Merged map is not centered and does not overlap the single maps correctly

[mmatteo-hub]

I am using the noetic version of the package and I am using trying to merge two maps obtained as projection of 3D maps into 2D. This is the launch file just for the map merge algorithm: map_merge.txt I assume to know the robot poses since I spawn them in that position. Each robot has its own /namespace/map and I publish a rigid transform from this to the /map frame to which I center the merged map. This is done via: <node name="$(arg robot1_ns)_MapAlign" pkg="tf" type="static_transform_publisher" args="$(arg robot_1_x) $(arg robot_1_y) $(arg robot_1_z) $(arg robot_1_Y) $(arg robot_1_P) $(arg robot_1_R) /map /$(arg robot1_ns)/map 50"/> <node name="$(arg robot2_ns)_MapAlign" pkg="tf" type="static_transform_publisher" args="$(arg robot_2_x) $(arg robot_2_y) $(arg robot_2_z) $(arg robot_2_Y) $(arg robot_2_P) $(arg robot_2_R) /map /$(arg robot2_ns)/map 50"/>

where `

<arg name="robot_1_x" value="-1.5"/>
<arg name="robot_1_y" value="-0.5"/>
<arg name="robot_1_z" value="0.0"/>
<arg name="robot_1_R" value="0.0"/>
<arg name="robot_1_P" value="0.0"/>
<arg name="robot_1_Y" value="0.0"/>

<arg name="robot2_ns" value="robot_2"/>
<arg name="agent_2" value="2"/>
<arg name="robot_2_x" value="1.5"/>
<arg name="robot_2_y" value="0.5"/>
<arg name="robot_2_z" value="0.0"/>
<arg name="robot_2_R" value="0.0"/>
<arg name="robot_2_P" value="0.0"/>
<arg name="robot_2_Y" value="3.14"/> `

are the same parameter of the map merge file.

How can I solve?

[Pratik-94]

I am facing a similar problem. For some reason the from_map node works alright for the example data provided, but does not work for my data. Can the quality of data be an issue? I also tried varying the parameters like estimation rate etc. but facing the same problem consistently...

[Pratik-94]

@mmatteo-hub can you attach the merged map and the individual maps here?

[mmatteo-hub]

@Pratik-94 I developed a new map merger on ROS2 starting from a pre-existing code. The new code is able to handle N agents knowing their starting position (so the corresponding transformation) wrt a reference frame.