TFG Luis Miguel Lopez Martin

In this repository we will track the process of the autonomous driving of an amazon deepracer model robot around certain tracks in Gazebo using reinforcement learning.

Preview

Control with teleoperator

Autonomous discontinous line following

Autonomous white lines lane following

Running the project

Prerequisites

In order to run the project, firstly install Python for ROS:

sudo apt install python3-rosdep build-essential

Then, clone this repository:

git clone https://github.com/RoboticsLabURJC/2022-tfg-luismiguel-lopez.git

Once cloned, compile the workspace with

cd tfg_ws
catkin_make -DPYTHON_EXECUTABLE=/usr/bin/python3

Passing the intrepret to the compilation to ensure the correct usage of the nodes.

If any issue pops up while compiling, install the required dependencies using the command

rosdep install --from-paths src --ignore-src -r -y

And source it accordingly

source devel/setup.bash

Model setting

According to the launch properties, models should be automatically launched when you execute the comand in order to load the simulation world. If this does not happen, make sure to copy the models of the project into the .gazebo folder on your user home. In order to do that, use the following command in the root of the workspace:

cp -r src/track_model/models/ ~/.gazebo

Launch requisites:

Firstly, make sure you have followed all the package configuration steps on this readme

Track model: World & Robot start

In order to test out the functionalities, we need to firstly launch the world using

roslaunch track_model deepracer.launch

This will launch the world, the robot and the according topics for both the camera and the actuators control.

Track Model: nodes launch

Teleop

In order to execute a teleop operation, you must first launch the control node:

rosrun track_model control.py

Once the control node is launched, launch the teleop with the following command:

rosrun track_model teleop.py

Follow line

In the same way the teleop works, firstly we need to launch the control

rosrun track_model control.py

Once the control node is launched, launch the follow line node with the following command:

rosrun track_model follow_line.py

Track model: Launch via launchers

With the launchers, you can launch either teleop or visual approach line following:

roslaunch track_model follow_line_launcher.launch

roslaunch track_model teleop_launcher.launch

NOTE: if any of the launching fails it may be related with this issue. Install ROSpkg with this command:

pip3 install rospkg