C++ API for JointController

[mogumbo]

Desired behavior

C++ API for JointController, JointPositionController, and JointTrajectoryController similar to the API found in gazebo::physics::JointController from Gazebo Classic. Specific methods I have used from gazebo::physics::JointController are GetJoints, GetPositions, GetPositionPID, GetVelocities, GetVelocityPIDs, SetPositionPID, SetPositionTarget, SetVelocityPID, and SetVelocityTarget.

Alternatives considered

You can interact with joint controllers using gz-transport, but results of these actions are not necessarily immediate and add complexity to your code if you need to receive a response.

Additional context

I'm currently using gz-sim8 on Ubuntu 24.04.

[scpeters]

I can imagine creating components like JointPositionControlTarget, JointVelocityControlTarget , JointPositionControlPID, JointVelocityControlPID that the joint control systems can try to read from the ECM in addition to gz-transport topics.

Would it also make sense to create a JointController C++ object similar to gz::sim::Model or gz::sim::Joint?

[azeey]

I support creating a gz::sim::JointController similar to gz::sim::Joint with all the PID functionality and having the current Joint controller plugins use that internally.

[yaswanth1701]

Hi everyone, can I work on this issue ?

[iche033]

yes please go ahead, I've assigned the issue to you

[yaswanth1701]

thanks a lot!

[yaswanth1701]

Hi, @iche033, which branch should I branch off from?

[iche033]

Can you target main for this new feature? We'll then backport to other release versions once it's merged.

[yaswanth1701]

Hi everyone. I want to ask a question regarding implementing the joint controller system. So, currently, the targetVel or targetPos are taken from the respective topics, but if it is set through the joint controller API for the same simulation step, which target value should be given preference? target topic value or JointControllerVelocityTarget/JointControllePositionTarget component value.

[iche033]

If one cmd is received after another in different time step, I would expect that the last received cmd overrides existing ones. However, it's in the same time step, I don't know if there is a strong preference for one type of cmd over another. We should print out an warning if this occurs and explicitly mention something like the JointController received both velocity and position commands in the same time step so position command will take precedence over velocity command (or the other way around)