Update the PID gains for all the relevant robots

[pattacini]

Follow-up of #3.

While doing this, we should publish a nice writeup on the community for documentation purposes.

[pattacini]

@martinaxgloria could be doing this update.

@mfussi66 and/or @Nicogene can act as sidekicks.

[pattacini]

We could grab the opportunity to take @martinaxgloria to the physical robot and start the training for using it.

[mfussi66]

We could grab the opportunity to take @martinaxgloria to the physical robot and start the training for using it.

To this end we can check the availability of iCubGenova11 or iCub3.

[S-Dafarra]

We could grab the opportunity to take @martinaxgloria to the physical robot and start the training for using it.

To this end we can check the availability of iCubGenova11 or iCub3.

iCub3 currently has the arms unmounted for iRonCub tests. You may want to double-check with @gabrielenava (not sure if he has access to this repo) to use it

[pattacini]

(not sure if he has access to this repo) to use it

Yep, this repo has been made public.

[gabrielenava]

iCub3 currently has the arms unmounted for iRonCub tests. You may want to double-check with @gabrielenava (not sure if he has access to this repo) to use it

we unmounted the arms for FT calibration tests, but the robot can be used. If you need to mount back the arms we can organize (cc @hosameldinmohamed @vpunithreddy)

[mfussi66]

iCub3 currently has the arms unmounted for iRonCub tests. You may want to double-check with @gabrielenava (not sure if he has access to this repo) to use it

we unmounted the arms for FT calibration tests, but the robot can be used. If you need to mount back the arms we can organize (cc @HosameldinMohamed @vpunithreddy)

Actually we would just need to move the head, thanks!

[mfussi66]

iCub3 booked for friday morning.

[martinaxgloria]

Yesterday I started modifying the PID values in the robots-configuration files. I started filtering the iCub with a head_version ≥ 2.0. Among those, I found some anomalies reported in the table above:

ROBOT NAME	SERIAL No.	DESCRIPTION
iCubBielefeld03	S/N: 019	No .xml files
iCubDarmstadt01 iCubNottingham01 iCubSheffield01 iCubTwente01	S/N: 025 S/N: 027 S/N: 028 S/N: 030	Device type: canmotioncontrol (no xml file concerning the head parameters)
iCubOsaka01 iCubGrenoble01 iCubParis02	S/N: 022 S/N: 023 S/N: 024	Device type canmotioncontrol (the icub_head.xml contains PID values strangely different wrt iCubGenova11, for example)

Just to be more precise, the configuration files relative to the robots listed above are not changed for at least 5 years.

cc @mfussi66 @pattacini @davidetome @Fabrizio69

[pattacini]

Hi @martinaxgloria

Good that you reported these anomalies, which need to be definitely discussed and analyzed 👍🏻

Just a comment regarding:

I started filtering the iCub with a head_version higher than 2.0.

⚠️ The condition shall be head_version $\ge$ 2.0.

[martinaxgloria]

⚠️ The condition shall be head_version ≥ 2.0.

Yes, I considered this, I just reported it wrong, sorry. I edited the previous comment with the right condition.

[martinaxgloria]

Update: I talked with @davidetome, who gave me some information about the anomalies observed.

ROBOT NAME	SERIAL No.	DESCRIPTION
iCubDarmstadt01 iCubNottingham01 iCubSheffield01 iCubTwente01	S/N: 025 S/N: 027 S/N: 028 S/N: 030	Device type: canmotioncontrol (there is only the mechanical head configuration file and not the motor control one)

The robots above have a head_version ≥ 2.0 but of type canmotioncontrol, while the rest of the body is embObjMotionControl.

ROBOT NAME	SERIAL No.	DESCRIPTION
iCubOsaka01 iCubGrenoble01 iCubParis02	S/N: 022 S/N: 023 S/N: 024	Device type canmotioncontrol (the icub_head.xml contains PID values strangely different wrt iCubGenova11, for example)

Those, instead, are entirely canmotioncontrol robots.

The PID values are reported above and, in particular, the first two columns regard the neck_pitch and neck_roll:

<group name="POS_PIDS">      
        <param name="kp">           6000          -6000         3200          -400          100           100           </param>       
        <param name="kd">           500           -500          3200          -400          1000          1000          </param>       
        <param name="ki">           5             -5            10            -4            10            10            </param>        
</group>      

while the PID values of iCubGenova11 and of the other embObjMotionControl robots are:

<group name="POS_PID_DEFAULT">  
        <param name="kp">                       -300        +300        </param>
        <param name="kd">                       -10         +10         </param>
        <param name="ki">                       -100        +100        </param>
</group>

[pattacini]

If I can make a general comment, I would proceed here iteratively.

First off, attack all the robots that do not have anomalies. For this, a preliminary PR can be opened. Of course, it's important to copy out the PID gains from the correct robot (@mfussi66 can point you at it).

Then (or in the meantime), we can understand how to deal with the anomalies. We may even decide not to touch them. The CAN-based robots are pretty old. TBD w/ @Nicogene too.

[martinaxgloria]

PR opened here.

[pattacini]

Nice message in the PR body!

[pattacini]

Hi @martinaxgloria

I've seen that you updated 3 robots with the latest commits, so how did you ensure full coverage of the robots? Which method did you use? Any robot still missing?

[martinaxgloria]

Before the last commit in which I uploaded the 3 robots you mentioned I followed this list maintained by @Fabrizio69. Since you pointed out that some robots were forgotten, I double-checked this file with the iKinGazeCrtl files in robots-configuration: in doing this, I saw that three robots I missed were not present in that list, this is why I didn't upload them with the others.

Actually, there are some robots with incongruences between the head_version in the iKinGazeCtrl.ini and what it is reported in that list of robots. Until now, I have uploaded only the ones with a coherent head_version between the two files.

[pattacini]

in doing this, I saw that three robots I missed were not present in that list, this is why I didn't upload them with the others.

Yeah, indeed, this is the pitfall of relying on lists that are "far" from the SW. They may be not up-to-date.

Better off performing a thorough check then.

[martinaxgloria]

Today I can do a search in folder in robots-configuration to check here the correct head_version and see if I forgot some robots.

[martinaxgloria]

After a further check, I found out that iCubLisboa01 was the last robot to be updated and I did it.

[pattacini]

Cool! I'll merge the PR then.

[pattacini]

For the leftover, I don't have a strong opinion. We may leave them unchanged.

@Nicogene can say the last word.

[Nicogene]

For the leftover, I don't have a strong opinion. We may leave them unchanged.

@Nicogene can say the last word.

I don't hear news about those robots in a while, I think we can mark as done this activity, and eventually, we will tackle also the remaining ones in another stint if needed

[pattacini]

Super! Thanks again @martinaxgloria!