ros2_map_tools

Requirements:

• PCL e.g. via sudo apt install libpcl-dev ros-${ROS_DISTRO}-pcl-ros
• OpenCV e.g. via sudo apt install libopencv-dev
• Eigen3 e.g. via sudo apt install libeigen3-dev
• Octomap
• python3 modules: cv2, numpy, scipy, yaml

colcon build --packages-select ros2_map_tools --cmake-args -DCMAKE_BUILD_TYPE=Release

pcl_to_map

Squashes 3D data files into 2D binary occupancy maps compatible with the ROS2 map_server.

Supported input formats: .pcd, .ply, .obj, .ot, .xyz

Params

• file: input filepath
• occupancy_likelihood_threshold: probability threshold for occupancy (applicable formats: .ot). Default: 0.5
• resolution: occupancy map resolution. Default: 0.05
• z_min: minimum point z-value for inclusion. Default: 0.0
• z_max: maximum point z-value for inclusion. Default: 1.8
• voxel_min_points: minimum point count per occupancy grid cell for occupancy. Default: 1
• transform: optional comma-separated floating point initial transform as a 6-tuple (tx,ty,tz,rx,ry,rz) or 7-tuple (tx,ty,tz,qx,qy,qz,qw)
• yaw: yaw value for map origin in metadata. Default: 0.0
• tf_ext: optional extension override for transform metadata. Default: follows extension of "cloud" param

Usage

1. In one terminal, load config/pcl_to_map.rviz in RViz, e.g.:

   rviz2 -d config/pcl_to_map.rviz
   

2. In a second terminal:

   ros2 run ros2_map_tools pcl_to_map --ros-args ...
   

3. In a third terminal, iteratively transform the cloud until the occupancy map looks OK:
   • translate:

     ros2 topic pub --once /pcl_to_map/translate geometry_msgs/msg/Vector3 "{x: 0.0, y: 0.0, z: 0.0 }"
     

   • rotate:

     ros2 topic pub --once /pcl_to_map/rotate geometry_msgs/msg/Vector3 "{x: 0.0, y: 0.0, z: 0.0 }"
     

4. To save (suggested to call from third terminal):

   ros2 service call /pcl_to_map/save std_srvs/srv/Trigger
   

   This saves a .png and .yaml pair that can be loaded with map_server (see below). The .yaml also contains the final cumulative transform from map frame to the raw input.
5. To verify (optional):
   1. Enable the "Map" display in RViz
   2. From some other terminal not running RViz:

      ros2 run nav2_map_server map_server --ros-args -p yaml_filename:=/path/to/map.yaml
      

      Note: you may need to also run

      ros2 run nav2_util lifecycle_bringup map_server
      

dense_map_builder

Builds dense maps from timestamp-synchronized nav_msgs/msg/Odometry and sensor_msgs/msg/PointCloud2.

ros2 run ros2_map_tools dense_map_builder --ros-args ...

Params:

• cloud_hz: optional cloud cache throttling frequency if > 0.0 . Default: -1.0
• trajectory: optional trajectory file in TUM format
• odom_topic: optional odometry topic. If omitted, incoming cloud msgs with stamps beyond the range in the trajectory file will be rejected
• voxel_size: voxel grid size. Default: 0.05
• voxel_min_points: minimum points per voxel for downsampling inbound point clouds. Default: 1
• occupancy_likelihood_threshold: probability threshold for occupancy. Default: 0.5
• out_path: outbound binary .ply path. Default: "cloud"

To save the dense cloud from another terminal once all your data has been published:

ros2 service call /dense_map_builder/save std_srvs/srv/Trigger

Note: This will clear the message buffers, unless there are no point clouds available.

An alternative trigger is publishing the optional path to a trajectory file after cloud accumulation:

ros2 topic pub /dense_map_builder/load_trajectory_and_save std_msgs/msg/String "data: /path/to/trajectory/file" --once

map_aligner_ui.py

A "UI" to align >=2 2D maps together, to generate new map metadatas and a final combined map. Future work may be towards automating this, given the assumption of maps that fit together cleanly and are suitable for alignment via feature matching.

ros2 run ros2_map_tools map_aligner_ui.py