Add an example with a passive joint

[christophfroehlich]

Adds a demo example with a passive joint: A pendulum hinged on the existing cart example.

I also added a hint that the physics of the passive joint are not rendered correct if position or velocity interface is used.

[ahcorde]

Suggested change

	The following example shows a cart with a pendulum arm. This uses the effort command interface for the cart's degree of freedom on the rail, and the physics of the passive joint of the pendulum is solved correctly.
	```bash
	ros2 launch gazebo_ros2_control_demos pendulum_example_effort.launch.py
	ros2 run gazebo_ros2_control_demos example_effort
	```
	*Attention*: If the position or velocity command interface is used instead, the motion of the pendulum is not calculated correctly and does not move at all, see the following example. This also holds true if a mimicked joint with position or velocity interface is used.
	```bash
	ros2 launch gazebo_ros2_control_demos pendulum_example_position.launch.py
	ros2 run gazebo_ros2_control_demos example_position
	```
	The following example shows a cart with a pendulum arm. This uses the effort command interface for the cart's
	degree of freedom on the rail, and the physics of the passive joint of the pendulum is solved correctly.
	```bash
	ros2 launch gazebo_ros2_control_demos pendulum_example_effort.launch.py
	ros2 run gazebo_ros2_control_demos example_effort

Attention: If the position or velocity command interface is used instead, the motion of the pendulum is not
calculated correctly and does not move at all, see the following example. This also holds true if a mimicked joint
with position or velocity interface is used.

ros2 launch gazebo_ros2_control_demos pendulum_example_position.launch.py
ros2 run gazebo_ros2_control_demos example_position

[christophfroehlich]

Do I understand correctly that you want to delete the hint with the position interface?

[christophfroehlich]

@ahcorde could you please give me an update about what you want me to change here? According to the reactions below, I'd suggest to add the hint with the position interface again.

[Fixit-Davide]

Sorry, what is the technical reason for the " If the position or velocity command interface is used instead, the motion of the pendulum is not calculated correctly and does not move at all"?

[christophfroehlich]

Sorry, what is the technical reason for the " If the position or velocity command interface is used instead, the motion of the pendulum is not calculated correctly and does not move at all"?

I never had a look into the gazebo code, but what I think from a mathematical/physical perspective what happens is:
If you use the effort interface, then the physics engine solves the whole rigid-body-dynamics, for example written in the standard matrix notation:

If we use a different interface (velocity, position) for a degree-of-freedom, then the states of the respective DOF are hardcoded and the coupling via M \DDot q is broken, because the acceleration of this DOF in \DDot q is not updated.

Maybe @ahcorde can verify my guess.

[ahcorde]

can you open an issue about this?

[christophfroehlich]

I'm not sure if this is a bug or a feature :D

[ahcorde]

the urdfs are different one is using a macro and the other doesn't.

I tested both examples, the effort one is working and I can see the passive joint moving, but the position one the passive joint is static.

Do you mind to take a look ? and reduce diff between the two new URDFs ?

[christophfroehlich]

the urdfs are different one is using a macro and the other doesn't.

I tested both examples, the effort one is working and I can see the passive joint moving, but the position one the passive joint is static.

exactly, that is the point I wanted to highlight and have adressed in the docs and opened the issue: it would have saved me some days of debugging of my project ;)

Do you mind to take a look ? and reduce diff between the two new URDFs ?

yes I'll have a look why the URDFs are different

[christophfroehlich]

@ahcorde I made both URFDs equal now (inertia was not important for the position case).

[ahcorde]

@christophfroehlich

The position joints are working now, can we close this issue #240 ?

Do you want to backport this ?

[christophfroehlich]

The position joints are working now, can we close this issue #240 ?

What do you mean? I don't think that this behavior changed, at least not with this PR. The pendulum does not move if the cart is commanded with position interface?

Do you want to backport this ?

yes please. the rolling version will die when gazebo classic is not released to noble anymore.

[ahcorde]

Now it's moving:

[ahcorde]

https://github.com/Mergifyio backport humble iron

[mergify]

backport humble iron

✅ Backports have been created

• #306 Add an example with a passive joint (backport #172) has been created for branch humble but encountered conflicts
• #307 Add an example with a passive joint (backport #172) has been created for branch iron but encountered conflicts

[christophfroehlich]

Now it's moving:

Passive joints were always "moving", but as if they were fixed in inertial frame. A movement of the parent link did not effect the respective joint.

For demonstration purposes, I change the initial pose to be in the lower equilibrium to make this clearer:
https://github.com/ros-controls/gazebo_ros2_control/assets/3367244/630cc8b8-6cb4-4dbc-9caa-948c085052e0