Custom Termination in TUDAT: Ending Propagation at a Specific Value

[luigigisolfi]

Question

User:
Hi everyone!
I am using propagation_setup.propagator.custom_termination as a termination condition for my simulation, which works well enough for my needs but hasn't the built-in capability to end the propagation at a specified value. My question was then if there is any way of using some TUDAT tools to make the custom termination to happen at a specific value of a custom calculated quantity.
As an example, I am using it as the following:

bodies = ... (SystemOfBodies object)
entry_fpa = ... (float)

def custom_condition(time:float) -> bool:
    # Calculate the inertial flight path angle of the body 'Capsule'
    capsule_state = bodies.get('Capsule').state
    state_position = capsule_state[0:3]
    state_velocity = capsule_state[3:6]
    dot_product = np.dot(state_position, state_velocity) / (LA.norm(state_position) * LA.norm(state_velocity))
    inertial_fpa = np.pi / 2 - np.arccos(dot_product)
    # Terminate if the inertial flight path angle exceeds a set value
    do_terminate = inertial_fpa >= entry_fpa
    return do_terminate
inertial_fpa_termination_settings = propagation_setup.propagator.custom_termination(custom_condition)

I would like the propagation to terminate when inertial_fpa ~ entry_fpa, is this possible in TUDAT?

[luigigisolfi]

Answer

Tudat-team:
I think you should be able to do this by creating a custom dependent variable, and then terminate on an exact value of this custom dependent variable. I see the custom dependent variable is not yet documented here:
https://py.api.tudat.space/en/latest/dependent_variable.html
but, in short, it would be an additional factory function in the long list, with the name:

custom_dependent_variable

It would have 2 inputs:
custom_function : A function that takes no inputs, and returns a numpy array as output
variable_size : The size of the numpy array returned by the custom_function
If you create your custom dependent variable correctly, I think as:

def inertial_fpa( ):
    # Calculate the inertial flight path angle of the body 'Capsule'
    capsule_state = bodies.get('Capsule').state
    state_position = capsule_state[0:3]
    state_velocity = capsule_state[3:6]
    dot_product = np.dot(state_position, state_velocity) / (LA.norm(state_position) * LA.norm(state_velocity))
    inertial_fpa = np.pi / 2 - np.arccos(dot_product)
    return np.array([inertial_fpa])

and create the dependent variable with
propagation_setup.dependent_variable.inertial_fpa(custom_dependent_variable, 1 )

you should be able to terminate on an exact value of this dependent variable.

[luigigisolfi]

Tudat-team:

Yes! You're right :)

[luigigisolfi]

User:
The custom variable works well when added to the dependent variables to save, however it is not working when used as a termination condition. Here's my code:

def inertial_fpa():
    capsule_state = bodies.get('Capsule').state
    state_position = capsule_state[0:3]
    state_velocity = capsule_state[3:6]
    dot_product = np.dot(unit_vector(state_position), unit_vector(state_velocity))
    inertial_fpa = np.pi / 2 - np.arccos(dot_product)
    return np.array([inertial_fpa])

inertial_fpa_termination_settings = propagation_setup.propagator.dependent_variable_termination(
    dependent_variable_settings=propagation_setup.dependent_variable.custom_dependent_variable(inertial_fpa, 1),
    limit_value=entry_fpa,
    use_as_lower_limit=False,
    terminate_exactly_on_final_condition=True,
    termination_root_finder_settings=root_finders.bisection(maximum_iteration_handling=root_finders.MaximumIterationHandling.accept_result_with_warning)
)

Everything is declared within a get_termination_settings(bodies, entry_fpa) function. It worked fine when I used the custom termination I showed in my first question.
As error I get this one:

Traceback (most recent call last):
  File "C:\Users\viero\PycharmProjects\ThesisProject\Just_aerocapture\just_aerocapture_numerical_uncertaintyPropagation.py", line 223, in <module>
    aerocapture_problem.fitness([interplanetary_arrival_velocity, flight_path_angle_at_atmosphere_entry])
  File "C:\Users\viero\PycharmProjects\ThesisProject\class_AerocaptureNumericalProblem.py", line 328, in fitness
    dynamics_simulator = numerical_simulation.SingleArcSimulator(
RuntimeError: Error, did not recognize double dependent variable type when making variable function: 48

[luigigisolfi]

Tudat-team:

Huh, after diving into the code this makes sense: the custom dependent variable is implemented as a vector-valued dependent variable, and termination can only be done on a single-valued dependent variable. I've added a small block of code to circumvent this, and with the dev package that should be built tomorrow morning you should be able to run your code. The custom dependent variable function should still return a vector, but if the vector is size 1, you can use it as a termination. Note that I haven't tested it, so I may be wrong.
If you see a new tudat and tudatpy build here:
https://dev.azure.com/tudat-team/feedstock-builds/_build
by tomorrow morning, and both are green, get the latest dev version of the conda packages, and rerun your code

[luigigisolfi]

User:
Just tried it and it works, thanks!

Also, are you planning to add some documentation on root finders anytime soon? I tried to guess how to use them based on Ctrl+clicking the module on pycharm, but I think adding it to the website documentation would help a lot :)

[luigigisolfi]

Tudat-team:

Indeed it would, but our resources for this the coming ~months probably won't allow for it. It is indeed an important point to add, but there are also a few points in the implementation that I would like to have streamlined before documenting it (so that we don't document it, change everything, and then have to start over). Until then, I'd recommend:

• Have a look at the actual Python exposure code
• Bombard us with questions