Aerial Exploration on Point Cloud Maps via Coverage Path Guidance

Zikang Yuan, Yuan Ren, Yixue Wang, Leyi Zhao, Shangzhe Sun, Chi Chen, Lijun Zhu, Xin Yang^† and Tim Cheng

^†Corresponding Author

• [2026.08.10] The source code of this work is released !
• [2026.07.30] This work is accepted by RA-L 2026 🚀 !

We sincerely thank Prof. Boyu Zhou and his team STAR for open-sourcing the excellent works EPIC and FALCON, which provided crucial insights for this work and significantly reduced the engineering implementation effort.

This work is a LiDAR-based aerial exploration package for avoiding revisitations on point cloud maps. Simulation and real-world experiments validate that our framework effectively eliminates revisitation redundancy and reduces exploration time against state-of-the-art baselines. (Click the image to view the video)

• Ubuntu 20.04
• ROS Noetic
• C++17

mkdir -p ~/EPICON/src
cd EPICON/src
git clone https://github.com/ZikangYuan/epicon.git
cd..
catkin_make

Download simulation maps from Google cloud address provided by EPIC and EDEN, create the folder MARSIM/map_generator/resource if it doesn't exist, and move the downloaded maps to this folder.

mkdir -p MARSIM/map_generator/resource
mv /path/to/downloaded/maps/*.pcd MARSIM/map_generator/resource/

source /devel/setup.bash && roslaunch epicon_planner garage.launch

You can replace garage with other maps. Three test scenarios are evaluated in our paper: cave, garage and city.

Our simulation environment is developed based on the GPU version of MARSIM. So if you don't have a GPU, you may need to make some necessary modifications to the simulator.

We would like to express our gratitude to the following projects, which have provided significant support and inspiration for our work:

• EPIC: An efficient framework for fast UAV exploration from which is our baseline.
• MARSIM: A lightweight point-realistic simulator for LiDAR-based UAVs upon which our simulator is built.
• FALCON: An efficient framework for fast UAV exploration, from which we evaluate SOP in large-scale scenarios.