October 23, 2020
This is a long-overdue release of Dymos that features several useful new features.
By taking advantage of recent changes to the OpenMDAO setup process, we can use introspection to determine a lot of things automatically now.
For instance, the user will no longer be required to specify the shape or default units of timeseries outputs or constraints (although they can override units if they choose).
Also on the topic of timeseries, dymos now supports glob patterns when adding ODE outputs to the timeseries.
Adding all 100 outputs of a large ODE to the timeseries output can now be accomplished in a single line:
phase.add_timeseries_output('*')
Introspection is also used for states, time, and controls.
States will get their default units and shape from the associated state rate variable now.
Controls, Polynomial Controls, and Parameters will now get their default units and shapes from their targets.
With the advent of automatic IndepVarComps in OpenMDAO, there no longer needs to be a distinction between InputParameters and DesignParameters.
From here on out, users just add `parameters` to a Trajectory or Phase.
A user can choose to make them design variables (in which case they perform as DesignParameters), or not.
Connecting an external variable to the parameter makes it function as a InputParameter.
From here on, we plan on making releases of Dymos roughly once per month.
Backwards Incompatible API Changes & Deprecations:
- `add_input_parameter` and `add_design_parameter` are **deprecated** and replaced with `add_parameter` [365](https://github.com/OpenMDAO/dymos/pull/365)
- The use of two-character location strings ('--' and '++') are replaced by 'initial' and 'final' when linking phases. [427](https://github.com/OpenMDAO/dymos/pull/427)
- The endpoint conditions component, used to pull out initial or final values of a variable in a phase, have been removed. Values should instead by pulled from the timeseries. [427](https://github.com/OpenMDAO/dymos/pull/427)
- The results of simulate are now automatically recorded to a database recorder 'dymos_simulation.db' and stored in a case named 'final'. Use `CaseReader('dymos_simulation.db').get_case('final')` to access the case. instead of the previous behavior `CaseReader('dymos_simulation.db').get_case(-1)` [437](https://github.com/OpenMDAO/dymos/pull/437)
- The RungeKutta transcription is **deprecated**. It is functionally equivalent to the GaussLobatto transcription with an order of 3, but since it's signficantly different under the hood it was being a drag on development.
One note on another backwards incompatibility.
Since we're using some newer OpenMDAO capabilities for the way in which parameters are handled, it is currently not possible to connect an upstream output to a parameter in dymos _when src_indices are used_.
We plan on fixing this issue with an upcoming OpenMDAO release.
If you rely on this capability, we recommend waiting for that update before using this version of Dymos, or using an intermediate "pass-through" component to pull the correct index from the original output and then passing _that_ as the value of a dymos parameter.
Enhancements:
* Dymos components are no longer listed in check_partials output by default. Use `dymos.options['include_check_partials'] = True` to override this for debugging. [438](https://github.com/OpenMDAO/dymos/pull/438)
* Any outputs can now be linked across phases. Added a more general and powerful `add_linkage_constraint` method, but the simpler `link_phases` method remains in place for simple continuity. [422](https://github.com/OpenMDAO/dymos/pull/422)
* Time, controls, and polynomial control units are now determined from targets. [412](https://github.com/OpenMDAO/dymos/pull/412)
* States now pull their default units and shapes from the rate source. [398](https://github.com/OpenMDAO/dymos/pull/398)
* Phase method `add_timeseries_output` now supports wildcards to allow multiple timeseries outputs to be added at once. [387](https://github.com/OpenMDAO/dymos/pull/387), #387
* Timeseries outputs automatically detect shape and default units. [380](https://github.com/OpenMDAO/dymos/pull/380)
* Added ph-adaptive refinement method that is capable of shrinking the grid. [379](https://github.com/OpenMDAO/dymos/pull/379)
* Control targets will automatically be set to a top-level input of the ODE, if present. [373](https://github.com/OpenMDAO/dymos/pull/373)
* State targets will automatically be set to a top-level input of the ODE, if present. [356](https://github.com/OpenMDAO/dymos/pull/356)
* Water-powered rocket MDO example added. [343](https://github.com/OpenMDAO/dymos/pull/343)
* State rates are included in timeseries outputs by default. [329](https://github.com/OpenMDAO/dymos/pull/329)
* By default, run_problem will use a problem recorder to record only the 'final' solution after an optimization. [328](https://github.com/OpenMDAO/dymos/pull/328)
* Move most setup functionality to configure to allow more introspection changes. [327](https://github.com/OpenMDAO/dymos/pull/327)
* Include examples using IPOPT via pyoptsparse. [311](https://github.com/OpenMDAO/dymos/pull/311)
* Parameters may now be selectively omitted from the timeseries outputs, but are included by default. [305](https://github.com/OpenMDAO/dymos/pull/305)
* Added a test of the distributed ODE capability. [304](https://github.com/OpenMDAO/dymos/pull/304)
* Replace `load_case` with `reinterpolate_solution`. [296](https://github.com/OpenMDAO/dymos/pull/296)
Bug Fixes:
* Fixed a bug that was causing matrix-shaped states to have connection errors. [441](https://github.com/OpenMDAO/dymos/pull/441)
* Fixed a bug where simulate failed when both standard and polynomial controls were present. [434](https://github.com/OpenMDAO/dymos/pull/434)
* Fixed an issue where parameters are promoted twice in some cases. This will cause errors in an upcoming versions of OpenMDAO. [429](https://github.com/OpenMDAO/dymos/pull/429)
* Fixed a bug that was causing matrix-shaped states to have connection errors. [441](https://github.com/OpenMDAO/dymos/pull/441)
* AnalysisError is now raised if scipy.integrate.solve_ivp fails during simulation. [400](https://github.com/OpenMDAO/dymos/pull/400)
* Fixed a bug involving GaussLobatto transcriptions `get_rate_source` method. [334](https://github.com/OpenMDAO/dymos/pull/334)
* Fix for simulate encountering errors when time options `input_initial` or `input_duration` were True. [317](https://github.com/OpenMDAO/dymos/pull/317)
* Fixed a bug in which shaped input parameters were breaking simulate. [301](https://github.com/OpenMDAO/dymos/pull/301)
* Sort linkage orders to make convergence more repeatable. [298](https://github.com/OpenMDAO/dymos/pull/298)
Miscellaneous:
* Switch to use preferred `assert_near_equal` method instead of `assert_rel_error` from OpenMDAO. [320](https://github.com/OpenMDAO/dymos/pull/320)
* Documentation is now handled via mkdocs instead of sphinx. [337](https://github.com/OpenMDAO/dymos/pull/337), [#332](https://github.com/OpenMDAO/dymos/pull/332)