WorldAngularVelocityCmd and WorldLinearVelocityCmd are ignored by Physics system

[scpeters]

Desired behavior

Currently the AngularVelocityCmd and LinearVelocityCmd components can be added to a model or link to set body-fixed velocity components (see Physics.cc:2406-2608. Components are also defined for WorldAngularVelocityCmd and WorldLinearVelocityCmd, but the Physics system ignores these components.

The Physics system should use these components so that velocities can be set in the world frame as an alternative to a body-fixed frame.

Alternatives considered

Implementation suggestion

This does raise a question what to do if an entity has both a *VelocityCmd and a World*VelocityCmd set; which one should be honored? Or should neither be honored?

I would suggest defining an order of precedence, perhaps preferring the body-fixed components since they were the first to be supported by the Physics system.

Additional context

[arjo129]

I suspect we should warn users if two plugins are setting VelocityCmd and World*VelocityCmd. My guess is the end user would be confused by the behavior if we don't warn them. We should have a defined ordering but at the very least by warning users, we would be preventing some degree of confusion.

On a related note, what happens if we set LinearVelocity* then try to apply a force? Or for instance the user sets a linear velocity and then tries to apply a force (it may be nice to support his at some point as it could clean up some of our tests)?

[azeey]

Let's discuss whether we should be encouraging the use of VelocityCmd or creating a new component specifically for setting initial velocities.

[scpeters]

Let's discuss whether we should be encouraging the use of VelocityCmd or creating a new component specifically for setting initial velocities.

We currently use JointPositionReset and JointVelocityReset to change joint states for a single timestep; so I suppose we could do the same with WorldLinearVelocityReset and WorldAngularVelocityReset.

[traversaro]

Let's discuss whether we should be encouraging the use of VelocityCmd or creating a new component specifically for setting initial velocities.

We currently use JointPositionReset and JointVelocityReset to change joint states for a single timestep; so I suppose we could do the same with WorldLinearVelocityReset and WorldAngularVelocityReset.

My two cents: I recently started contributing to a project that uses ros2_control + Gazebo/gz, and the use of control/cmd naming for operations that are actually fiddling with the physics engine in a way that has no direct relation in the real world has been indeed a major source of confusion (see also https://discord.com/channels/1077825543698927656/1080763334384697430/1196933980252029099). So I am super in favor of using Reset as opposed to Cmd in this context.

fyi @AleTarsi @pattacini