Specify used F/T sensors in wholebodydynamics with an explicit parameter

[traversaro]

TODO

cc @gabrielenava @ahmadgazar remember me to update this issue

[gabrielenava]

ping :-)

[ahmadgazar]

I am sorry I don't get what do you mean by the explicit parameter of the F/T sensors? @traversaro

[fjandrad]

I think this is meant to allow to enable/disable and FT sensor from the perspective of the estimation without the need to modify the urdf model.

[traversaro]

Exactly, something like having a remapper for SensorsList (I think there was a similar requirement coming from Andy guys) instead of using all the FT sensors specified in the URDF.

[fjandrad]

Procedure to remove sensor at the foot

Remove the sensor tag in the the urdf depending on the sensor to be eliminated l_foot_ft_sensor or r_foot_ft_sensor.
In this example it will be the left foot sensor.
Since at the foot, there is no other joint after the sensor, wholebodydynamics will collapse the following links to the parent of the ft sensor, the l_ankle_2 link.
Due to this all references to l_foot or l_sole as a link should be changed to l_ankle_2 in the wholebodydynamics.xml file this includes the skindynlib section but more importantly the following lines:

 <param name="/wholeBodyDynamics/left_foot/cartesianEndEffectorWrench:o">(l_foot,l_sole,l_sole)</param>

 <param name="l_foot">(l_foot_dh_frame,5,5)</param>

The attach part of the ft sensor should also be commented
<elem name="l_foot_ft_sensor">left_leg-eb7-j4_5-strain</elem>

[traversaro]

It is is difficult for me to address this in the short term, moving from Assignee: Traversaro to Status: Help Wanted.

[fjandrad]

Considering I'll dive in into wholebodydynamics device to include the temperature section I might as well take a look at this since is useful for testing ft sensor configurations quickly.
@traversaro we could use the meeting on Monday so you can tell me how you were planning on doing this, to get a head start

[traversaro]

Yes!

[fjandrad]

cc @gabrielenava

[fjandrad]

For developing the temperature compensation the multipleAnalogSensorsNames variable was inserted, it allows to signal which are the sensors that are a MAS. We might include a similar variable for selecting which sensors to use. Things to keep in mind are:

• To keep a similar logic as the other variables we will add the sensors to be considered.
• The new variable should have priority over multipleAnalogSensorsNames
• Handle the case where multipleAnalogSensorsNames could be a valid name but it was removed by the new variable. Should send a warning or fail if a name is inserted that is not there?

[traversaro]

Should send a warning or fail if a name is inserted that is not there?

Fail. Fail early, fail fast.

[fjandrad]

@prashanthr05 this is the issue about enabling/disabling the ft sensors

[traversaro]

@prashanthr05 this is probably related to the problem of excluding the FT sensors on the hip that we were discussing yesterday.

[prashanthr05]

I think this is meant to allow to enable/disable and FT sensor from the perspective of the estimation without the need to modify the urdf model.

Ideally, this would be a nice feature to have.
But commenting out the sensors in the URDF seems to be a workaround solution for quick hacks.

[S-Dafarra]

Let me revive this issue, as it happens often, also during demos, that we need to avoid using an F/T from the robot. Editing the URDF is usually something not so convenient

[fjandrad]

Wow its been a while. This takes me back haha. I think I was very close to having this ready or that might just me editing my past memories. What you want to do might be related to this function

whole-body-estimators/devices/wholeBodyDynamics/WholeBodyDynamicsDevice.cpp

Line 767 in 5fb9aee

bool WholeBodyDynamicsDevice::openMultipleAnalogSensorRemapper(os::Searchable &config)

. I vaguely remember activating one or the other depending if a sensor had the temperature component or not.

[fjandrad]

It could also be that you can modify the estimator in here https://github.com/robotology/idyntree/blob/759fd719b3ad7f6c63f565d99250dd7174177161/src/estimation/src/ExtWrenchesAndJointTorquesEstimator.cpp#L133 . Could also provide a list of the sensors there.

[traversaro]

Thanks @fjandrad !

Let me revive this issue, as it happens often, also during demos, that we need to avoid using an F/T from the robot. Editing the URDF is usually something not so convenient

As discussed in person, probably we another thing we could implement that would be convenient is to actually get the FT sensors to use from YARP instead of from the URDF. In general you would not like to do that (as if there is some kind of error on the YARP side, you only use a subset of the sensors without realized that), but in case some FT sensors are missing you want to do that.