cartographer-tuner 0.1.2

A toolkit for Google-Cartographer tuning.

Cartographer Tuner

Helper package for Cartographer tuning.

Installation

pip install cartographer-tuner

System Requirements

This package requires the following external tools to be installed (for some functionality):

• cartographer - Google Cartographer SLAM library
• cartographer_ros - ROS 2 integration for Cartographer
• rviz - ROS2 integration for RViz

Dependecy installation:

sudo apt-get install \
    ros-${ROS_DISTRO}-cartographer-ros \
    ros-${ROS_DISTRO}-cartographer \
    ros-${ROS_DISTRO}-cartographer-rviz

PGM Map metrics

The tool implements metrics presented in 2D SLAM Quality Evaluation Methods.

Example

• pgm-corner-count /path/tp/map.pgm
• pgm-occupied-proportion /path/to/map.pgm
• pgm-enclosed-areas /path/to/map.pgm

Map building

The tool may build 2D map out of rosbag file via cartographer.

Note: rosbag file must contain topics named according to cartographer requirements: /imu, /scan, etc

Example:

lua-to-pgm \
    --bag_filename=/data/bags/2011-01-28-06-37-23 \
    --configuration_directory=/opt/ros/humble/share/cartographer_ros/configuration_files \
    --configuration_basenames=mit_stata.lua \
    --map_filestem=/data/maps/2011-01-28-06-37-23_test 

Configuration evaluation

One can combine map building and evaluation in one step.

Example:

lua-pgm-metrics \
    --bag_filename=/data/bags/2011-01-28-06-37-23 \
    --config_dir=/opt/ros/humble/share/cartographer_ros/configuration_files \
    --config_basename=mit_stata.lua

Configuration optimization

One can optimize their configuration using metrics described above and one of available optimizators:

• Grid Search:

  config-grid-search \
      --bag_filename=/data/bags/2011-01-28-06-37-23 \
      --config_dir=/opt/ros/humble/share/cartographer_ros/configuration_files \
      --config_basename=mit_stata.lua \
      --grid '{"map_builder.pose_graph.constraint_builder.ceres_scan_matcher_3d.translation_weight": [0.1, 1, 10, 100], "map_builder.pose_graph.constraint_builder.ceres_scan_matcher_3d.rotation_weight": [0.1, 1, 10, 100]}' \
      --metrics corner_count,enclosed_areas_count,occupied_proportion \
      --output=/data/ceres_weights_metrics.csv