Anders, Looking at your log file, it is apparent that your GPS is not happy. You have repeated status drops to 1:
I would look at your GNSS antenna connection/placement and make sure everything is secure and that the antenna has a clear view of the sky and that the antenna and module are both away from any source of interference.
I can also see that the number of satellites in use at any given time is typically 6-7 with as few as 4. I see that you are using the Swift Piksi. We have a couple of the Pkisk Multi’s as well as the Duro Inertial. I presume that your Piksi is limited to the GPS constellation only? Even with only one constellation, you should be getting a better result than what you are getting. Also, What correction source are you using and how are your corrections being sent to the rover? The SBPH entry in the log indicates that there are many CRC errors in the raw data injection messages. This indicates that the correction data is only intermittently reaching the GPS.
Despite your very intermittent GPS results, the EKF seems to do OK most of the time however, there are a number of critical errors reported in the log file. These will correspond with your vehicle stopping during AUTO missions. The vehicle did not stop due to lack of RTK fix (as this happened literally 100+ times in this log…) The failsafe errors seem to have occurred due to the EKF INAV not being happy. The errors seem to correspond with the periods of large yaw innovations.
I haven’t been concerned with vibration levels since I was using Arducopter more than 5 years ago, so I’m not sure what vibrations levels are acceptable on rover. Unfortunately I do not have access to any log files at work to check, but at first glance, your vibration levels look quite high. This may be contributing to the EKF issues.
I am not sure what your use case is or how close to the path you need your rover to stay, but if you want the rover to stay closer to the path after turning then I can give you some tips.