franka_inria_inverse_dynamics_solver

Contents

This dynamics solver for the real Franka Emika (FER) robot is based on a model identified by the following paper:

C. Gaz, M. Cognetti, A. Oliva, P. Robuffo Giordano, and A. De Luca, “Dynamic Identification of the Franka Emika Panda Robot With Retrieval of Feasible Parameters Using Penalty-Based Optimization,” IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 4147-4154, Oct. 2019, doi: 10.1109/lra.2019.2931248.

The group at INRIA provided the library for the computation of the dynamic parameters, composed by four header files (one per function) and several source files with different implementations. On top of these, the library InverseDynamicsSolverFrankaInria has been developed, to include the following files:

• get_CoriolisMatrix.hpp: given the 7X1 vector of positions and the 7X1 vector of velocities, both in joint space, it computes the 7x7 matrix of torques related to centrifugal and Coriolis effects;
  • get_CoriolisMatrix_old.cpp: computes the Coriolis matrix by using temporary variables to store intermediate results of the computation;
    • This function is not actually needed by the library implementation, and is kept for legacy purposes only;
  • get_CoriolisMatrix.cpp: computes the Coriolis matrix directly, without using intermediate variables;
• get_FrictionTorque.hpp: given the 7X1 vector of velocities, it computes the 7x1 vector of torques due to friction;
• get_GravityVector.hpp: given the 7x1 vector of positions in joint space, it computes the 7x1 vector of torques due to gravity;
• get_MassMatrix.hpp: given the 7x1 vector of positions in joint space, it computes the 7x7 inertia matrix.
  • get_MassMatrix.cpp: computes the inertia matrix by using temporary variables to store intermediate results of the computation;
    • This function is not actually needed by the library implementation, and is kept for legacy purposes only;
  • get_MassMatrix_parziale.cpp: computes the inertia matrix directly, without using intermediate variables.

The dynamics solver implements the inverse_dynamics_solver::InverseDynamicsSolver class.

How to build

To build this package, run the following from the root of your colcon workspace:

colcon build --packages-up-to franka_inria_inverse_dynamics_solver
source install/setup.bash

Demo

You can evaluate the solver using the demo, currently configured in a launch file. The demo reads a bag file containing a sequence of sensor_msgs/msg/JointState messages and, for each state, computes the corresponding torques according to the InverseDynamicsSolverFrankaInria solver, which are saved in another bag file.

Run the demo

To launch the demo, run the following:

ros2 launch franka_inria_inverse_dynamics_solver evaluate_solver_franka.launch.py

By default, the demo reads the panda_exciting_trajectory.db3 bag file, and produces the output under the panda_exciting_trajectory_real_torques folder (created under the current working directory), with the computed torques written to the /torques topic. You can change this configuration with

ros2 launch franka_inria_inverse_dynamics_solver evaluate_solver_franka.launch.py input_bag:=<my_bag_file> output_bag:=<my_output_folder> topic:=<my_output_topic>

Visualize the results

Please refer to the parent class documentation to visualize the results, i.e. the evaluation of joint torque signals on input_bag, stored in output_bag.

Citation

If you find this work useful, please cite it as

@article{Gaz_2019,
    title={Dynamic Identification of the Franka Emika Panda Robot With Retrieval of Feasible Parameters Using Penalty-Based Optimization},
    volume={4},
    ISSN={2377-3774},
    url={http://dx.doi.org/10.1109/lra.2019.2931248},
    DOI={10.1109/lra.2019.2931248},
    number={4},
    journal={IEEE Robotics and Automation Letters},
    publisher={Institute of Electrical and Electronics Engineers (IEEE)},
    author={Gaz, Claudio and Cognetti, Marco and Oliva, Alexander and Robuffo Giordano, Paolo and De Luca, Alessandro},
    year={2019},
    month=oct,
    pages={4147–4154}
}

Changelog for package franka_inria_inverse_dynamics_solver

1.0.0 (2025-04-13)

• [REF] Rename .h files to .hpp files
• Merge pull request #2 from unisa-acg/humble Set repo URL, rename test launch files, document CO reproducibility
• [MAK] Add repo URL
• Merge pull request #1 from unisa-acg/humble [DOC] Add references for papers
• [DOC] Add references for paper in which the dynamic models have been identified
• Contributors: Enrico Ferrentino, Vincenzo Petrone

0.1.1 (2025-04-03)

• [MAK] Create CHANGELOG files
• [MAK] Commit for release
• Contributors: Vincenzo Petrone