Request on DROID end effector config update

[Everloom-129]

Dear DROID Team,

Following the instructions from the DROID, we initially uploaded the standard configuration file for the Robotiq gripper.

Updating Inertia Parameters for Robotiq Gripper
In order to accurately represent the robot model with the Robotiq gripper we need to update the end-effector configuration. From the Desk UI you may access this option under Settings > End-Effector as outlined here.

Please add the following configuration file via the UI.

Issue:
Original endeffector-config.json:

{"mass":0.9,"centerOfMass":[0,0,0.057],"transformation":[1,0,0,0,0,1,0,0,0,0,1,0,0,0,0,1],"inertia":[0.002768,0,0,0,0.003149,0,0,0,0.000564]}

However, since we have mounted a ZED Mini camera to the robot wrist. This additional hardware shifts the CoM and alters the inertia tensor significantly away from the standard Robotiq configuration.

Because of this mismatch between the configured physics model and the actual payload, which brings approx. 20% RMSE trajectory tracking errors

Planned: (100, 7), Actual states: (100, 7)

=== Planned vs Actual Robot States (100 steps) ===
Position error (m):  mean=0.0856, max=0.6484
Rotation error (rad): mean=0.3175, max=2.2570
Gripper error:       mean=0.0699, max=0.3329
Total RMSE: 0.2207

My Questions:

1. Does the Franka Desk or FCI API provide a Load Identification utility to automatically estimate the Mass, CoM, and Inertia Tensor of a custom payload?
2. If manual calculation is required, could you provide best practices for calculating the Inertia Tensor (, etc.) for a composite end-effector?
3. Could these high tracking errors (max 0.64m) be caused solely by the payload mismatch, or should we also be looking into impedance/stiffness parameters in our controller?

Any guidance on properly configuring this custom payload to reduce tracking error would be greatly appreciated.

Best regards,

Jie Wang

GRASP Lab