What does yarp::dev::IVelocityControl::stop() actually do at the lowest level?

[pattacini]

We all know that velocity commands are integrated by the FW to retrieve position references to follow by means of the PID.

Whenever a yarp::dev::IVelocityControl::stop() is issued, what actually happens at the lowest level? Two possibilities are given:

1. The FW simply stops integrating as it received a velocityMove(0.0).
2. The FW sets the internal position reference equal to the current encoder reading.

Is there any third one?

I'm a bit concerned about hypothesis 1: because of a poorly tuned PID, the joint will be still moving for a while after the stop command is received.

I hope the implementation resorts to 2, indeed.

Second question

Does the joints subset version of the stop command travel over the rpc channel or over the streaming channel? Can we be sure that the stop command is never lost?

/cc @randaz81 @barbalberto

[pattacini]

Backstage

I'm asking this because while teleoperating the iCub I've realized that the fingers - which are controlled in velocity and whose control is a bit degraded - do not stop exactly upon stop reception but, instead, proceed moving for a short while afterwards.

[ale-git]

I can answer about the joints controlled by 2FOC boards (thus not regarding the fingers), where the stopping method is 2 (new reference = encoder reading).

[pattacini]

@ale-git cool thanks 👍
What about the MC4 (controlling the fingers) then?

[barbalberto]

All stop commands are sent using rpc.

[pattacini]

Thanks @barbalberto 👍
Thus, only one doubt (the main one) remaining.

[randaz81]

The implementation 2 is already used on MC4 boards.

    if (_control_mode[axis] == MODE_IMPEDANCE_POS || \
        _control_mode[axis] == MODE_POSITION || \
        _control_mode[axis] == MODE_IMPEDANCE_VEL || \
        _control_mode[axis] == MODE_VELOCITY || \
        _control_mode[axis] == MODE_MIXED) \
        { \
            _set_vel[axis] = 0; \
            _set_acc[axis] = 0; \
            _desired_vel[axis] = 0; \
            init_trajectory (axis, _position[axis], _position[axis], 1); \
        } \

[pattacini]

Great, thanks @randaz81
Hence I'd need to better inquire what I've been observing with the fingers!