Using the odometry from LOAM

[MigVega]

You mention in your paper and the comments of your launch files that one could/should use an accurate odometry system.
However, I don't know the requirements or how to achieve that.

I tried to integrate into one single workspace, A-LOAM, and DLL, and run them in parallel; however, even changing the frame names did not work.

The closest example I found to recreate that was in the mbzirc.launch file. However, it says that the odometry is already in the bag file.
In this case, my question is, how can I save the odometry resulting from LOAM into a bag file so that it is usable by DLL?

I appreciate any help you can provide. I want to compare my method with yours in a large-scale dataset. I believe that DLL is one of the best localization algorithms. Therefore, It would be an interesting comparison!

[fercabe]

Dear Miguel Angel,

DLL follows the usual ROS standards for odometry, that is, a TF that transform the robot base_link into the odometric frame.

You only need to specify the corresponding parameters in the launch files, they are the following:

• base_frame_id: Defines the robot base_link frame name, defualt value is "base_link"
• odom_frame_id: Defines the odometric frame name, the default name is "odom"
• global_frame_id: Define the global frame name, default value is "world". The odometric frame will be transformed into the global frame so that the Lidar sensing information matches the provided map.

You can find a good explanation about how TF is used in localization in the AMCL package, in section 3.1.6.

The mbzirc launch should work straight the way with the parameters of the launch file, if this is not the case, just let me known. We lately performed some updates in the grid map that might affect the initial position of the launch file.

Un abrazo!
Fernando Caballero

[MigVega]

Thanks for the detail explanation!

For A-LOAM these are the corresponding frames, right?

• global_frame_id: "/camera_init"
  -odom_frame_id: "/laser_odom";
  https://github.com/HKUST-Aerial-Robotics/A-LOAM/blob/e51f88c12b483ce42dfa219783712c7234dc8c08/src/laserOdometry.cpp#L513C45-L513C45
• base_frame_id: "/aft_mapped"
  https://github.com/HKUST-Aerial-Robotics/A-LOAM/blob/e51f88c12b483ce42dfa219783712c7234dc8c08/src/laserMapping.cpp#L219/

[fercabe]

Hi Miguel,

The name of the different frames will depend on your sensor setup and system configuration, I would need to have a look at the full sensor pipeline to be sure about the frame names and relation. In principle, looks ok for me.

Fernando

[fercabe]

Dear Miguel, did you solve this issue? Something new to comment?