[Franka robot] franka_description fix weird inertia

[MaximilienNaveau]

No description provided.

[krishADUG]

the inertia values of inertial.yaml file for both main and humble-devel match.

[MedericFourmy]

To provide more context, these inertial paremeters are copied from this publication: https://inria.hal.science/hal-02265293

Reading through the method (section V), they explain that franka "dynamic parameters" (inertias, masses, CoM location), cannot be retrieved directly from their dynamical identification procedure. They are rather inferred from "dynamic coefficients" vector that can be retrieved uniqually and is lower dimensional. Then in a second step, dynamic parameters are obtained from dynamic coefficients by imposing box constraints (positive intertia matrix and masses, CoM in a predefined box, etc.). These constraint values can be found in this file: bounds_gM.csv. For instance masses are constrained to be between 0 and 10 kg.

Definition of the the function "dynamic parameters" -> "dynamic coefficients"

What I get from all of this is that any set of dynamical parameter (masses, etc.) that maps to the same dynamical coefficients lead to the same dynamical behavior of the robot (e.g. pin.aba will outputs the same accelerations). This is maybe why the values we have in inertial.yaml seem weird to us.

TLDR: Probably not an issue in my opinion

[MaximilienNaveau]

Maybe use the https://gitlab.laas.fr/gepetto/figaroh framework on Panda.