Autonomous Drones with ROS

- A group project by Milena Eisemann, Samyak Jain, Abhinav Utkarsh, and Fabienne Greier from TU - Munich

Table of Contents

1. Introduction
2. Video
3. Test Bench Hardware
4. Setup and Installation
5. Running the Project
6. Current Problems

---

Introduction

In the summer semester of 2023, we took a course at TU Munich called Introduction to ROS. As part of this course, we were assigned with a project that involved the following components (See the visual overview ):

1. Unity Simulation Environment: A base version was provided, which we were free to adjust and improve.

2. ROS-Simulation-Bridge: This feature served as the communication bridge between the simulation and other ROS nodes via TCP. Again, a base version was provided to us, but we had the flexibility to make any necessary adjustments.

3. Quadrotor Controller: This component enabled position control of the drone. As with the other parts, a basic version was supplied, but we could modify it as required.

4. State Machine: we had to manage the state machine for our drone, which controlled tasks such as takeoff, exploring, and landing at the specified location.

5. Perception Pipeline: This required us to convert the depth image, firstly to a point cloud and then to a voxel-grid representation of the environment.

6. Path Planner: We were tasked with creating a path planner that would generate a path through the environment to the goal location.

7. Trajectory Planner: Following the path planner, we also had to design a trajectory planner based on the path found.

The path- and trajectory-planning is implemented in 2D by collapsing all obstacles into the x-y plane.

One of the requirements of the project was to implement a custom message or service definition which we did for our state_machine. The detailed documentation can be found in the documentation directory.

Video

Watch our demo video below:

Test Bench Hardware

This project was implemented and tested on a system running Ubuntu 20.04 with ROS Noetic Full Desktop installed. The specific hardware specifications of the system (a Razer Blade 14 2021 model) are as follows:

• Processor : AMD Ryzen 9 5900X
• Processor Cores: 8
• Memory Size: 16 GB
• Memory Speed: 3200 MHz
• Graphics Card : RTX 3070

All the drone flight tests were performed on this system within a virtual environment running Ubuntu 20.04 with 6 cores and 12GB of Memory.

Setup and Installation

We were following the instruction doc: Ubuntu Installation Guide (https://docs.google.com/document/d/1HaGEXkqa_M8hBGSx2cSsKWAnSgEtfGKO1poNUynt7DE/edit).

Install ros-noetic-desktop-full as a base.

Then install the required ROS packages: single command :

sudo apt-get install python3-catkin-tools ros-noetic-costmap-2d ros-noetic-move-base-msgs ros-noetic-move-base ros-noetic-pointcloud-to-laserscan ros-noetic-explore-lite ros-noetic-octomap-server ros-noetic-rtabmap-ros ros-noetic-rotate-recovery ros-noetic-octomap-rviz-plugins

Finally, make sure the two drone files from the course instructor have the right permissions to execute in the 'devel/lib/simulation' folder.

Then finally run

catkin build

and

source devel/setup.bash

Running the Project

To start all the required processes including the Unity simulation and rviz, simply use the launchfile group4.launch in the src directory.

roslaunch src/group4.launch

6.2 Recording & Playback

We use Rosbag for recording and playback, but need to rename the bagfile to recording.bag in the workspace. Commands: play: roslaunch simulation playback.launch record: rosbag record -a -o recording.bag

Current Problems

• Lizard Problem
• Elastic controller maneuver
• Confiscated pointclouds on each other
• Sparse 2D scan
• Unexplored regions because of high inflation radius

Disclaimer: Since this is a course project, the files have been hidden to eliminate any possibility of malpractice from future course attendees.