\ Gazebo/ROS based virtual environment to reinforcement learning research on USVs.
First of all you should have ROS Noetic and Gazebo 11 install in your system. For installation instructions see http://wiki.ros.org/noetic.
Once you have ROS and Gazebo installed, to make the environment run you will need to do:
1 - Create a workspace for the project
$
mkdir -p ~/eboat_ws/src
2 - Change to the src directory
$
cd ~/eboat_ws/src
3 - Clone de repository
$
git clone https://github.com/medialab-fboat/eboat_gz_1.git
4 - Change to the workspace directory
$
cd ~/eboat_ws
5 - Compile the source
$
catkin_make
6 - Source de ROS package
$
source ~/eboat_ws/devel/setup.bash
6.1 - Alternatively, you can insert the command above in your .bashrc file
$
echo "source ${HOME}/eboat_ws/devel/setup.bash" > ${HOME}/.bashrc
7 - Launch the environmen
$
roslaunch eboat_gazebo ocean.launch
\ To launch the GUI with boat controls, you sould do:
1 - Install miniconda from https://docs.conda.io/en/latest/miniconda.html
2 - Create an anaconda environment using the YML file in ~/eboat_ws/src/eboat_gz_1/eboat_control/scripts/
$
conda env create --file ${HOME}/eboat_ws/src/eboat_gz_1/eboat_control/python/script/ctr_gui.yml
3 - Make the script executable
$
chmod +x ~/eboat_ws/src/eboat_gz_1/eboat_control/scripts/controls_v0.py
4 - Change the path on the python script to point your ctrgui anaconda environment
5 - Launch eboat_gazebo (If you did not follow installation step 6.1, you might need to execute installation step 6 before starting the environment)
$
roslaunch eboat_gazebo ocean.launch
6 - Run the python script
$
~/eboat_ws/src/eboat_gz_1/eboat_control/scripts/controls.py
\
OBS: Alternatively, if you do not wish to make the python code executable, you can replace steps 3 to 6 with the following steps.
3' - Activate the environment
$
conda activate ctrgui
4' - Launch eboat_gazebo
$
roslaunch eboat_gazebo ocean.launch
5' - Run the python script
$
python ~/eboat_ws/src/eboat_gz_1/eboat_control/scripts/controls.py
\ To make Gym environment avaiable for RL developing, you must do:
1 - Install miniconda from https://docs.conda.io/en/latest/miniconda.html
2 - Create an anaconda environment using the YML file in ~/eboat_ws/src/eboat_gz_1/
$
conda env create --file ${HOME}/eboat_ws/src/eboat_gz_1/gazgym.yml
3 - Activate the environment
$
conda activate gazgym
4 - Clone the GazeboGym environment from GitHub in your home folder
$
git clone https://github.com/erlerobot/gym-gazebo.git
5 - Got to the GazeboGym home
$
cd ~/gym-gazebo
6 - Install the environment
pip install -e .
\
This project contains some gym environments predifined. You can use them or create your own environments for your RL tasks. The environments are stored in ~/eboat_ws/src/eboat_gz_1/eboat_gym_gaz/envs/
\ * The original YML file will install tensorflow-gpu and pythorch-gpu. In case you do not have a compatible GPU available, remove tensorflow-gpu and pythorch-gpu from the YML file.
\ Frequently used commands:
Closes everything:
$
sudo killall -9 gazebo gzserver gzclient python rosmaster rosout
Open Tensorboard:
$
cd ~/eboat_ws/src/eboat_gz_1$tensorboard --logdir logs