Pose-graph SLAM with Open Karto as Front-end and GTSAM as Backend
The optimization routine used is Levenberg-Marquadt provided by gtsam, you can easily change it to something else (Powell's dog leg etc.). This package is currently tested with Ubuntu 16.04 LTS + ROS Kinetic
Use the given launch files. The build_map_w_params launch files uses a given param file, which you can edit to change the behavior of open_karto. Make sure to use the right scan topic. Also, the package will expect an odometry tf to be published to between odom frame and base_link (or base_footprint).
-
Download and install Intel MKL from https://software.intel.com/en-us/mkl
-
Download and install GTSAM 4.x, instructions are in GTSAM readme. $ git clone https://bitbucket.org/gtborg/gtsam.git
-
Download and put open karto in your catkin_ws/src from https://github.com/ros-perception/open_karto
Head on over to my blog post http://sauravag.com/2017/07/an-practical-introduction-to-pose-graph-slam/ where you can ask questions.