This repository is used for UAV dynamic small obstacles avoidance. It is a complete system for lidar-based UAV, including FAST-LIO slam, time-accumulated KD-Tree mapping and kinodynamic A* search modules. It is able to avoid dynamic small obstacles (down to 20mm diameter bars) by running at 50Hz.
Related paper: "Avoiding dynamic small obstacles with onboard sensing and computating on aerial robots", available on arxiv now https://arxiv.org/abs/2103.00406.
Related video: https://youtu.be/pBHbQ_J1Qhc
As the same as the prerequisites as FAST-LIO.
Ubuntu >= 18.04.
ROS >= Melodic. ROS Installation
PCL >= 1.8, Follow PCL Installation.
Eigen >= 3.3.4, Follow Eigen Installation.
Clone the repository and catkin_make:
cd ~/catkin_ws/src
git clone https://github.com/hku-mars/dyn_small_obs_avoidance.git
cd ..
catkin_make
source devel/setup.bash
roslaunch path_planning demo.launch
You can download our demos rosbag from https://drive.google.com/drive/folders/1knQwnrbwunGIvXzOL6wWKkCtOkGE22bG?usp=sharing And play the bag by:
rosbag play XXX.bag
rostopic pub /goal geometry_msgs/PoseStamped '{header: {stamp: now, frame_id: "slam_map"}, pose: {position: {x: 5.0, y: 0.0, z: 1.0}, orientation: {w: 1.0}}}'
You can change the target point by setting different value of 'x','y','z' in the above command.
You can also run this planning program as a module with other point cloud input from other sensors (like depth camera D435i). But we have not yet test it on depth cameras. First, change the point cloud topic to your point cloud topic name in path_planning/launch/demo_withNOlidar.launch
<remap from='/your_pointcloud_topic' to='/cloud_registered'/>
Then, start the program:
source devel/setup.bash
roslaunch path_planning demo_withNOlidar.launch
Thanks for FAST-PLANNER(Zhou, Boyu and Gao, Fei and Wang, Luqi and Liu, Chuhao and Shen, Shaojie. Robust and efficient quadrotor trajectory generation for fast autonomous flight), FAST-PLANNER.