EN | 中文

Overview of gym-gazebo

An OpenAI gym extension for using Gazebo known as gym-gazebo! This work can put gym environment with gazbeo, then you would like putting robot into gazebo with code applying gym. You can also visit the official github here gym-gazebo. If you use ROS2, the better way to you is visiting the newest version gym-gazebo2.

Summary

Because the official github which install version to Ubuntu16.04 has been deprecated, and the package in author's github has many question that has been closed without issues, Here I provide installation of mine and some hint to problem solving.

The original Installation of authors is here original and here INSTALL. If this installation wouldn't help you, visiting the author's github and submit a issue.

Prerequisites

ubuntu16.04
ROS-Kinetic (visit the official web here.)
Gazebo 7.14
openai-gym (visit gym github here.)
anaconda3 (install anaconda, click here.)
python=2.7 (with anaconda env below.)

The package you need to install:

numpy=1.16.2
matplotlib=2.2.3

If you want to train it with GPU here, you should install cuda

cuda=9.0
libcudnn7.3

Installation of conda env

Create an environment to run them.

conda create -n gymenv python=2.7
source activate gymenv

Install gym and gym-gazebo

# install gym
git clone https://github.com/openai/gym.git
cd gym
pip install -e .

# install gym-gazebo
git clone https://github.com/zhaolongkzz/gym_gazebo_kinetic
cd ~/gym_gazebo_kinetic
pip install -e .

Here some package you need to install first in your conda environment, if not, it will appear some errors when you run the code.

conda install numpy matplotlib

pip install rospkg catkin_pkg

pip install defusedxml

conda install scikit-image

pip install netifaces

Installation of ROS

First installing some ROS dependencies below:

sudo apt-get install \
cmake gcc g++ qt4-qmake libqt4-dev \
libusb-dev libftdi-dev \
ros-kinetic-octomap-msgs        \
ros-kinetic-joy                 \
ros-kinetic-geodesy             \
ros-kinetic-octomap-ros         \
ros-kinetic-control-toolbox     \
ros-kinetic-pluginlib	       \
ros-kinetic-trajectory-msgs     \
ros-kinetic-control-msgs	       \
ros-kinetic-std-srvs 	       \
ros-kinetic-nodelet	       \
ros-kinetic-urdf		       \
ros-kinetic-rviz		       \
ros-kinetic-kdl-conversions     \
ros-kinetic-eigen-conversions   \
ros-kinetic-tf2-sensor-msgs     \
ros-kinetic-pcl-ros \
ros-kinetic-navigation

Other dependencies and libraries:

# Failed to find libusb
sudo apt-get install libusb-dev
# ftdi.h: No such file or directory
sudo apt-get install libftdi-dev

sudo apt-get install ros-kinetic-sophus

sudo apt-get install ros-kinetic-ar-track-alvar-msgs

sudo apt-get install ros-kinetic-rqt-joint-trajectory-controller

Update your gazebo:

sudo sh -c 'echo "deb http://packages.osrfoundation.org/gazebo/ubuntu-stable `lsb_release -cs` main" > /etc/apt/sources.list.d/gazebo-stable.list'
wget http://packages.osrfoundation.org/gazebo.key -O - | sudo apt-key add -
sudo apt-get update
sudo apt-get install gazebo7

Quickstart

1.Compile all the packages

Note: All the command in ROS, I recommand executing in terminal without anaconda, this may lead some interference with your dependencies.

I have alter some github package or version in files, use gazebo_ros_kinetic.repos in my github here, please.

cd gym_gazebo_kinetic/envs/installation
bash setup_kinetic.bash

Put model file into your workspace/src folder.

cd gym_gazebo_kinetic/envs/installation
bash turtlebot_setup.bash

After first two steps above, you will find five lines being added in your ~/.bashrc:

source /home/zzl/gym-gazebo/gym_gazebo/envs/installation/gym_ws/devel/setup.bash
export GAZEBO_MODEL_PATH=/home/zzl/gym-gazebo/gym_gazebo/envs/installation/../assets/models
export GYM_GAZEBO_WORLD_MAZE=/home/zzl/gym-gazebo/gym_gazebo/envs/installation/../assets/worlds/maze.world
export GYM_GAZEBO_WORLD_CIRCUIT=/home/zzl/gym-gazebo/gym_gazebo/envs/installation/../assets/worlds/circuit.world
export GYM_GAZEBO_WORLD_CIRCUIT2=/home/zzl/gym-gazebo/gym_gazebo/envs/installation/../assets/worlds/circuit2.world
export GYM_GAZEBO_WORLD_CIRCUIT2C=/home/zzl/gym-gazebo/gym_gazebo/envs/installation/../assets/worlds/circuit2c.world
export GYM_GAZEBO_WORLD_ROUND=/home/zzl/gym-gazebo/gym_gazebo/envs/installation/../assets/worlds/round.world

Here open a new termianl, and uncomment your conda env in your .bashrc. Then using it with below:

source activate gymenv

cd gym_gazebo/examples/scripts_turtlebot
python circuit2_turtlebot_lidar_qlearn.py

2.Open gazebo

Open another Terminal:

cd gym-gazebo/gym_gazebo/envs/installation/
source turtlebot_setup.bash
# here the number is set in your code, default 12346
export GAZEBO_MASTER_URI=http://localhost:12346
gzclient

Display a graph showing the current reward history by running the following script:

cd examples/utilities
python display_plot.py

Picture

LICENCE

MIT License

FAQ

Q1: If encountering that ImportError: No module named msg like below:

Traceback (most recent call last):
  File "/home/zzl/gym_gazebo_kinetic/gym_gazebo/envs/installation/gym_ws/src/hector_gazebo/hector_gazebo_thermal_camera/cfg/GazeboRosThermalCamera.cfg", line 5, in <module>
    from driver_base.msg import SensorLevels
ImportError: No module named msg

A1: Run driver_base first, like that catkin_make -DCATKIN_WHITELIST_PACKAGES="driver_base".

Q2: If you encounter that ImportError: cannot import name _validate_lengths like below:

from numpy.lib.arraypad import _validate_lengths
ImportError: cannot import name _validate_lengths

A2: You'd better to upgrade your scikit-image, like that pip install --upgrade scikit-image. I test the version of 0.14.2 works well.

Q3: If encountering like below:

ERROR: cannot launch node of type [robot_state_publisher/robot_state_publisher]: robot_state_publisher

A3: Need to run sudo apt-get install ros-kinetic-robot-state-publisher

Q4: Here your protobuf is disturbed by conda or wrong version of your protobuf.

[libprotobuf FATAL google/protobuf/stubs/common.cc:61] This program requires version 3.6.1 of the Protocol Buffer runtime library, but the installed version is 2.6.1.  Please update your library.  If you compiled the program yourself, make sure that your headers are from the same version of Protocol Buffers as your link-time library.  (Version verification failed in "google/protobuf/any.pb.cc".)
terminate called after throwing an instance of 'google::protobuf::FatalException'
  what():  This program requires version 3.6.1 of the Protocol Buffer runtime library, but the installed version is 2.6.1.  Please update your library.  If you compiled the program yourself, make sure that your headers are from the same version of Protocol Buffers as your link-time library.  (Version verification failed in "google/protobuf/any.pb.cc".)
Aborted (core dumped)
[gazebo-1] process has died [pid 25049, exit code 134, cmd /home/zhaolong/gym_gazebo_kinetic/gym_gazebo/envs/installation/gym_ws/src/gazebo_ros_pkgs/gazebo_ros/scripts/gzserver --verbose -e ode /home/zhaolong/gym_gazebo_kinetic/gym_gazebo/envs/installation/../assets/worlds/circuit2.world __name:=gazebo __log:=/home/zhaolong/.ros/log/1593dbda-496c-11e9-8009-54bf648baa4d/gazebo-1.log].
log file: /home/zhaolong/.ros/log/1593dbda-496c-11e9-8009-54bf648baa4d/gazebo-1*.log

A4: conda install protobuf=3.5.2 in conda, here I recommand that version 2.6 is better for ubuntu16.04, so run all of the code in 2.6, please. Note that compile all the package with catkin_make without conda env, i.e. commenting the lines about conda. And after all the packages is compiled well, then open a new ternimal, and uncomment conda env, run python circuit2_turtlebot_lidar_qlearn.py

Q5: Lack of h files.

fatal error: spnav.h: No such file or directory
fatal error: cwiid.h: No such file or directory

A5:

sudo apt-get install libspnav-dev
sudo apt-get install libcwiid-dev

Q6:

/usr/include/gazebo-7/gazebo/msgs/msgs.hh:24:37: fatal error: ignition/math/Inertial.hh: No such file or directory

A6:

sudo apt-get install libignition-math2-dev