Skip to content

Latest commit

 

History

History
118 lines (79 loc) · 4.14 KB

install.md

File metadata and controls

118 lines (79 loc) · 4.14 KB

How to Install CoSTAR

Note: CoSTAR installation has only been tested on ROS Indigo (Ubuntu 14.04 LTS). For instructions on Indigo installation, please see here. There is a prototype install script available here that you can try out as well.

Installation

We provide two setup scripts to create most of your workspace:

NOTE THAT THESE SCRIPTS ARE THE PREFERRED WAY TO INSTALL THIS PACKAGE. They are extensively tested on travis and you should not see any problems.

The second option is Docker. This should create a container with a version of the CTP environment.

If none of these options work on your platform for whatever reason, you can always try a manual install.

Docker Install

Still in development. Simply run:

sudo docker run -ti alee156/costar

to get started.

Manual Install

The following are manual installation instructions for the CTP package and simulation. They are not guaranteed to be up to date; follow at your own risk.

Prerequisites

TTS can be installed either as a ROS catkin package or as an independent python package. Most features will work just fine if it is used without ROS.

  • To install TTS as a ROS package, just git clone it into your catkin workspace, build, re-source, and start running scripts.
  • To install TTS as an independent python package, use the setup.py file in the python directory.

To install the python packages on which TTS depends:

pip install h5py Theano pygame sympy matplotlib pygame gmr networkx dtw pypr gym numba pyyaml PyPNG

To use the CoSTAR plan system, you will need to install the following software packages:

  • Python (tested version 2.7.12)
  • Git (tested version 1.9.1)
  • ROS (tested ROS Indigo/Kinetic, Ubuntu 14.04 and 16.04 LTS)
  • Catkin Build Tools
  • OpenAI Gym -- mostly deprecated but important for acouple things
  • TensorFlow
  • Keras 2
  • Keras-RL
  • Bullet3

You can download all the required packages to use CoSTAR with ROS Indigo from the Ubuntu repositories with this command:

# set your ros distro 
export ROS_DISTRO=indigo

# install rosdep and catkin
sudo apt-get install -y python-catkin-pkg python-rosdep python-wstool python-catkin-tools ros-$ROS_DISTRO-catkin

# init your rosdep (if you have not already done so)
sudo rosdep init
rosdep update

Python Packages

For convert_jigsaws.py tool:

  • pip install opencv-python for video tools
  • pip install pillow for image resizing tools

Nvidia CudaNN

GPU setup for tensorflow is fairly easy, and you can find the instructions on the website. If it causes problems:

sudo pip uninstall tensorflow && sudo pip install tensorflow-gpu --upgrade

Then download the appropriate CudaNN libraries. For Ubuntu 14.04 and Tensforflow 1.4, these are easily installed with:

sudo dpkg -i libcudnn6_6.0.21-1+cuda8.0_amd64.deb 
sudo dpkg -i libcudnn6-dev_6.0.21-1+cuda8.0_amd64.deb

Step 1. Get CoSTAR Planning and Simulation Packages from Git

After this, you can download our various prerequisites from git. These packages are not qutonamtically installed from the package manager.

cd path/to/your/catkin_ws/src
# Contains larger binary files -- meshes, etc.
git clone gttps://github.com/jhu-lcsr/costar_objects.git

# Utility associated with costar_objects repo
git clone https://github.com/cburbridge/python_pcd.git

# Simple DMP library
git clone https://github.com/cpaxton/dmp.git --branch indigo

# Robotiq 85 gripper library
git clone https://github.com/cpaxton/robotiq_85_gripper.git

# Used to install ROS dependences from archive.
rosdep install -y --from-paths ./ --ignore-src --rosdistro $ROS_DISTRO

Step 2. Build catkin workspace

Change directory into catkin workspace folder and run:

catkin build