Minor improvements: clean-up examples, throw_pretty in Python actuaction, and remove memory allocation in quasiStatic

[cmastalli]

The changes on the examples are targeted to extend them along a set of cases that used inverse-dynamics (instead of the current forward dynamics).

Note that the inverse-dynamics based optimal control is still an ongoing work which will be released in Crocoddyl v2.

[cmastalli]

@wxmerkt -- do you know why the ROS-CIs are not passing?

[wxmerkt]

The momentum cost test failed. I restarted it in case it was flaky. However, it's likely that we are already using Pinocchio 2.6.3 on the ROS-CI which has the fixed centroidal momentum derivatives in case that has an impact here or requires adjustments to the unit test.
30 - pybinds-momemtum_cost (Failed)

[cmastalli]

The momentum cost test failed. I restarted it in case it was flaky. However, it's likely that we are already using Pinocchio 2.6.3 on the ROS-CI which has the fixed centroidal momentum derivatives in case that has an impact here or requires adjustments to the unit test.
30 - pybinds-momemtum_cost (Failed)

I tested locally, the error appears with Pinocchio 2.3.6 (i.e., after the changes in stack-of-tasks/pinocchio#1474).

It seems that we don't need anymore these lines: https://github.com/loco-3d/crocoddyl/blob/devel/include/crocoddyl/multibody/residuals/centroidal-momentum.hxx#L50-L52. At least, the test is passing if I remove them.

@jcarpent -- I need your help here. Should I remove those lines?

[jcarpent]

@jcarpent -- I need your help here. Should I remove those lines?

Yes, you have to.

[proyan]

Yes, I put these lines as a local solution for the bug. Since we have now fixed it in pinocchio as well, we should also update the minimum version to 2.6.3

[proyan]

Also, it looks like LAAS-CNRS is missing from the copyright in all residual models. Since the residual-models were refactored from the original cost-models, the copyright should remain.

[cmastalli]

Also, it looks like LAAS-CNRS is missing from the copyright in all residual models. Since the residual-models were refactored from the original cost-models, the copyright should remain.

Done!

The code is ready to be merged; however, gitlab-ci is failing since we need to deploy 2.6.3 in robotpkg. @nim65s -- do you know when it is expected the new release.

[cmastalli]

I will merge this PR as ROS-CI is checking all the unittests (and I don't want to wait for robotpkg release)

[nim65s]

We could have used a version check instead of requiring the latest version.

[cmastalli]

We could have used a version check instead of requiring the latest version.

I indeed though about this possibility, but it seems more convenient to constrain everyone moving to another version as this is a bug.
We could still add the version check (as we just merged it), if you believe it is a good idea.

[proyan]

@cmastalli Accepting PRs when CI is failing is never a good idea. It might not be the cleanest way forward, but it is better to leave a PR open than to accept it without the CI.

[cmastalli]

@cmastalli Accepting PRs when CI is failing is never a good idea. It might not be the cleanest way forward, but it is better to leave a PR open than to accept it without the CI.

There is redundancy, the ROS-CI is testing the same bits but with Pinocchio 2.6.3.