Calibrating multiple motors (hardware stop) using the MC4-PLUS board

[plinioMoreno]

We are creating the configuration files for our robot Vizzy, which will use the ETH-based boards (the repository of the files is here). We are running into issues when calibrating two DC motors at the same time. When running the calibration at the same time (see config files), the pwm is commanded and the motors start to move to reach the hard stop, but the motors do not move to the home position, and the message we get in the console is that the calibration is finished and worked properly.

To figure out if the problem is from the files, we run the calibration independently for each motor using these config files, and in this case the calibration works as expected.

Since the parameters of the calibration for both tests were the same, we wonder how can we figure out the problem. Any help is very appreciated.

cc @marcoaccame @ale-git @Ryo-Naohara

[marcoaccame]

Hi @plinioMoreno,
we had a look at the xml files and we surely have some basic suggestions. For instance, the file head-eb05-j0_1-mec.xml seems wrong in the the COUPLING and JOINTSET_CFG groups.

To give more focused suggestions, however, we also need to know some more about the mechanical setup.

We can start from one point: the step-by-step first movement of joints 0 and 1.

Step by step movement of joints 0 and 1.

To debug your setup you can begin to calibrate and move joint 0 and 1 considering them independent joints. That is what you have probably done using the other file you cited. After you are sure that the motors move the joints in the way you expect (in direction and angle and PID) you go forward.

At this point if the joints are independent, the task is over, otherwise if the joints are coupled you must add the coupling rule specific of your mechanical setup.

Let's see how to do that in steps:

1. Move joint 0 and 1 (or the motors) in independent mode.

   • You should change head-eb05-j0_1-mec.xml so that the COUPLING and JOINTSET_CFG groups are:

        <group name="COUPLINGS">
   		<!-- marco.accame: use identity matric for independent joints / motors  
             -->
           <param name="matrixJ2M">  
               1.00    0.00    0.00    0.00
               0.00    1.00    0.00    0.00
               0.00    0.00    1.00    0.00
               0.00    0.00    0.00    1.00 
           </param>
   
           <param name="matrixM2J">  
               1.00    0.00    0.00    0.00
               0.00    1.00    0.00    0.00
               0.00    0.00    1.00    0.00
               0.00    0.00    0.00    1.00 
           </param>
            
           <param name ="matrixE2J"> 
               1.00    0.00    0.00    0.00    0.00    0.00
               0.00    1.00    0.00    0.00    0.00    0.00
               0.00    0.00    1.00    0.00    0.00    0.00
               0.00    0.00    0.00    1.00    0.00    0.00
           </param>
   
       </group>
   
       <group name="JOINTSET_CFG"> 
   		<!-- marco.accame: use two jointsets for independent joints 
             -->        
           <param name= "numberofsets"> 2 </param>
           <group name="JOINTSET_0">
               <param name="listofjoints">  0             </param>
               <param name="constraint">    none          </param>
               <param name="param1">        0             </param>
               <param name="param2">        0             </param>
           </group>
           <group name="JOINTSET_1">
               <param name="listofjoints">  1             </param>
               <param name="constraint">    none          </param>
               <param name="param1">        0             </param>
               <param name="param2">        0             </param>
           </group>        
       </group>
   

   • keep or modify the following files according to the properties of your setup: head-eb05-j0_1-mc.xml, head-eb05-j0_1-mc_service.xml, the GENERAL and LIMITS groups of head-eb05-j0_1-mec.xml.

   • change head-calib.xml so that you have:

     <param name="CALIB_ORDER"> (0) (1) (2) (3) (4) </param> 

   • edit the general.xml so that you skip calibration:

     <param name="skipCalibration">    true  </param>

   • launch yarprobotinterface and then yarpmotorgui --calib

   • attempt calibration on joint 0 and then on joint 1.

   • if everything is OK you can also revert skipCalibration to be false again and restart yarprobotinterface.

2. move joint 0 and 1 (or the motors) in coupled mode. See in here an example of coupled joints: the pitch and roll of the neck of iCub.

   • find out the coupling rule between joint 0 and joint1 and add it inside head-eb05-j0_1-mec.xml.

   • in teh same file add one jointset only w/ both the joints.

   • calibrate teh two joints together:

     <param name="CALIB_ORDER"> (0 1) (2) (3) (4) </param> 

[plinioMoreno]

Hi @marcoaccame,

Thanks for the informative reply.

Let me summarize the mechanicals of our robot: Most of the joints (head, torso and arms) are independent joints. The only coupled joints are the two eyes, similarly to iCub (version and vergence coupled in control software) for which I should follow head-eb21-j2_5-mec.xml, is that right?

The following picture and table show more details about the mechanicals, where we have four serial chains of revolute joints: left arm, right, arm, left eye and right eye:

We were able to calibrate sucessfully the two joints! We changed the COUPLING and JOINTSET_CFG as you explained on our file:

    <group name="COUPLINGS">
		<!-- Check if this is right  -->
        <param name="matrixJ2M">  
            1.00    0.00    0.00    0.00
            0.00    1.00    0.00    0.00
            0.00    0.00    1.00    0.00
            0.00    0.00    0.00    1.00 
        </param>
		<!-- Check if this is right  -->
        <param name="matrixM2J">  
            1.00    0.00    0.00    0.00
            0.00    1.00    0.00    0.00
            0.00    0.00    1.00    0.00
            0.00    0.00    0.00    1.00 
        </param>
		<!-- Check if this is right  -->
        <param name ="matrixE2J"> 
            1.00    0.00    0.00    0.00    0.00    0.00
            0.00    1.00    0.00    0.00    0.00    0.00
            0.00    0.00    1.00    0.00    0.00    0.00
            0.00    0.00    0.00    1.00    0.00    0.00
        </param>

    </group>

    <group name="JOINTSET_CFG"> <!-- To be deleted  -->
        <param name= "numberofsets"> 2 </param>
        <group name="JOINTSET_0">
            <param name="listofjoints">  0           </param>
            <param name="constraint">    none          </param>
            <param name="param1">        0             </param>
            <param name="param2">        0             </param>
        </group>
        <group name="JOINTSET_1">
            <param name="listofjoints">  1           </param>
            <param name="constraint">    none          </param>
            <param name="param1">        0             </param>
            <param name="param2">        0             </param>
        </group>
    </group>

We are able to run the calibration either in sequence

<param name="CALIB_ORDER"> (0) (1) </param>

or at the same time on the two independent motors

<param name="CALIB_ORDER"> (0 1) </param>

We noticed that the startupPosThreshold had to be updated to 3.5 degrees for the calibration to finish properly (file) Do you know how to reduce this threshold?

[pattacini]

Hi @plinioMoreno

I should follow head-eb21-j2_5-mec.xml, is that right?

Right 👍🏻
More precisely, both the matrices: matrixJ2M and matrixM2J.

We noticed that the startupPosThreshold had to be updated to 3.5 degrees for the calibration to finish properly (file) Do you know how to reduce this threshold?

Maybe, @ale-git can give you some heads-up on the meaning and functionality of startupPosThreshold.

[plinioMoreno]

Hi @pattacini,

I saw that the matrices matrixJ2M and matrixM2J consider that the first joint is the right eye and the second one is the left eye as per commented here. How do those J2M and M2J matrices look like if the joints are in reversed order (first left eye and then right eye)?

[pattacini]

Hi @plinioMoreno,

The transformation between version|vergence to/from right|left eyes is computed according to https://icub-tech-iit.github.io/documentation/icub_kinematics/icub-vergence-version/icub-vergence-version.

The comment you referred to is a bit misleading as joints are only version and vergence and not the eyes, although the order of the eyes is correct.

If you invert the order of the eyes, then you have:

    <param name ="matrixM2J"> 
        1.000   0.000   0.000   0.000
        0.000   1.000   0.000   0.000
        0.000   0.000   0.500   0.500
        0.000   0.000   0.500  -0.500   
    </param>

matrixJ2M is merely the inverse of matrixM2J.

Also, be careful, as the numbers in the matrices should be fixed as per robotology/robots-configuration#228.

[ale-git]

The precision of the startup movement could depend on two factors:

• the precision of the control itself, and thus you should check the tuning of the position PID
• if reaching the startupPosition requires high torque maybe the startupMaxPwm is too low?