tag:ros-noetic

Python - Gazebo Simulation Environment for a UAV with Geometric Control

This repository includes Python codes for the position control a UAV in a Gazebo simulation environment, using geometric controllers.

Features

• Developed using Python
• Uses a geometric controller that works great with aggressive maneuvers
• Uses Gazebo as the physics engine
• Uses only minimal Gazebo/ROS codes so that if something breaks on ROS end in the future, it is easier to fix
• Has a nice GUI for controlling the UAV
• Can run your experiments, log data, and plot at the end of the trajectory at a click of a button
• Estimator, controller, and trajectory generators are in their own Python classes, if you need to test your own estimator, controller, or a trajectory, you only need to modify the respective class

Why Python?

• Python makes developing/debugging easier and shorter (no compiling)
• Can easily find modules or libraries for different tasks

Which controller is used for the UAV control?

• A geometric controller with decoupled-yaw attitude control is used
• The controller is published in:

  @InProceedings{Gamagedara2019b,
      title={Geometric controls of a quadrotor uav with decoupled yaw control},
      author={Gamagedara, Kanishke and Bisheban, Mahdis and Kaufman, Evan and Lee, Taeyoung},
      booktitle={2019 American Control Conference (ACC)},
      pages={3285--3290},
      year={2019},
      organization={IEEE}
  }
  

• Implementation of the same controller in C++ and Matlab can be found at https://github.com/fdcl-gwu/uav_geometric_control

How to test my own controller?

• Make sure the desired trajectory generates the desired states required by your controller.
• Simply update the Controller class with your controller.
• Make sure your modified controller class outputs the variable force-moment vector (fM).

Which estimator is used for the state estimation?

• The estimator defined in the following paper is implemented here (except the sensor bias estimation terms):

  @InProceedings{Gamagedara2019a,
      author    = {Kanishke Gamagedara and Taeyoung Lee and Murray R. Snyder},
      title     = {Real-time Kinematics {GPS} Based Telemetry System for Airborne Measurements of Ship Air Wake},
      booktitle = {{AIAA} Scitech 2019 Forum},
      year      = {2019},
      month     = {jan},
      publisher = {American Institute of Aeronautics and Astronautics},
      doi       = {10.2514/6.2019-2377}
  }
  

• Matlab implementation of the above controller can be found at https://github.com/fdcl-gwu/dkf-comparison.
• Note that the Matlab implementation has a delayed Kalman filter that has not been implemented here. Only the non-delayed parts inside DelayedKalmanFilter.m is utilized here.

How to test my own estimator?

• Make sure your estimator provides the states for the UAV control, or if not create a separate estimator using the current estimator as the base class.
• Update the Estimator class with your estimator.

How to test my own trajectory?

• Add your trajectory generation function in Trajectory class.
• Replace that with an unused mode in the calculate_desired function inside the Trajectory class.

Setting-up

:bangbang: If you are running this on a virtual machine, please make sure that Gazebo can run at real-time speed. It is known that this simulation exhibits unintended behavior if the "real-time factor" of the Gazebo simulation is not closer to 1.0 (See issue#3).

Setting-up the repository

1. Clone the repository.

   git clone https://github.com/fdcl-gwu/uav_simulator.git
   

2. Update the submodules.

   cd uav_simulator
   git submodule update --init --recursive
   

Dependencies

You have to options here:

1. Installing everything locally
2. Running a docker container

Installing everything locally is probably the fastest way, but you may have to instal dependencies manually, or may have to deal with package version changes. Docker solves this by streamlining all the dependencies, up-to the OS. For example, if you are on Ubuntu 22.04 and want to test the ROS-Melodic version, docker will be the only way.

If you want to install everything locally, follow Local Install. If you want to run a docker container instead, skip to Docker Setup.

Local Install

1. ROS: this repository has been developed using ROS Noetic, on Ubuntu 20.04. If you are on ROS Melodic with Ubuntu 18.04, please checkout ros-melodic branch before installing dependencies.
2. Python GTK libraries for GUI (not required if you opt to not to use the GUI)

   sudo apt-get install python3-pip python3-gi
   

3. Python modules: these libraries must be installed in the system
   1. NumPy
   2. Pandas
   3. Matplotlib

   python3 -m pip install numpy pandas matplotlib
   

Now, skip to Setting-up the plugins and Gazebo.

Docker Setup

The instructions here assume you are on Ubuntu. This has not been tested on other OS versions.

1. Install docker following official instructions.
2. If you are not already there, cd uav_simulator
3. Enable xhost (required for Gazebo and GUI): xhost +
4. Build the docker image: docker build -t uav_simulator . (see following paragraph if you just want to pull the built image instead)
5. Run a container: bash docker_run.sh

The last command will start a docker container, install all the dependencies, and mount the local directory there. The first time you run the build command will take a while as it installs all the libraries.

You can skip the build command altogether by pulling the built docker from the Docker Hub with the following command. This is NOT required if you are building it locally using the build command.

docker pull kanishgama/uav_simulator:ros-noetic
bash docker_run.sh

After that, you only need to run the bash docker_run.sh every time you need to run the simulation. Since this mounts the local repository inside the docker, you just need to change the code in your local repository, and it will be automatically update inside the docker.

For running the code, simply follow Setting-up the plugins and Gazebo, and onwards.

Setting-up the plugins and Gazebo

You only need to do the followings once (unless you change the Gazebo plugins)

1. Make the plugging.

   # From uav_simulator
   catkin_make
   

2. Source the relevant directories (NOTE: you need to do this on every new terminal).

   # From uav_simulator
   cd devel && source setup.bash && cd ../
   

Running the simulation environment

1. In the current terminal window, launch the Gazebo environment:

   # From uav_simulator
   roslaunch uav_gazebo simple_world.launch 
   

2. Once the Gazebo is launched, run the rover code from a different rover terminal (if you already don't know, you may find tmux a life-saver):

   # From uav_simulator/scripts
   python3 main.py
   

   If you change the Python code, simply re-run the Python code. The code has been tested with Python3.8.10, which comes default with Ubuntu 20.04.

Tips

1. Every time you change the simulation environment, you have to kill the program, catkin_make and re-run it.
2. If you do not make any changes to the simulation environment, you only need to kill the Python program.
3. The UAV will re-spawn at the position and orientation defined in reset_uav() in rover.py when you run the Python code.

Control Guide

• Simply click on the buttons on the GUI to control the UAV.
• You can easily switch between each trajectory mode simply clicking on the radio buttons.
• Stay mode simply commands the UAV to stay at the current position.
• When take-off, stay, and circle trajectories end, the UAV switched to the "manual" mode.
• When the UAV is in manual, you can use following keys (these are not case sensitive):
  • WASD: to move in horizontal plane
  • P: increase altitude
  • L: decrease altitude
  • Q: yaw rotation in anti-clockwise direction
  • E: yaw rotation in clockwise direction
• At any point of the flight, you can use following keys (these are not case sensitive):
  • M: kill motors
  • 0-5: set the flight mode without clicking on the GUI
• Please not that the GUI must be in focus for any of the above keys to work.
• If you want to change the above keyboard shortcuts, you can do so by editing on_key_press function in gui.py.

Running Unit-Tests

• Make sure you are in the main directory.
• Run python -m unittest.
• Unit tests have only been tested on Python 3.9.
• Currently, unit test only covers the matrix_utils.py module.