multirobot-map-merge.md

Evalyn Berleant, Kelly Duan

This FAQ section assumes understanding of creating TF listeners/broadcasters, using the TF tree, namespacing, and launching of multiple robots. This tutorial also relies on the ROS gmapping package although different SLAM methods can be substituted.

Setting up launch files for gmapping

Gmapping will require namespaces for each robot. If you want to use the launch file turtlebot3_gmapping.launch from the package turtlebot3_slam, be sure to pass values to the arguments set_base_frame, set_odom_frame, and set_map_frame to use the robot namespaces. For instance, if the default value is base_footprint, the namespaced value will be $(arg ns)/base_footprint. You can also directly call the node for slam_gmapping inside the namespace (below it is $(aarg ns)) with

<node pkg="gmapping" type="slam_gmapping" name="turtlebot3_slam_gmapping" output="log">
    <param name="base_frame" value="$(arg ns)/base_footprint"/>
    <param name="odom_frame" value="$(arg ns)/odom"/>
    <param name="map_frame"  value="$(arg ns)/map"/>
</node>

Using multirobot_map_merge package

1. First navigate to the ros documentation for multirobot_map_merge. Install depending on the version (newer ROS versions may need to clone directly from the respective branches of m-explore).

2. catkin_make

3. If not using initial poses of robots, simply use map_merge.launch as is. In the initial file, map_merge has a namespace but by removing the namespace the merged map will be directl published to /map.

4. If using initial poses of robots, you must add parameters within the robot namespace named map_merge/init_pose_x, map_merge/init_pose_y, map_merge/init_pose_z, and map_merge/init_pose_yaw.

Using merged map for cost map

The current map merge package does not publish a tf for the map. As such, one must create a TF frame for the map and connect it to the existing tree, making sure that the base_footprints of each robot can be reached from the map, before using things such as move_base.

Saving multiple/namespaced maps

Maps can be saved directly through the command line, but multiple maps can also be saved at once by creating a launch file that runs the map_saver node. If running the map_saver node for multiple maps in one launch, each node will also have to be namespaced.

When using map_saver, be sure to set the map parameter to the dynamic map of the robot (will look like namespace/dynamic_map, where namespace is the name of the robot’s namespace), instead of just the at namespace/map topic.

Example:

<node name="map_saver" pkg="map_server" type="map_saver" 
    args="-f $(find swarmbots)/maps/robot$(arg newrobots)" output="screen">
        <param name="map" value="/robot$(arg newrobots)/dynamic_map"/>
    </node>

will set the map to be saved as the one specified in the parameter, meanwhile the argument -f $(find swarmbots)/maps/robot$(arg newrobots) will set the save destination for the map files as the maps folder, and the names of the files as robot$(arg newrobots).pgm and robot$(arg newrobots).yaml.

To save a map merged by the map_merge package, do not namespace the map_merge launch file and instead have it publish directly to /map, and map saver can be used on /map.

References

• Multirobot map merge package