Deprecated SolveIVP transcription. `simulate` method now uses the ExplicitShooting transcription without derivatives. #898

robfalck · 2023-01-31T15:32:45Z

Summary

SolveIVP transcription is now deprecated.
Added SimulationPhase which is returned by phase.get_simulation_phase.

does not support optimization
currently only supports a single default timeseries

Changes to ExplicitShooting

num_segments, order, compressed, segment_ends are no longer deprecated. This makes the interface to create the transcription the same as the others. If using these options rather than grid, it will use a GaussLobattoGrid by default.
fixed a units bug in the related VandermondeControlInterpComp

Changes to grid_data

grid_data now treats transcription_order argument the same whether it is a scalar or a length-1 array.

Related Issues

Backwards incompatibilities

None

New Dependencies

None

…g transcription without derivative propagation instead of the SolveIVP transcription. Fixed a get_cmap deprecation warning from matplotlib in the dymos plots.

…volve constraints and objectives have been overridden to inform the user that they are not to be applied to SimulationPhase. Fixed an issue in pseudospectral_timeseries_output_comp which was causing it to fail if the intput and output grids were the same but the output subset was not 'all'. Fixed a few indexing issues in grid_data. ExplicitShooting phases are now allowed to simulate.

…3.25.1 has been out for a while

Un-deprecated num_segments, compressed, segment_ends, and order options to ExplicitShooting. Defining the transcription using these options will now result in getting a GaussLobattoGrid by default. Fixed a bug in ExplicitShooting handling of matrix-shaped states. Fixed a units issue in TauComp.

…ring.

…f the groups IVC

… example

review-notebook-app · 2023-02-01T11:56:00Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

…dates for one more test.

… 3.17.0

robfalck · 2023-02-01T17:18:17Z

dymos/examples/aircraft_steady_flight/test/test_aircraft_cruise.py

@@ -83,7 +83,7 @@ def test_cruise_results_gl(self):
        p.setup()

        p['phase0.t_initial'] = 0.0
-        p['phase0.t_duration'] = 1.515132 * 3600.0
+        p['phase0.t_duration'] = 3600.0


initial value was out of bounds of the design variable.

robfalck · 2023-02-01T17:18:29Z

dymos/examples/aircraft_steady_flight/test/test_aircraft_cruise.py

@@ -181,7 +181,7 @@ def test_cruise_results_radau(self):
        p.setup()

        p['phase0.t_initial'] = 0.0
-        p['phase0.t_duration'] = 1.515132 * 3600.0
+        p['phase0.t_duration'] = 1.0 * 3600.0


initial value was out of bounds of the design variable

robfalck · 2023-02-01T17:19:49Z

dymos/examples/brachistochrone/test/test_brachistochrone_subprob_reports.py

-            # # There is the nominal problem, the simulation problem, and a subproblem for each segment in the simulation.
-            self.assertEqual(len(report_subdirs), 12)
+            # There is the nominal problem, the simulation problem, and a subproblem for the simulation.
+            self.assertEqual(len(report_subdirs), 3)


changed due to having fewer subproblems involved.

robfalck · 2023-02-01T17:20:29Z

dymos/examples/brachistochrone/test/test_ex_brachistochrone.py


    def test_ex_brachistochrone_radau_uncompressed(self):
        ex_brachistochrone.SHOW_PLOTS = True
        p = ex_brachistochrone.brachistochrone_min_time(transcription='radau-ps',
                                                        compressed=False)
        self.run_asserts(p)
        self.tearDown()
-        if os.path.exists('ex_brach_radau_uncompressed.db'):


Handled by @use_tempdir

naylor-b · 2023-02-01T17:36:04Z

docs/dymos_book/examples/robertson_problem/robertson_problem.ipynb

+    "\n",
+    "Since this is purely an integration with no controls to be determined, a single call to `run_model` will propagate the solution for us. There's no need to run a driver. Even the typical follow-up call to `traj.simulate` is unnecessary.\n",
+    "\n",
+    "Technically, we could even solve this using a single segment since segment spacing in the explicit shooting transcription determine the density of the control nodes, and there are no controls for this simulation.\n",


determine --> determines

naylor-b · 2023-02-01T17:39:08Z

dymos/phase/phase.py


-        # Copy over any simulation options from the simulate call.  The fallback will be to
+        # Copy over any simulation options from the simulate call.  The ffallback will be to


typo: ffallback

naylor-b · 2023-02-01T17:43:32Z

dymos/phase/test/test_time_targets.py

-                exp_out = p.model.phase0.simulate()
+                exp_out = p.model._get_subsystem('phase0').simulate()
+
+                time_comp = exp_out.model.phase0._get_subsystem(f'time')


fstrings are not needed in this block

naylor-b · 2023-02-01T17:47:08Z

dymos/transcriptions/solve_ivp/solve_ivp.py

@@ -32,6 +32,10 @@ class SolveIVP(TranscriptionBase):
        Dictionary of optional arguments.
    """
    def __init__(self, grid_data=None, **kwargs):
+        om.issue_warning('The SolveIVP transcription is deprecated. The simulate methods in Dymos now use a '


'Dymos now use a the ExplicitShooting...'

…work.

johnjasa

Looks great to me!

johnjasa · 2023-02-01T20:09:31Z

dymos/transcriptions/explicit_shooting/ode_integration_comp.py

@@ -664,8 +665,8 @@ def _f_primal(self, t, x, theta):
        ----------
        t : float
            The current value of the integration variable.
-        y : np.array
-            The augmented state vector.
+        x : np.array


Some part of this docstring doesn't jive; right after the function definition, it says "where y is the augmented state vector" but now there's no y? And dtheta_dz isn't to be seen either? Maybe just updating the comments is all that's needed

robfalck added 4 commits January 30, 2023 16:10

Started reworking a new simulate capability that uses ExplicitShootin…

152fb37

…g transcription without derivative propagation instead of the SolveIVP transcription. Fixed a get_cmap deprecation warning from matplotlib in the dymos plots.

added deprecation notice for SolveIVP

d370fbd

undo matplotlib change in this PR.

69bdaa4

robfalck requested review from johnjasa and kaushikponnapalli January 31, 2023 15:32

robfalck added 7 commits January 31, 2023 14:04

Fall back to the old way of initializing phase values until OpenMDAO …

24c7f2d

…3.25.1 has been out for a while

typo

0873b18

fixed a units issue in VandermondeControlInterpComp. Fixed poor docst…

5f79af6

…ring.

All rate sources in ode_evaluation_group are expected to be outputs o…

8f0857b

…f the groups IVC

tolerance adjustment and some slight changes to the robertson problem…

4a77f7e

… example

Cleaning up a few final tests.

371c1a1

robfalck added 5 commits February 1, 2023 07:26

Merge branch 'master' of https://github.com/OpenMDAO/dymos into i889

1ecc972

Fix to make sure SimulationPhase respects atol and rtol. Tolerance up…

a1fc4a8

…dates for one more test.

replaced use_tempdirs in test_time_targets

34ef146

Not using System.set_val yet, so minimum OpenMDAO version set back to…

d8f94bd

… 3.17.0

pep8

dfa86a5

robfalck commented Feb 1, 2023

View reviewed changes

naylor-b requested changes Feb 1, 2023

View reviewed changes

feedback from bret

ca41c12

robfalck mentioned this pull request Feb 1, 2023

phase.add_timeseries_output not working for promoted variable with ExplicitShooting #895

Closed

naylor-b approved these changes Feb 1, 2023

View reviewed changes

remove unnecessary code in get_simulation_phase that was duplicating …

6c87537

…work.

johnjasa approved these changes Feb 1, 2023

View reviewed changes

kaushikponnapalli approved these changes Feb 1, 2023

View reviewed changes

robfalck added 2 commits February 3, 2023 19:05

updated actions to use checkout@v3

e1a7778

Merge branch 'master' of https://github.com/OpenMDAO/dymos into i889

9993e22

robfalck merged commit b0d6657 into OpenMDAO:master Feb 6, 2023

robfalck deleted the i889 branch February 1, 2024 21:06

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Deprecated SolveIVP transcription. `simulate` method now uses the ExplicitShooting transcription without derivatives. #898

Deprecated SolveIVP transcription. `simulate` method now uses the ExplicitShooting transcription without derivatives. #898

robfalck commented Jan 31, 2023 •

edited

Loading

review-notebook-app bot commented Feb 1, 2023

robfalck Feb 1, 2023

robfalck Feb 1, 2023

robfalck Feb 1, 2023

robfalck Feb 1, 2023

naylor-b Feb 1, 2023

naylor-b Feb 1, 2023

naylor-b Feb 1, 2023

naylor-b Feb 1, 2023

johnjasa left a comment

johnjasa Feb 1, 2023


		# Copy over any simulation options from the simulate call. The fallback will be to
		# Copy over any simulation options from the simulate call. The ffallback will be to

Deprecated SolveIVP transcription. simulate method now uses the ExplicitShooting transcription without derivatives. #898

Deprecated SolveIVP transcription. simulate method now uses the ExplicitShooting transcription without derivatives. #898

Conversation

robfalck commented Jan 31, 2023 • edited Loading

Summary

Related Issues

Backwards incompatibilities

New Dependencies

review-notebook-app bot commented Feb 1, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

johnjasa left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Deprecated SolveIVP transcription. `simulate` method now uses the ExplicitShooting transcription without derivatives. #898

Deprecated SolveIVP transcription. `simulate` method now uses the ExplicitShooting transcription without derivatives. #898

robfalck commented Jan 31, 2023 •

edited

Loading