Dynamic graph head implementation of classical MPC (rigid contact), torque-feedback MPC (low-pass actuation model) and force-feedback MPC (soft contact model) for the KUKA iiwa LBR 14. This contains the code necessary to reproduce the hardware experiments results described in our paper submission :
- S. Kleff, et. al, "Force Feedback in Model-Predictive Control: A Soft Contact Approach" PDF (under review)
Demo script imports the OCP setup in force_feedback_mpc repo and runs the MPC either in a real-time simulation in PyBullet using SimHead or on the real robot.
Each controller's parameters (i.e., OCP definition, costs, constraints, discretization, and MPC parameters : frequency, solver settings, etc.) are available in the config directory.
The controller's scripts are available in the controllers directory.
The dirs plot contains utilities to visualize experimental data, and the data used in the paper figures and videos is available in data.
@unpublished{kleff:hal-04572399,
TITLE = {{Force Feedback in Model-Predictive Control: A Soft Contact Approach}},
AUTHOR = {Kleff, S{\'e}bastien and Jordana, Armand and Khorrambakht, Rooholla and Mansard, Nicolas and Righetti, Ludovic},
URL = {https://hal.science/hal-04572399},
NOTE = {working paper or preprint},
HAL_LOCAL_REFERENCE = {Rapport LAAS n{\textdegree} 24093},
YEAR = {2025},
MONTH = Jun,
PDF = {https://hal.science/hal-04572399v2/file/force_feedback_article_second_submission.pdf},
HAL_ID = {hal-04572399},
HAL_VERSION = {v2},
}