How to use SLAM & dual EKF method?
I have setup my robot with a single EKF instance fusing wheel odom + IMU, and it's working pretty well. I also have a GPS setup that is accurate to ~3cm. RTAB provides a gps input to assist with loop-closure, which works well, but with our accurate gps I think our system could be improved with the dual ekf (and is seen in clearpath's husky).
It's a similar setup to Clearpath's Husky, where there is both dual EKF and SLAM (they use gmapping). You can see here that they have gmapping set to provide map->odom_gmapping, while their dual ekf provides the map-> base_link and odom->base_link transform. I am confused by how this tf tree would work, as their local and global costmaps are in the odom frame, and I haven't found where else the odom_gmapping frame is otherwise used. If it's not used at all, what is the point of broadcasting the map->odom_gmapping transform?
So the question is how does the separation in odom and odom_gmapping work? My issue right now is both the gps-included EKF and RTABmap would provide map->base_link.
