Best Practices for Connecting cuVGL and VSLAM Nodes

I have a question regarding the communication between cuVGL and VSLAM. As I understand it, I should create a node that subscribes to the visual_localization/pose topic (published by cuVGL) and republishes it to visual_slam/initial_pose, so that every successful localization triggers a new initialization in the map. In this case, do I need to deactivate the publish_map_to_base_tf parameter (cuVGL), or is the correct adjustment of the TF tree not yet implemented in the visual_slam/initial_pose process?

The logic for when to trigger localization should be implemented in this connector node — for example, depending on the VSLAM status, such as when the VSLAM pose jumps too much or loses its position in the map?

Any advice would be appreciated.

vslam


cuVGL

Isaac ROS Visual Global Localization — isaac_ros_docs documentation

isaac_ros_visual_slam — isaac_ros_docs documentation