# epicon: coverage path guidance for aerial exploration that tries not to fly twice > A C++ package published in IEEE Robotics and Automation Letters in 2026, built inside a ROS workspace on Ubuntu with a bundled MARSIM simulator, where the problem being solved is not coverage itself but redundancy, and the evaluation baselines EPIC and SOP from FALCON are the frameworks the authors credit with cutting their engineering effort. **ZikangYuan/epicon** — [RA-L 2026] A LiDAR-based aerial exploration framework via coverage path guidance on point cloud maps - Repository: https://github.com/ZikangYuan/epicon - Stars: 54 · Forks: 0 - Language: C++ - License: GPL-3.0 - Published: 2026-08-27 · Updated: 2026-08-27 · Language: en - Canonical page: https://hysenlabs.com/projects/zikangyuan-epicon ## The problem is redundancy, not coverage The subtitle is the clearest statement of the contribution: a LiDAR-based aerial exploration package for avoiding revisitations. The introduction repeats it, describing the work as a package for avoiding revisitations on point cloud maps, and reports that simulation and real-world experiments validate that the framework effectively eliminates revisitation redundancy and reduces exploration time against state-of-the-art baselines. So the input is a point cloud map and the objective is to cover it without flying over the same ground twice, which is a different problem from building a coverage path in the first place. A video is linked from the header, and the paper is in IEEE Robotics and Automation Letters, volume 11, number 10, pages 11299 to 11306, dated 2026, with Zikang Yuan as first author and Xin Yang as corresponding author. ## A news timeline that dates the release precisely Four dated entries record how the project became available. On July 30, 2026 the work was accepted by RA-L 2026. On August 10 the source code was released. On August 15 it became early accessible in IEEE Xplore. On August 23 it was officially published. That sequence matters for a user because it means the repository post-dates the acceptance and that the published version is the one in the repository, with no separate pre-print to reconcile. The README also carries a citation request for anyone using the project in research, and the author list on the paper entry is slightly longer than the header list, adding the final author as Kwang-Ting Cheng. ## The tree is a simulator plus two planners At the top level there is a CMakeLists.txt, a LICENSE, a README, a doc directory, a MARSIM directory and two planner directories, global_planner and local_planner. That split is the architecture in one line: the simulator is vendored rather than treated as an external dependency, and the planning logic is separated into a global layer and a local layer. A vendored simulator is a deliberate trade in a research release, since it makes the experiments reproducible without pinning a separate simulator version, and at the cost of making the repository large to build. The licence is GPL-3.0, which is worth noting for anyone planning to reuse the planner code in a closed system. ## The build is four commands inside a catkin workspace The quick start assumes a ROS workspace layout and uses catkin: ```bash mkdir -p ~/EPICON/src cd EPICON/src git clone https://github.com/ZikangYuan/epicon.git cd.. catkin_make ``` The directory is created as a src folder inside a workspace named EPICON, the repository is cloned into it, and the build runs from the workspace root with catkin_make. The test environment is stated as Ubuntu 20.04, ROS Noetic and C++17, so the compiler standard and the distribution are fixed rather than flexible. There is no rosdep step documented and no list of system packages beyond that, which suggests the vendored simulator carries most of what the planners need. ## Maps are downloaded from two external projects, not shipped The datasets are the first thing that will stop a new user, because they are not in the repository. The instruction is to download simulation maps from the Google cloud addresses provided by EPIC and by EDEN, create the folder MARSIM/map_generator/resource if it does not exist, and move the downloaded maps into it: ```bash mkdir -p MARSIM/map_generator/resource mv /path/to/downloaded/maps/*.pcd MARSIM/map_generator/resource/ ``` The file extension in that move is pcd, so the expected format is point cloud data, which follows from the point cloud map framing. The destination path is inside the vendored simulator, which means the maps and the simulator are coupled: change where MARSIM reads from and the launcher will not find them. ## Running it is one roslaunch, and the map name is the argument Once the workspace is built and the maps are in place, the run command is: ```bash source /devel/setup.bash && roslaunch epicon_planner garage.launch ``` The package name in the launch call is epicon_planner, which is where the local planner directory in the tree ends up. The argument is a launch file named after a scenario, and the README says you can replace garage with other maps and names the three scenarios evaluated in the paper: cave, garage and city. That list is also the shortest way to know which maps to download, since a fourth map you find elsewhere has not been part of the published evaluation. ## The simulator needs a GPU, and the baselines are named One warning is prominent enough to quote in spirit: the simulation environment is developed based on the GPU version of MARSIM, so if you do not have a GPU you may need to make some necessary modifications to the simulator. That is the honest caveat of the release, and it applies to the experiments rather than to the planner. The credits section names three upstream projects with a role for each. EPIC is an efficient framework for fast UAV exploration and is the baseline. MARSIM is a lightweight point-realistic simulator for LiDAR-based UAVs, and the simulator here is built upon it. FALCON is another efficient framework, and SOP from it is evaluated in the large-scale scenarios. The acknowledgements thank Boyu Zhou and the STAR team for open-sourcing EPIC and FALCON, saying those works provided crucial insights and significantly reduced the engineering implementation effort. ## Conclusion epicon fits a robotics researcher working in ROS Noetic on UAV exploration who wants a released implementation and the maps its authors evaluated on, since the code, the simulator and the launch files are all in one repository. It is a poor fit if you have no GPU, because the bundled simulator is the GPU version of MARSIM and the README says you may have to modify it, and it is tied to a specific Ubuntu and ROS pair. Before you build it, read the citation request in the README, download the maps from the sources named rather than assuming they are included, and expect to compile a fairly large C++ tree with catkin. ## FAQ ### What is the epicon in this repository? It is not the supplement brand. Here EPICON is a LiDAR-based aerial exploration package for avoiding revisitations on point cloud maps, published in IEEE Robotics and Automation Letters in 2026 and implemented in C++ for ROS. ### How do I build and run epicon? Create a workspace src folder, clone the repository into it and run catkin_make, then source the setup file and roslaunch epicon_planner with one of the evaluated scenario launch files such as garage. The test environment is Ubuntu 20.04, ROS Noetic and C++17. ### Where do the epicon test maps come from? They are downloaded from the Google cloud addresses provided by the EPIC and EDEN projects, then moved into the MARSIM/map_generator/resource folder inside the vendored simulator. They are not included in the repository. ### Does epicon need a GPU? For the simulation, yes in practice. The environment is built on the GPU version of MARSIM, and the README says that without a GPU you may need to make some necessary modifications to the simulator. The planner itself is C++ code compiled through catkin. ## Sources - [Official README](https://github.com/ZikangYuan/epicon#readme) - [Project repository](https://github.com/ZikangYuan/epicon) - [README](https://github.com/ZikangYuan/epicon/blob/main/README.md) - [Releases](https://github.com/ZikangYuan/epicon/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/zikangyuan-epicon