Problem while integrating GPS data into robot_localization

Hi to Tom and everybody, I'm currently making some experiments with robot_localization, and I'm having some problems with the integration of GPS data for global localization of a non-holonomic mobile robot constrained to a 2D plane (no Z coordinate and no roll/pitch). Here is the overview of my setup for odometry which is very similar to the one described in [this section](http://docs.ros.org/kinetic/api/robot_localization/html/integrating_gps.html#notes-on-fusing-gps-data) of the wiki where the use of a dual EKF node is discussed.  - X and Y are bidimensional coordinates - Vx and Vy are linear velocities - H*, X** and Y** are angular velocity and linear accelerations, respectively - The topic name is inside parentheses - The provided TF transformation is inside square brackets This is my **launch file** (only the part relevant to replay the bag file): [ false, false, false, false, false, false, true, true, false, false, false, false, false, false, false ] [ false, false, false, false, false, false, false, false, false, false, false, true, true, true, false ][43.542883, 11.571365, 0] [ false, false, false, false, false, false, true, true, false, false, false, false, false, false, false ] [ true, true, false, false, false, false, false, false, false, false, false, false, false, false, false ] [ false, false, false, false, false, false, false, false, false, false, false, true, true, true, false ] And below there is the output of `rqt_graph` which I suppose it's correct, except I don't know why "navsat_transform" is not subscribed to topic "/odometry/imu" even if I specified it with a remap.  [This](https://www.dropbox.com/s/bljxrfje7xgmp52/data.bag?dl=0) is the link to download the **bag file** that can be used with the above launch file. When launched, this is the TF tree I'm seeing with `rosrun rqt_tf_tree rqt_tf_tree` that makes sense to me.  [Here](https://www.dropbox.com/s/6yguf9f9779inja/odometry.rviz?dl=0) you can download the RVIZ config file to obtain the same visualization I'm using. While replaying the bag file, the `odom_link-->base_link` transformation is correctly calculated by the Kalman filter (I'm having pretty good accuracy, too!), but it seems there are problems with the `map-->odom` transformation, because I see these warnings multiple times: [ WARN] [1485356716.681560393]: Transform from odom_link to map_link was unavailable for the time requested. Using latest instead. [ WARN] [1485356718.683456544]: Transform from map_link to base_link was unavailable for the time requested. Using latest instead. Other than that, the estimated global pose start diverging after a few seconds and the coordinates become huge, it is not so clear in the following image, but this is what I see in RVIZ after setting "map" as the fixed frame.  I would like to obtain something similar to what was presented by Tom here, but the GPS data is giving a really unpredictable behavior. Does anybody having these kind of issues with a similar setup? Many thanks, Guido