Right value of magnetic_declination_radians

edit

Thanks Tom for the answer.

If the declination is around 0.152425 radians, would you expect that the EKF yaw to have a 0.152425 offset relative to the IMU yaw?

I am still struggling with these concepts and I hope to get robot_localization working outdoors soon. What kind of error would you expect using an Ardusimple RTK GPS (cm accuracy) with an XSens MTi-630 IMU? I calibrated the IMU but I can't get repeatability on the results.

No, declination has nothing to do with the EKF. It's solely a concept that applies to navsat_transform_node. It's just used to correct your raw IMU data (which points to magnetic north, and not true north) in navsat_transform_node. The EKF's heading is utterly independent of anything in navsat_transform_node. In fact, your EKF might not even fuse IMU data. It might just fuse yaw velocity from your wheel encoders.