/IL_ROS_HSR

For performing various learning-based tasks with ROS-based robots.

Primary LanguagePython

IL_ROS_HSR

The goal of this package is to support various learning-based robotics tasks involving the HSR and the Fetch robots.

Right now, it supports the HSR and bed-making (as suggested by the package name), but we will eventually add more tasks and seamless support for the Fetch.

We also have code for training and deploying networks, using PyTorch 0.4.1.

Installation

Requirements: Ubuntu 16.04, ROS Kinetic, HSR package libraries.

  • Make a Python 2.7 virtualenv. We use --system-site-packages because we have ROS installed system-wide.

    virtualenv --system-site-packages --python=python2  <path_to_env_name>

  • Install HSR_CORE.

  • Install this package by running python setup.py develop, which ensures that your changes in this code are immediately reflected (i.e., no re-installation is required).

To get started connecting with the HSR, see Getting_Started.md.

Code Structure

Use main/ for any scripts to run for your experiments.

Use scripts/ for supporting scripts, such as for plotting results.

Within src/il_ros_hsr, there are three main sub-packages:

  • core: Contains core utilities across multiple projects. This will gradually be phased out in favor of HSR_CORE; it is only left here for backwards compatibility.
  • p_pi: Use for different application tasks, such as bed-making, each in their own sub-directories.
  • nets: For developing any Deep Learning code.

TODO

  • Get a requirements.txt file.
  • Phase out il_ros_hsr.core in favor of hsr_core.
  • Extend support for the Fetch robot and make package robot-agnostic.

History

This code was used for, among other things, the bed-making project, so some of the scripts and READMEs reflect that code usage. To get the exact code snapshot that was used for the paper, clone this other GitHub repository instead and set it up in your own virtualenv. That repository will not be updated further.