Full Map Posterior Grid-based Fast-SLAM

This project extends the OpenSLAM GMapping framework and it's ROS wrapper to support computing Full Map Posteriors besides the Most-Likely Map according to the paper "Closed-Form Full Map Posteriors for Robot Localization with Lidar Sensors" by Lukas Luft, Alexander Schaefer, Tobias Schubert and Wolfram Burgard.

Usage

Installation

Downloading

In a computer with ROS installed, download or clone the repository:

foo@bar:-$ git clone --recurse-submodules https://github.com/joseab10/FMP_gmapping

Compiling

In a command line terminal:

1. Ensure that the ROS environment is properly set

   foo@bar:-$ source /opt/ros/${ROS_DISTRO}/setup.bash

2. Change directory to repository

   foo@bar:-$ cd <dir>/FMP_gmapping

3. Compile using Catkin

   foo@bar:-$ catkin_make
   > Base path: /…/FMP_gmapping
   > Source space: /…/FMP_gmapping/src
   > Build space: /…/FMP_gmapping/build
   > Devel space: /…/FMP_gmapping/devel
   > Install space: /…/FMP_gmapping/install
   > ####
   > #### Running command: "cmake …
   > /…/FMP_gmapping/src
              ⋮
   > [ 98%] Built target slam_gmapping_replay
   > [100%] Linking CXX executable /…/FMP_gmapping/devel/lib/gmapping/slam_gmapping
   > [100%] Built target slam_gmapping

4. Set the ROS environment to the FMP_gmapping workspace

   foo@bar:-$ source devel/setup.bash
   

Running

Full Map Posterior SLAM

To execute the full SLAM stack, including:

• saving debug log files
• generating files for the translational and rotational errors
• computing the Ground Truth and Odometry Maps
• Saving all the occupancy maps
• Generating and saving the Posterior Distribution Maps

a python wrapper for the launch file exist in the map_simulator submodule for running everything.

foo@bar:-$ rosrun map_simulator experiment.py

Alternatively, the launch files from the map_simulator submodule can also be executed normally.

In order to start a launch file, with the environment properly set to the FMP_gmapping workspace, run:

foo@bar:-$ roslaunch map_simulator <launch_file>

where:

• <launch_file> is one of the many files in src/map_simulator/launch/

However, this will not automatically add a timestamp to the results directories, nor will it redirect the ROS debug and log files to the experiment directory.

Test Datasets

The workspace includes several benchmark datasets in the data/ directory, as well as preconfigured roslaunch files in src/fmp_gmapping/launch/ to start using it right away.

In order to start a launch file, with the environment properly set to the FMP_gmapping workspace, run:

foo@bar:-$ roslaunch fmp_gmapping <launch_file>

where:

• <launch_file> is one of the many files in src/fmp_gmapping/launch/

This will start the following processes and nodes:

• roscore: ROS Master process
• gmapping node: Main Grid-based SLAM node
• rosbag play node: Simulation node
• rviz: Program for the visualisation of the step-by-step results

Example:

The following command runs the preconfigured SLAM algorithm on the corrected odometry dataset of building 079 in the University of Freiburg:

foo@bar:-$ roslaunch fmp_gmapping fr_079_corr.launch

Standalone node

In order to use FMP_Gmapping as a standalone node for your own projects, simply run it as any other ROS node with your required remappings and parameters:

foo@bar:-$ rosrun gmapping slam_gmapping

For more details on the specifics, check the README.md file in the slam_gmapping node source directory.