GPS discrepancy in flight causes erratic behavior (Emlid RTK)

I just performed the first test flight of my X8 octocopter while using the Emlid Reach M+ RTK. I had a good RTK fixed position and started a simple waypoint grid mission. The first flight line seamed fine but then during the second line the drone violently rolled momentarily. This happened a few more times before I abandoned the Auto mission. There did not appear to be an issue flying in Pos Hold mode, only Auto mode. I see from the logs that there is a large skew of the POS parameter during the erratic movements. I am using a Here2 as GPS 1 and the M+ as GPS2. Any ideas?

Mine does exactly the same when it looses the rtk fix. It violently changes gps position. That is why i stopped using l1 rtk receivers. I know some people have had good results but i really struggle to use emlid l1.
Since i shifted to an f9 gps i have a strong fix and a lot less probs.

Thanks for the insight. I am wondering if it would help to move the M+ to the GPS 1 position and the Here2 to GPS 2.

I sent you a PM regarding your F9 GPS. Can you share your setup? Thanks

I had similar issues with combo of a Reach M+ and Drotek M8P (both RTK). The problem went away when I set it to just use one GPS. From your log I noticed you have GPS_AUTO_SWITCH = 1, “Use Best.” It may be better to set it to 0 to use GPS1 or 3 to use GPS2. You also have no GPS offsets, which may cause issues when switching from one GPS to another.

I took another look at the log and there is a large difference between the positions reported by each GPS and it is switching between the two. It looks like it is using GPS2 for most of the time but had some moments where it switched to GPS1. GPS1 is consistently reporting a lower HDOP, but when GPS2 has an RTK fix (status 6) that is recognized as the better position. GPS2 status is switching between 6 and 3, where as GPS is consistently at 4. The M+ isn’t going between status 5 and 6 (float and fix) but from 6 to 3, which may indicated a problem with sending corrections rather than a problem keeping fix.

Thanks for the info. When you say GPS offsets what do you mean exactly? I do not see a parameter that relates to this.

http://ardupilot.org/copter/docs/common-sensor-offset-compensation.html