Performance Tuning for Reliable Grasping While Walking and Determination of Object Manipulation Workspace.

[vigisushrutha23]

At the moment there is a bottleneck for the grasping performance due the issue robotology/icub-tech-support#1641. This prevents us from performing any meaningful bimanual experiments. Once this is solved, we have the following tasks.

• Ensure that for at least one cardboard the real robot has consistent grasping and manipulation capabilities.
• Ensure that the same is repeatable while walking.
• Obtain plots of footsteps, arm joint extremes to quantitatively analyse the shift in robot foorptint while moving the object This should give us a rough possible workspace for the object manipulation with real object.
• Perform initial experiments with just Walking+Bimanual in the area behind the desk in robot arena with manual manipulation commands.

[mebbaid]

@vigisushrutha23 Today I did some work during my slot on the shoulder joints tracking. The observation is as follows

• Low level tracking in stiff mode is great (e.g. l_shoulder_roll plot below).
  

• When we invoke the walking to control the shoulders using commands comming from the bi-manual manipulation, we change the mode of the joint to complient hence the poor tracking.

• The fix is very simple, we need simply (in our fork/branch of the walking) to change those lines https://github.com/SimoneMic/walking-controllers/blob/3c594f6bbbfbf04a7dac6c674697564ee4ced195/src/WalkingModule/app/robots/ergoCubSN000/dcm_walking/iFeel_joint_retargeting/robotControl.ini#L25-L26

and those lines
https://github.com/SimoneMic/walking-controllers/blob/3c594f6bbbfbf04a7dac6c674697564ee4ced195/src/WalkingModule/app/robots/ergoCubSN000/dcm_walking/iFeel_joint_retargeting/robotControl.ini#L33-L34

to all read true.

[S-Dafarra]

• When we invoke the walking to control the shoulders using commands comming from the bi-manual manipulation, we change the mode of the joint to complient hence the poor tracking.

In this case, the culprit is robotology/icub-tech-support#1547

[vigisushrutha23]

We were able to reliably grasp and manipulate a soft lunch bag while walking as seen in the video below. We The object roll motions seemed reliable and repeatable and will be preferred in the upcoming tests. We also tested the full forward motion in the end where the robot lifts the object and extends the hand completely and that is effective as well.
https://github.com/hsp-iit/project-ergocub/assets/5215434/df32326c-10c5-47fc-b310-1758c7c0621a

[vigisushrutha23]

Closing Old issues