My skid steer bathy boat is all over the place. Can’t keep a steady heading. This is this boat’s third season with scans every month compiling hundreds of kilometers. It’s never had an issue until now. It’s running rover 4.2.3, cube black with dual blended gps (just changed to see if it would make a difference from the previous single Here gps which it didn’t). The last job I was scanning I had to take manual control and finish the last 5km myself. I kept getting velocity variance errors. Today’s log had no errors, but still all over the place. Can anyone see an obvious issue from the log? I’m a copter person, I have no idea lol!
I don’t see any smoking guns. It looks reasonably well behaved, though it does oscillate a bit on the straights. You could turn the steering gains up just a little bit - it undershoots pretty consistently. There’s a lot of noise in the speed output, but that might be a physical limitation and not a tuning issue.
I’m not super well versed in EKF diagnosis, but I took a peek at the innovations, and they seem reasonable as far as I can tell. I also don’t see any one IMU misbehaving as if there’s a hardware issue.
Usually sudden issues like this are hardware failures vs tuning or software issues. I’d like to see a log where the errors do occur.
Incidentally, Rovers don’t really care about TAKEOFF waypoints. You could skip that step in your mission planning.
I also had a quick look at the logs and I don’t see anything too bad. I agree with Yuri’s comments about sudden problems normally being caused by hardware changes.
It might help a little to set the GPS_POS_XYZ parameter values so that when the vehicle rotates the EKF can compensate for the GPS moving.
I’ll set the offsets and see what happens. I’ve changed nothing from the original build. Followed the tuning documentation and everything worked great! It’s probably done 500km without a hitch until now.
Like I said, I’m a copter user lol! Auto waypoint and off we go. I’ll try increasing the steering gains tomorrow.
This log is a bit more pronounced, but no velocity error on this one. I’m sure this is related to the velocity variance error I had previously as the telemetry for ground speed is all over the place. I usually keep a close eye on speed as my max hull speed is about 3.4 knots. I try to stay around 3.2, anything over and I’m doubling my current. Before this problem started, the speed would vary maybe .1 or .2 knots. This of course depends on gps quality, but it was always pretty steady. Now the ground speed jumps all over from 1.5 to 4, but looking at the boats it looks like it’s going the same speed lol! I wish I had some of last month’s scan logs, but I had turned logging off being I had a bad card.
Happy boat. On track and straight as an arrow. I changed three things all at once, not great for debugging a problem. I went from firmware 4.2.3 to 4.4.0 beta. Input my IMU position and the position of gps1 and gps2. The third was changing EK3_ABIAS_P_NSE from .02 to .003. I don’t know if this value was because of the firmware update, but looking at a previous tlog I loaded, the previous parameters indicated EK3_ABIAS_P_NSE was at .003. I assume EK3_ABIAS_P_NSE .02 is the new default. Here’s this long from tonight’s flight (boating session lol).
for 4.3 and up. Anyhow, everything is back to normal. The indicated speed doesn’t jump around anymore as well.
1 Like
Hi @X1aero1,
Glad that you’re making progress.
We did change the EK3_ABIAS_P_NSE parameter from 0.003 (in Rover-4.2) to 0.02 (in Rover-4.3 and higher). This is the PR that made the change:
The PR does talk about EKF velocity issues when the vehicle is stationary for long periods of time and when using an RTK GPS. It looks like you’ve got a very fancy GPS on that vehicle.
The PR listed above is suppose to fix the issue and is included in 4.4 so I suspect it was the firmware upgrade that resolved the problem. I also suspect hat setting the EK3_ABIAS_P_NSE to the default of 0.02 will be fine.
EDIT: actually I guess the fancy GPS in the first picture is not used by the autopilot but perhaps only used by the sonar? I guess though that adding another GPS may have slightly improved the GPS quality and triggered the same issue (mentioned in the PR) that could be caused by an RTK GPS. This explanation is all a bit hand-wavey and trying to explain after the fact but perhaps it’s correct.
Glad also to see progress and continued discussion. I’m afraid I don’t have much to add regarding the EKF analysis, but I hope to learn from any further findings. It certainly seems logical to raise what I assume is a noise floor parameter.
1 Like
The big gps antenna is a Novatel pinwheel receiving rtk, but it’s just for the sonar. Navigation is handled by two blended first gen here gps pucks. Originally there was only one here gps, but I thought maybe two would help😂 Of course it didn’t make any difference.