Keep Acrobot model files in sync

[jwnimmer-tri]

In https://github.com/RobotLocomotion/drake/tree/master/drake/examples/Acrobot, we have Acrobot.sdf and Acrobot.urdf. The files should be interchangeable, so we should ensure that the two files remain sync'd, in terms of the resulting MultibodyTree and their dynamics.

We had a program that performed this check at https://github.com/RobotLocomotion/drake/blob/e8ecbe96a6608b0ab61390b3dce06b6dc3fe4bfe/drake/multibody/rigid_body_system1/test/compareRigidBodySystems.cpp, but it got overtaken by events. We should consider reimplementing it, or else removing one of the two duplicated model files, or etc.

[amcastro-tri]

The acrobot benchmark will probably prove useful in verifying the two different models.

[amcastro-tri]

cc'ing @clalancette

[EricCousineau-TRI]

Per #11009 / #11011, I want to delete examples/Acrobot.{urdf,sdf} and only use the benchmark versions.

Once the examples models are deleted / fully slated for deprecation (and not used at all), we can close this issue.

[jwnimmer-tri]

@RussTedrake cares a lot about what the examples look like. I suggest getting his explicit approval on any plan that touches examples/acrobot, before merging any PRs.

[RussTedrake]

Just answered in #10294 . There is downstream code that depends on these (in underactuated).

[jwnimmer-tri]

Due to #6619 and #10294, I should probably own this issue as well.

[jwnimmer-tri]

Re-assigning to @mitiguy to resolve. In #13593 we're apparently adding yet another acrobot, buried inline in a unit test. That's beyond what I can retrofit.

[mitiguy]

@jwnimmer-tri I am not creating another acrobot in PR #13593. I think it important to be able to create small hand-crafted fixtures/models (e.g., a kinematic-only double pendulum) for testing purposes without reliance on other models that have other dependencies.

To make this point clearer, I am adding the following comment to the test fixture (model):
// Note: The applied forces (such as gravity) on this system are irrelevant as
// the plan for this text fixture is limited to testing kinematics for an
// instantaneous given state. The mass/inertia properties of this plant are
// irrelevant and there are no actuators, sensors, controller, or ports.

Note: As you deem best, feel free to unassign this issue to me.

[jwnimmer-tri]

I'll just close this as a duplicate of #10294. I'm not sure that the SDF vs URDF sync is the big deal, vs the C++ implementations. If we do care about the files in particular, we can have it as a subtask of #10294.