VITAL

Deep Learning Based Vision Pipeline for Near Real-Time Autonomous UAV Landing Support With Added Robustness

What is VITAL?

This repository includes an image processing pipeline for accurate landing pad detection and relative pose estimation of the UAV w.r.t. the landing pad, while also providing a human detection in local environment.

Highlights

• ROS1/ROS2 compatible.

• Deep Learning based pipeline.

• Real-time performance.

• Modular - the system is modular and can be easily extended with new functionalities. It is easly adaptable to different cameras and landing targets.

• Safety - the system is able to detect the presence of a human near the landing site.

• Autonomous landing - the system is able to provide information about the target's position continuously, which allows to build a pipeline for landing on moving targets.

• Precision landing - the system is able to detect landing target with a precision of 1 cm verified using Opti-Track.

Prerequisites

• ROS Melodic/Noetic or ROS2. Pipeline was tested on Ubuntu 18.04 (NVIDIA Jetson Xavier NX) with ROS Melodic and Ubuntu 20.04 (PC) with ROS Noetic and ROS2 Foxy.
• MAVROS. MAVROS is a ROS-Node that allows to communicate with the autopilot. It is used to get the altitude of the UAV.
• requirements.txt - contains all the required python packages. To install them run python3 -m pip install -r requirements.txt in the virtual environment.

Installation

ROS Melodic/Noetic

• cd catkin_ws/src

• git clone https://github.com/PUTvision/VITAL.git

• cd ../

• catkin build

ROS2 Foxy

• cd ros2_ws/src

• git clone https://github.com/PUTvision/VITAL.git

• cd ../

• colcon build

Config documentation

All configuration parameters from config.yaml are described in config docs.

Usage

ROS Melodic/Noetic

• source devel/setup.bash

• roslaunch visual_landing_provider visual_infer.launch

ROS2 Foxy

• source install/setup.bash

• ros2 launch visual_landing_provider visual_infer.launch.py

For developers

If you:

• want to understand how the system works
• want to build your own models
• want to adapt the pipeline to your needs

then you should read the developer docs.

Citation

If you use this code for your research, please cite our paper:

(SOON)

Acknowledge

We acknowledge the support of the TERRINet project and would like to express our gratitude to the GRVC team at Sevilla for their hospitality and help. Read more here.