Week 9. Landing on the working infrastructure

1 minute read

This week we are going to install the necessary libraries to have the TensorFlow library installed as well as some examples from this repository.

To Do

The tasks proposed for this week are


Some of the robotics libraries that will be used during the development of this work are:

  • JdeRobot (open toolkit for developing Robotics applications).
  • Gazebo (simulator).
  • ROS (Robot Operating System).

With these tools we will have a part of the infrastructure where the learning algorithms will be trained later by “Reinforcement Learning”.

Install Docker images.

To install the docker you need the Dockerfile files from this repository with the command:

docker build -t cuda_cudnn_ros_melodic_python3 .


docker build -t ros_melodic_gazebo .

Replicate examples from the repository

As a first objective I wanted to launch a test world of Gazebo inside the container as well as the GzWeb client to be able to visualize it in the web browser (outside the container).

To do this:

  • Start the container:

    docker run -ti --rm -p 11311:11311 -p 11345:11345 -p 8080:8080 -p 7681:7681 ros_melodic_gazebo bash
  • Run the following instructions inside the container:

    roscore &
  • We clone and run a test world of the repository:

    git clone https://github.com/RoboticsLabURJC/2019-phd-alberto-martin
    cd 2019-phd-alberto-martin/gym-pyxis/gym_pyxis/envs/gazebo/assets/launch
    roslaunch turtlebot3_followline.launch
  • In another terminal (for simplicity) we run the Gazebo (web) client inside the container:

    cd gzweb
    npm start

The result is that the world can be seen in the browser as in the images:

Image week 9 - 1 Image week 9 - 2
Samples of Gazebo Worlds launched by ROS inside the Docker.



First approach to the environment where the work will be developed.