Open REALM: Real-time Aerial Localization and Mapping

This is the repository for Open REALM, a real-time aerial mapping framework. It is currently in alpha state, so don't expect a polished and bugfree application.

Feel free to fork and contribute. Let's make mapping fast again! :)


Demonstration Video


The proposed framework stands on the shoulders of giants, namely the open source implementation for visual SLAM and stereo reconstruction. Please read the references below.

For a detailed description of the underlying ideas refer to my thesis:


The pipeline is designed for multirotor systems with downlooking 2-axis gimbal stabilized camera and alignment of the magnetic heading of the UAV with the image's negative y-axis. It should be easily extendable to other cases in the future, but at the moment I won't recommend to use it on any other.

What it can do

What NOT to expect


OS ROS Distribution Build Status
Ubuntu 16.04 ROS Kinetic Build Status
Ubuntu 18.04 ROS Melodic Build Status

For ROS installation please refer to:

Other dependencies are installed using the script:

chmod u+x

Do not proceed to the next step before you executed this script.

Optional Dependencies

CUDA for stereo reconstruction with plane sweep lib

-> Refer to

Please note, that installing CUDA can sometimes be troublesome. If you are facing an error like

*fatal error: cuda_runtime.h: No such file or directory*

often times adding the CUDA directory to the .bashrc does the trick. If you use CUDA 9.0 for example, you should

echo 'export CPATH=/usr/local/cuda-9.0/include:$CPATH' >> ~/.bashrc 


Linux (both Ubuntu 16.04 and 18.04)

# Create and init a catkin workspace
mkdir -p catkin_ws/src
cd catkin/src

# Clone Open REALM git and compile
git clone

Option 1: Configuration WITHOUT cuda

# Make sure you are in your catkin_ws, not src
# Configure catkin and cmake, blacklist cuda dependent packages
catkin init --workspace .
catkin config --blacklist psl --cmake-args -DCMAKE_BUILD_TYPE=Release
catkin build

Option 2: Configuration WITH cuda

# Make sure you are in your catkin_ws, not src
# Configure catkin and cmake, no blacklisted packages, densifier with cuda
catkin config --blacklist "" --cmake-args -DCMAKE_BUILD_TYPE=Release -DDENSIFIER_WITH_CUDA=True
catkin build

Quick Start

Step 1: Download the test dataset:

Step 2: Unzip the dataset with a tool of your choice, e.g.

tar -xvzf open_realm_edm_dataset.tar.gz

Step 3: We provided as well a set of configuration files in realm_ros/profiles as the corresponding launch files in realm_ros/launch to run the test dataset. The only thing you have to do is modify the path in the launch file:

node pkg="realm_ros" type="realm_exiv2_grabber" name="realm_exiv2_grabber" output="screen"
    param name="config/id" type="string" value="$(arg camera_id)"/>
    param name="config/input" type="string" value="PUT THE TEST DATASET'S ABSOLUTE PATH HERE"/>
    param name="config/rate" type="double" value="10.0"/>
    param name="config/profile" type="string" value="alexa"/>

Note: The exiv2 grabber node reads images and exiv2 tags from the provided folder and publishes them for the mapping pipeline.

Step 4: Launch the pipeline you want to use.


OpenREALM can also be used with a docker. The docker is based on Ubuntu 18.04 and all files related to it are in the docker folder of this main repository. Testing it is very simple:

1.Install Docker from

Docker Install

2.Build the Docker image using the script in docker folder.


3.Run the Docker image using the script in docker folder.


This script can be run from any folder in the host system. The Working Directory will be mounted in the docker. The dataset should ideally be kept in this same folder. Then change the path of the dataset in the launch file as described previously and run the test.


Visual SLAM

[1] Raúl Mur-Artal, J. M. M. Montiel and Juan D. Tardós. ORB-SLAM: A Versatile and Accurate Monocular SLAM System. IEEE Transactions on Robotics, vol. 31, no. 5, pp. 1147-1163, 2015. (2015 IEEE Transactions on Robotics Best Paper Award).

Stereo Reconstruction

[2] Christian Häne, Lionel Heng, Gim Hee Lee, Alexey Sizov, Marc Pollefeys, Real-Time Direct Dense Matching on Fisheye Images Using Plane-Sweeping Stereo, Proc Int. Conf. on 3D Vison (3DV) 2014


[3] P. Fankhauser and M. Hutter, "A Universal Grid Map Library: Implementation and Use Case for Rough Terrain Navigation", in Robot Operating System (ROS) – The Complete Reference (Volume 1), A. Koubaa (Ed.), Springer, 2016.

[4] T. Hinzmann, J. L. Schönberger, M. Pollefeys, and R. Siegwart, "Mapping on the Fly: Real-time 3D Dense Reconstruction, Digital Surface Map and Incremental Orthomosaic Generation for Unmanned Aerial Vehicles"