I've been puzzled by the concept of dual EKF for days, and really appreciate it if someone could kindly help me out. I checked this post about dual ekf, but I am still confused. Thanks.
In robot localization package's wiki section Integrating GPS Data, we have
If you want to fuse data from a GPS into your position estimate, one potential solution is to do the following:
Run one instance of a robot_localization state estimation node that fuses only continuous data, such as odometry and IMU data. Set the world_frame parameter for this instance to the same value as the odom_frame parameter. Execute local path plans and motions in this frame.
Run another instance of a robot_localization state estimation node that fuses all sources of data, including the GPS. Set the world_frame parameter for this instance to the same value as the map_frame parameter
In the context of using robot location package to fuse IMU and GPS, my understanding is that the wiki recommends us to use two ekf nodes. the first one fuses only IMU, and the second ekf fuses GPS.
At the end of the day, the localization module is supposed to continuously output state estimation via the ekf_localization_node 's published Topics (odometry/filtered and accel/filtered), so the two ekf nodes will publish their own separate messages via these topics. My bewilderment is what happened to the "fusion". Specifically
- Are we supposed to post process the messages from the two ekf nodes, and fuse them somehow?
- Or maybe there is already some interactions occurred between the two ekf nodes, and the messages they output are already the fused result, except the first one report the result in odom frame, and the second one report the result in map frame? If this is the case, how the two ekf nodes interact with each each, and get IMU and GPS data fused?
A note that might be of importance, currently my understanding is that both odom frame and map frame are world fixed frame, and the transform between them are fixed once defined. The odom frame refers to the starting position of the robot, while map frame refers to the origin of the map.
Post score: 1