unique temporary directory is created.

  • discount_factor – Discount factor for the solver. If not provided, acados default weighting is used (i.e., 1/N_horizon for intermediate stages, 1 for terminal stage).

  • n_batch_init – Initially supported batch size of the batch OCP solver. Using larger batches will trigger a delay for the creation of more solvers.

  • num_threads – Number of threads used in the batch solver.

  • verbose – Whether to print the code generation output.

leap_c.diff_mpc.utils.create_solver.make_ocp_sensitivity_compatible(sensitivity_ocp: acados_template.AcadosOcp)[source]

Make the given ocp compatible with sensitivity computation.