This is the main repository for my master's thesis: "Vision-based safe autonomous UAV landing with panoramic sensors".

⚙ Dependencies

The repository was developed and tested with:

Ubuntu 20.04
ROS Noetic
Pytorch 1.11+cu113 (But newer versions of Pytorch should work as well)
PICAM360 panoramic camera module
PX4 (follow the instructions on the main documentations)

Prepare your workspace:

mkdir -p safelanding_ws/src
cd safelanding_ws/src
git clone https://github.com/phuoc101/detection_msgs.git
git clone https://github.com/phuoc101/yolov7_ros.git
git clone https://github.com/phuoc101/safe_landing.git
cd ..
catkin build

🚀 Run the software

# Source ROS environment
source <path_to_ros-ws>/devel/setup.bash #or .zsh, depending on your shell

The following commands run the program in Gazebo simulation. There are 2 options: hover and wait or adaptively select landing zone:

# Start PX4 simulation
roslaunch px4 roslaunch px4

## !!! CHOOSE ONE
# Adaptive emergency landing
roslaunch safe_landing offboard_live_sim.launch yolo_weights:=<your_weight> video_device:=/dev/video0 visualize_yolo:=true conf_thres:=0.5

# Adaptive emergency landing
roslaunch safe_landing offboard_live_sim_adaptive.launch yolo_weights:=<your_weight> video_device:=/dev/video0 visualize_yolo:=true conf_thres:=0.5

# The yolo image is published under topic /yolov7/image_raw
rosrun rqt_image_view rqt_image_view

🕹 Demo

Link to demonstration video

Citation

@article{nguyenVisionbasedSafeAutonomous2023,
  title = {Vision-Based Safe Autonomous {{UAV}} Docking with Panoramic Sensors},
  author = {Nguyen, Phuoc Thuan and Westerlund, Tomi and Pe{\~n}a Queralta, Jorge},
  year = {2023},
  journal = {Frontiers in Robotics and AI},
  volume = {10},
  issn = {2296-9144},
}