Report of on-site visit of 20-24/04/2015 to Sheffield

[pattacini]

This is a list of things deserving attention/improvement which we spotted while working with iCubSheffield01 during the WYSIWYD integration week held in Sheffield from 20th to 24th of April 2015.

Mechanics

• During ARE fingers calibration, the left thumb did not behave as expected, while the right thumb did calibrate nicely. The abduction of the left thumb (joint 8) does not move via the Gui, whereas by manually rotating the corresponding screw we managed to make it move. See The left arm must be helped during calibration #37. Endorsed.
• The right arm collides with the torso cover during the first calibration. Apparently, this might be due to a tiny deformation of the cover of the torso in the correspondence of the collision.

Software

• robotInterface never run after the first motor start-up. No matter how long we waited, some boards were not being heard of. Further, we had the same problem after a fault button event followed by robotInterface shutdown. In both the above situations, a second motor power-cycle made robotInterface start correctly. See Sporadic failures of robotInterface when the robot is just switched on #36. Endorsed.
• Frequent HW faults have been triggered throughout the week, in the neck and in the wrist.
• The ARE kinesthetic teaching did fail. The teaching went on in admittance mode and, looking into the code, this happens when the function setTorque() returns false. We tested it only once, but it's somehow weird that the admittance control took over the torque mode. Clarified.
• Reaching accuracy is low: see Reaching accuracy with ETH robots is low icub-main#140. Endorsed.
• There's the need for updating yarp binaries: see [ci]: need for updated yarp packages robotology-legacy/wysiwyd#12. Endorsed.

System configuration

The icubsrv laptop has been shipped back with the following wrong settings:

• The xml file used to launch the calibrated cameras was not the updated version; as result, the left and right image streams were swapped. This was quite a nasty bug to find, causing lots of problem with stereo vision. Fixed.
• Somehow, there were two bash files devoted to contain iCub related env variables: .bash_iCubrc and .bashrc_iCub, with mixed configuration. The standard file must be the former, while .bashrc was pointing to the latter, messing up the overall system.
  Fixed.

Good news

That's the sore points over, now the good news:

• Thanks to the new pronosupination, we have been experiencing an improved matching between the kinematic predictions and the stereo-vision observations in the depth2kin experiment.
• We managed to successfully integrate the stereo-vision and the IOL object-recognition into the wysiwyd system!

/cc @lukeboorman @davidetome @super-ste @marcoaccame @randaz81 @lornat75

[lornat75]

@mbrunettini @davidetome: check the issue with the configuration of the laptop:

Somehow, there were two bash files devoted to contain iCub related env variables: .bash_iCubrc and .bashrc_iCub, with mixed configuration. The standard file must be the former, while .bashrc was pointing to the latter, messing up the overall system.

Just to make sure we do not end up with the same problem in the future.

[pattacini]

@randaz81 reported that torque mode for iCubSheffield01 was purposely disabled, therefore ARE kinesthetic teaching was clearly not available.

[lukeboorman]

We are getting more frequent HW faults on neck joint 0 and I assume this then puts joint 1 into idle...

After startup (viewed with robotMotorGui) and during testing using the ball demo. I can send the robotInterface log file.

Could this be hardware?

[mbrunettini]

Hi,
the installation instructions here:
http://wiki.icub.org/wiki/Generic_iCub_machine_installation_instructions
reports only the correct .bashrc_iCub file, as Lorenzo told.
@davidetome : do you confirm you followed the above instructions? In case you’re using an HD image, can you chack it?

Thanks

Matteo