GPS Yaw Issues.

[yishaiSilver]

I am working on a real robot on humble, and I am seeing several issues:

• My magnetometer isn't being incorporated into the EKF, or perhaps it is incorrectly? If I set the yaw' values of my wheel odometry to be false, there is no yaw at all on the position estimate.
• The starting position as shown by mapviz is always facing East. This is most likely connected to the issue above. I'm not sure how this would be addressed.
• EKF consistently takes way too long. Maybe it's because my GPS is taking too long? I don't know, but I'll see big spikes in CPU usage for EKF nodes (interestingly, it seems to be both global and local nodes that spike, so maybe it's my wheels).

For background:

• My GPS is operating at ~0.5Hz to ~1Hz. I am using a SparkFun RTK solution with correction data coming from UNAVCO. I think the correction data is coming in at a rate of about ~1Hz. I was worried that the closest station may be too far (~40Km), but I have been seeing few issues. In fact, I have mapviz up and running (there were issues with the mapviz launch file: I had to remap /fix to /gps/fix), so I can see that my GPS coordinates are valid.
• My wheel odometry is just fine.

Anyways, I think it's possible much of these issues are regarding robot_localization, so I wanted to ask if there are any sources for using the fuse package instead.

[yishaiSilver]

Nevermind, I misread the sensor_msgs/Imu documentation and set the first index of the orientation covariance matrix to -1 (since I'm not providing roll). A dumb mistake, but it meant my yaw measurements weren't being considered.