This is the source code to the 2021 IEEE ICDL paper "Human hand motion retargeting for dexterous robotic hand".
You can find 2 branches here:
- main - has python 3 support. This updated code includes an implementation with Mujoco physics simulator and pose optimization with particle swarm optimization (PSO). Source code was developed in 2019 and I cannot guarantee compatibility with current version of Mujoco 3.x.
- python2.7_vrep - old version of the code which directly relates to the paper. It support VRep simulator (now CoppeliaSim). I cannot guarantee compatibility with the current version of CoppeliaSim.
If this work is helpful, please consider using following entry for the citation:
@inproceedings{orbikInverseReinforcementLearning2021,
title = {Inverse Reinforcement Learning for Dexterous Hand Manipulation},
booktitle = {2021 {{IEEE International Conference}} on {{Development}} and {{Learning}} ({{ICDL}})},
author = {Orbik, Jedrzej and Agostini, Alejandro and Lee, Dongheui},
year = {2021},
month = aug,
pages = {1--7},
publisher = {IEEE},
address = {Beijing, China},
doi = {10.1109/ICDL49984.2021.9515637},
urldate = {2023-02-25},
isbn = {978-1-72816-242-3},
langid = {english}
}
This package works with Python 3. To use it with ROS Kinetic or newer:
cd ~/catkin_ws
rm -rf devel build
catkin build
source devel/setup.zsh # or setup.bash
cd src
git clone https://github.com/ros/geometry
git clone https://github.com/ros/geometry2
cd ..
rosdep install --from-paths src --ignore-src -y -r # not sure if needed
catkin build --cmake-args \
-DPYTHON_EXECUTABLE=/usr/bin/python3 \
-DPYTHON_INCLUDE_DIR=/usr/include/python3.6m \
-DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.6m.so
source devel/setup.zsh # or setup.bash
(source https://answers.ros.org/question/326226/importerror-dynamic-module-does-not-define-module-export-function-pyinit__tf2/ and ros/geometry2#259)