Add friction + coulomb + threshold parameters

[sgiraz]

What's new:

• As per Friction model coulomb + viscous cannot be used without rebuilding the firmware icub-firmware#396 removes the old kbemf torque parameter in favor of the new torque control parameters:
  • viscousPos
  • viscousNeg
  • coulombPos
  • coulombNeg
  • verlocityThres

The new friction compensation law is the following one (see also the source code):

float PID_do_friction_comp(PID *o, float32_t vel_fbk, float32_t trq_ref)
{
    const float32_t MICRO { 1000000.0 }; // conversion from Nm into microNm
    const float32_t coulomb_pos_converted = o->coulomb_pos_val * MICRO;
    const float32_t coulomb_neg_converted = o->coulomb_neg_val * MICRO;
    
    const float32_t viscous_neg = o->viscous_neg_val * vel_fbk;
    const float32_t viscous_pos = o->viscous_pos_val * vel_fbk;
    
    if (vel_fbk <= -o->velocityThres_val)
    {
        return o->Ktau*(-coulomb_neg_converted + viscous_neg + o->Kff*trq_ref);
    }
    else if (vel_fbk > o->velocityThres_val)
    {
        return o->Ktau*(coulomb_pos_converted + viscous_pos + o->Kff*trq_ref);
    }
    else
    {
        return o->Ktau*(o->Kff*trq_ref);
    }
}

Note:

• The following robots will be affected by this changes:

#	robot
1	ergoCubSN000
2	ergoCubSN001
3	iCubEdinburg01
4	iCubErzelli01
5	iCubErzelli02
6	iCubGenova02
7	iCubGenova07
8	iCubGenova08
9	iCubGenova10
10	iCubGenova11
11	iCubLisboa01
12	iCubNancy01
13	iCubPargue01
14	iCubShanghai01
15	iCubShenzhen01
16	iCubSingapore01
17	iCubTemplates
18	iCubValparaiso01
19	iCubWaterloo01
20	iCubZagreb01

cc @isorrentino @marcoaccame @Nicogene

[pattacini]

Hi @sgiraz

To guarantee back-compatibility, we need to make sure that the values of viscousPos and viscousNeg equate to the values of kbemf¹. This is true for most of the robots but NOT for those where kbemef <> 0.

Thus, there are robots requiring a fix.

Footnotes

1. See https://github.com/robotology/icub-firmware/pull/410/files. ↩

[sgiraz]

To guarantee back-compatibility, we need to make sure that the values of viscousPos and viscousNeg must equate the values of kbemf1. This is true for most of the robots but NOT for those where kbemef <> 0.

Hi @pattacini,

Changes applied! Thanks for your feedback.

[pattacini]

Thank you @sgiraz!

I'll wait a bit for @isorrentino and/or @GiulioRomualdi to check the changes for ergoCub before merging.

[pattacini]

/remind December 18, merge the PR.

[octo-reminder]

⏰ Reminder
Monday, December 18, 2023 10:00 AM (GMT+01:00)

, merge the PR.

[isorrentino]

The modifications seem okay. Obviously, the parameters provided are not the ones that make the joints work in torque, but with those values we are confident that nothing will happen if the torque control is activated.

[octo-reminder]

🔔 @pattacini

, merge the PR.