trajectory_tracking

This is a ROS application to test trajectory tracking algorithms for mobile robots using Gazebo as simulator. The robot that has been used for this project is Turtlebot 2 which is an affordable mobile robot widely used for research.

Algorithms that have been used and are working properly so far:

• Numerical method controller using Euler's approximation.
• PID controller

These algorithms have been tested, and are properly working for the following trajectories:

Trajectory	Euler Controller	PID Controller
linear	✅	✅
circular	✅	❌
squared	✅	❌
lemniscate	✅	❌
epitrochoid	✅	❌
lissajous	✅	❌

Installation

Before cloning this repository, you should install ROS and Gazebo. It is highly recommended to use Ubuntu as OS. I also suggest installing Gazebo before ROS in order to avoid Gazebo version conflicts.

Installing Gazebo

In order to install Gazebo, you can follow the instructions provided in this tutorial, or you can just execute this command line:

$ curl -ssL http://get.gazebosim.org | sh

Installing ROS

The ROS version used in this project is kinect. In order to install ROS, you can follow the instruction in this tutorial, or you can execute the following command lines:

$ sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu $(lsb_release -sc) main" > /etc/apt/sources.list.d/ros-latest.list'
$ sudo apt-key adv --keyserver hkp://ha.pool.sks-keyservers.net:80 --recv-key 421C365BD9FF1F717815A3895523BAEEB01FA116
$ sudo apt-get update
$ sudo apt-get install ros-kinetic-desktop-full
$ sudo rosdep init
$ rosdep update

Finally you should source the enviroment variables to avoid doing it every time you want to use ROS:

$ echo "source /opt/ros/kinetic/setup.bash" >> ~/.bashrc
$ source ~/.bashrc

You could optionally install rosinstall in order to download packages easily:

$ sudo apt-get install python-rosinstall

Installing Turtlebot package

Once ROS and Gazebo have been installed, you have to install the Turtlebot package. Run the following command line:

$ sudo apt-get install ros-kinetic-turtlebot-gazebo

In order to make sure that the installation process was successful, execute the following command line, which will open a Gazebo world with some objects and a Turtlebot between them:

$ roslaunch turtlebot_gazebo turtlebot_world.launch

In case you get an error when executing the previous command, just restart your computer and try again.

Creating a workspace

Now, it is time to create a workspace, so you can create it in your home directory by executing the following commands, in this case the workspace will be named turtlebot_ws

$ mkdir -p ~/turtlebot_ws/src
$ cd ~/turtlebot_ws/src/
$ catkin_init_workspace
$ cd ~/turtlebot_ws/
$ catkin_make

Once the workspace has been created source it:

$ source ~/turtlebot_ws/devel/setup.bash

Cloning repository

Now it is possible to clone this repository. Because this repository is a ROS package, it should be cloned inside ~/turtlebot_ws/src/:

$ cd ~/turtlebot_ws/src/
$ git clone https://github.com/bit0001/trajectory_tracking.git

After cloning the repo, you have to run catkin_make again:

$ cd ~/turtlebot/
$ catkin_make

Finally, this package is ready to be used.

Usage

Plotting a trajectory

It is possible to visualize the trajectories that the mobile robot can follow:

$ cd ~/turtlebot_ws/src/trajectory_tracking/src/
$ python -m trajectory

This will run a console application in which one can list all available trajectories, and one can also plot a trajectory.

Running a simulation

Open a terminal window, and you have to source the workspace, and execute a launch file in order to initialize Gazebo:

$ source ~/turtlebot_ws/devel/setup.bash
$ roslaunch trajectory_tracking turtlebot_world.launch

This will open Gazebo in a world where a turtlebot is shown in the middle of a room. Now, open a new terminal, source again the workspace, and run the file trajectory_tracking twist.py:

$ source ~/turtlebot_ws/devel/setup.bash
$ rosrun trajectory_tracking twist.py cmd_vel:=cmd_vel_mux/input/teleop

Finally, open again a new terminal, source the workspace and run the file control.py:

$ source ~/turtlebot_ws/devel/setup.bash
$ rosrun trajectory_tracking control.py <controller_name> <trajectory_name> [simulation_time]

Where controller_name could be either euler or pid, and trajectory_name could be either linear, circular, or squared. The simulation_time is a positive number, and if it is not given, a default time is used. Example:

$ rosrun trajectory_tracking control.py euler squared 60.0

The previous command uses the the euler method controller to follow a squared trajectory during 60 seconds. Once an experiment is completed, results are plotted and shown, and when plot windows are closed, simulation data is stored in a database, which is stored in the root directory of this project, creating a table which name follows the following format: controller_trajectory_YYYY_mm_dd_HH_MM_SS.

Plotting results of the last simulation

Once a simulation has been completed, it is possible to plot again the results obtained in that simulation. In order to achive this, one has to run the module plotter and pass the absolute path to the database results.db as an argument of the command-line:

$ cd ~/turtlebot_ws/src/trajectory_tracking/src/
$ python -m plotter /absolute/path/to/database/results.db

Plotting results of a simulation specifying its name

It is also possible to plot again the results of a simulation by specifying its name:

$ cd ~/turtlebot_ws/src/trajectory_tracking/src/
$ python -m plotter /absolute/path/to/database/results.db simulation_name

In order to see the list of simulation names use the --sims flag:

$ cd ~/turtlebot_ws/src/trajectory_tracking/src/
$ python -m plotter /absolute/path/to/database/results.db --sims

Comparing the results of two simulations

It is possible to compare the obtained results in two different simulations. There is one mandatory requirement that is comparing two simulations of the same trajectory, i.e., the controller can be different.

Although at first glance it seems that the time that both simulations lasted should be the same, it is possible to force a comparison. When a comparison is forced, the smallest time of the two simulations is taken, and the other simulation is limited to the smallest time.

To compare two simulations use:

$ cd ~/turtlebot_ws/src/trajectory_tracking/src/
$ python -m plotter /absolute/path/to/database/results.db sim1_name sim2_name

To force the plot comparison of two simulations use:

$ cd ~/turtlebot_ws/src/trajectory_tracking/src/
$ python -m plotter /absolute/path/to/database/results.db sim1_name sim2_name --f