OmniMapper is a modular, multimodal framework for solving SLAM problems. It provides a plugin-based architecture for easy integration of new sensors and feature types.
For more information on the design of OmniMapper, please refer to the publication: OmniMapper: A Modular Multimodal Mapping Framework.
This version of OmniMapper has been tested on Ubuntu 16.04 with ROS Kinetic.
OmniMapper depends primarily on GTSAM 4.0 and PCL. These libraries also have a variety of dependencies (many of which you may already have installed).
Install Cmake, Boost, Git (used by all libraries):
sudo apt-get install cmake libboost-all-dev git
Install PCL dependencies (FLANN and VTK):
sudo apt-get install libflann-dev libvtk6-dev
Install Intel Thread Building Blocks (TBB) and GSL (used by GTSAM and OmniMapper):
sudo apt-get install libtbb-dev libgsl-dev
Install Google Perf Tools (used by OmniMapper):
sudo apt-get install libgoogle-perftools-dev
Clone the GTSAM repository from BitBucket and checkout the most recently tested commit:
git clone https://bitbucket.org/gtborg/gtsam.git
cd gtsam
git checkout 6f8bfe0
Compile GTSAM and install:
mkdir build && cd build
cmake ..
make -j`nproc`
sudo make install
Install PCL 1.7.2 from Github:
git clone https://github.com/PointCloudLibrary/pcl.git
git checkout tags/pcl-1.7.2 -b 1.7.2
mkdir build && cd build
We need to build PCL against the Eigen library included in GTSAM as it includes custom patches. Modify the path below to your cloned GTSAM repo:
cmake -DEIGEN_INCLUDE_DIR="/path/to/gtsam-4.0/gtsam/3rdparty/Eigen" ..
make -j`nproc`
sudo make install
Get the most recent version of OmniMapper from Github:
git clone https://github.com/CogRob/omnimapper.git
From the omnimapper directory, compile OmniMapper and install:
mkdir build && cd build
As with PCL, we need to build against the Eigen library included in GTSAM as it includes custom patches. Modify the path below to your cloned GTSAM repo:
cmake -DEIGEN_INCLUDE_DIRS="/path/to/gtsam-4.0/gtsam/3rdparty/Eigen" ..
make -j`nproc`
sudo make install
If you don't already have ROS installed on your system, follow the instructions for installing ROS Kinetic here.
Also install the following packages for PCL and OpenNI-compliant RGB-D sensors:
sudo apt-get install ros-kinetic-perception-pcl ros-kinetic-openni-launch
Now, clone the omnimapper_ros repository into your ROS workspace:
https://github.com/CogRob/omnimapper_ros.git
Clone the most recent version of CSM (Canonical Scan Matcher) from Github into your ROS workspace:
git clone https://github.com/AndreaCensi/csm.git
We must tell omnimapper_ros where to find CSM. We have provided some config files in the omnimapper_ros repository to make this easier.
From omnimapper_ros/csm_config, copy the cmake folder into the root folder of the csm package.
Next, copy the local-csm-config.cmake file from omnimapper_ros/csm_config into csm/local_config. Rename this file to local-your-computer's-hostname.cmake (find your hostname by running hostname at the command prompt).
Now, from the CSM directory, run the install_quickstart.sh script.
If this is successful, CSM has now been compiled and installed to csm/deploy.
Copy the csm directory from omnimapper_ros/csm_config into csm/deploy/lib/.
Also, copy the libcsm.so file from csm/deploy/lib, to /usr/local/lib.
Now, edit the CMakeLists.txt for omnimapper_ros. Find the line where CSM_DIR is set and change the path to where your CSM is installed, e.g. /path/to/csm/deploy/lib/csm. This tells catkin where to find the cmake config file for CSM.
These steps should allow omnimapper_ros to compile with no issues. (Yes, this is annoying and hacky, and hopefully will be updated to a better installation process sometime in the future.)
Now, from the root of your ROS workspace, run:
catkin_make
You should now be ready to run OmniMapper. See the next section for instructions.
Further documentation available at: https://github.com/CogRob/omnimapper/wiki