Using `set_duration_balance` without `solve_segments` in pseudospectral phases should raise exceptions in certain situations.

[robfalck]

Description

Set duration bounds relies on the residual variable being a dependent upon the phase duration. When pseudospectral phases are being solved by the optimizer, there is no direct link between the phase duration and the state values.

Expected Behavior

Using set_duration_balance with the pseudospectral and Birkhoff transcriptions should raise an exception unless the transcription has solve_segments set to forward or backward for the duration balance variable.

If the duration balance is some function of the a solve_segment variable, dymos cannot detect that this situation exists. In this scenario, the user must set solve_segments to True at the transcription level (or, equivalently, for all states).

set_duration_balance will have a new argument, loc='initial'|'final'. The default value is to use the final value, which is the current behavior. However, if solve_segments='backward', then the loc of the associated duration balance must be initial.

Example

TBD

Dymos Version

1.10.1-dev

Relevant environment information

No response