This Repository contains the code used for training the manipulation policies in simulation, before deploying it to the real world using the ros package found here: https://github.com/eugval/sim2real_dynamics_robot
In order to run this project, the following dependencies need to be satisfied.
Mujoco : http://www.mujoco.org/
Openai gym : https://github.com/openai/gym
Openai mujoco-py : https://github.com/openai/mujoco-py
urdf_parser_py : https://github.com/ros/urdf_parser_py
PyKDL : If the pip version does not work, it will need to be compiled from scratch by following the instructions
at orocos/orocos_kinematics_dynamics#115
Robosuite (version 0.1.0): https://github.com/StanfordVL/robosuite
The following forked repositories: kdl_parser, the branch fixing_setup_py from https://github.com/eugval/kdl_parser/tree/fixing_setup_py simple-pid, the branch allow arrays from https://github.com/eugval/simple-pid/tree/allow_arrays
Simpler dependencies are listed (that can be recursively installed) in requirements.txt.
Each of the following need to be installed independently by navigating to the corresponding folder and running pip install -e .
.
robosuite-extra
implements extra environments and functionalities for the robosuite framework
sim2real-policies
implements the methods for training and evaluating the different sim2real polcies.
reality-calibration-characterisation
implements code to identify and characterise the different sources of error
when transferring from simulation to the real world for the given tasks, as well as scripts to optimise simulator
parameters.
Our environments are created using Robosuite (https://github.com/ARISE-Initiative/robosuite), part of which we modified in robosuite-extra to suite our project.
If you are using this code in your work, please also consider citing our paper:
@inproceedings{
valassakis2020crossing,
title={Crossing the Gap: A Deep Dive into Zero-Shot Sim-to-Real Transfer for Dynamics},
author={Valassakis, Eugene and Ding, Zihan and Johns, Edward},
booktitle={International Conference on Intelligent Robots and Systems (IROS)},
year={2020}
}