Hello. As the title suggests, I am noticing a strange behavior with the GPS values whenever I enter the Auto plot application on the Amiga Brain. First, there seems to be some weird drift or otherwise shift to the rover’s position in the application even when not moving. Note, my GPS was fixed with an accuracy around ~1cm, with me changing the GPS threshold in the settings to about 2cm for testing reasons. The filter also stated that it was converged, and I previously changed the dimensions in the robot geometry section and calibrated it to accurately reflect GPS and IMU positiom.
Here were the GPS PVT and relposned values i took moments before booting up the app:
Another phenomena that coincides with the drift in robot position is a drift in GPS values. I can drive around a whole soccer field with no problem, getting accurate and consistent updates to my GPS feed for long periods of time un-interupted. However, when I boot up the autoplot app, the gps values start spiking upwards, eventually diverging and then stating that it “lost RTK corrections”.
From what I’ve seen on the guides, the only thing I can image being incorrect are my measurements for the GPS and IMU. However, if there are any other settings or other pieces of information I am missing any help would be greatly appreciated.
Hey Henry,
thanks for your question. So, autoplot breadcrumbs drifting off are usually related to miss-calibrated IMUs.
So, if you are okay with the measurements (make sure you also got the units right. if it is showing meters and a untoogled button, your measurements should be in meters), can you try to go to Settings > IMU calibration and, while the robot is in a flat position and fully stoped, you click to calibrate?
Thanks for the reply Marcello!
I have double-checked all measurements, making sure they were accurate and drew everything out to make sure there were no gaps in my measuring logic. I did make sure I was on flat ground, fully stopped when I calibrated these settings.
I have also tried mirroring all the values (swapping the sign) in-case i got the direction of “front” messed up. None of these solutions have worked unfortunately, and I have the same problem of GPS values diverging. Does the update rate of the GPS coordinates have to be faster for that app than they are? Currently they are updated about every second. The only other thing I can think of is maybe the OAK’s or some other physical configuration of the robot is bad?
(OAK0 is the one facing the camera)
I see, just to make sure we are talking about the same calibration here… did you check the process of IMU Calibration?
Here’s a screenshot:
Yeah, I adjusted the robot geometry, then hit the Start Calibration button, and let it finish.
okay. That seems odd. I will send you an email to set up a time to watch the behavior and do some debugging.
1 Like
After some quick experimenting using a Emlid RS2+ as my basestation, my results were much better. I could successfully run the autoplot app without the GPS diverging like it was before. I cannot prove it, but I am certain that the issue was with the clash between GPS refresh rate and how the unscented kalman filter was sampling old GPS data, causing a wide probability distribution in the model. I will mark this as the solution for this thread, but I am curious about one final bug.
For some reason, even with an accurate filter, accurate GPS, accurate geometry offset, active WiFi connection, and auto control being on, I could not follow paths.
When trying to click on why the button was greyed out, the text box was empty. I am wondering if there are any more steps to get the follow track working that I am missing? For example, does the rover need to be over the starting position and at the proper orientation to start following track?
(The orange line in the bottom of the photo is a path generated by a script of mine, while the middle photo is one tracked and saved by the autoplot app itself)
Hey @henrydalrymple53,
glad to hear the RTK worked fine. It looks like he frequency of the public service (maybe affected by the distance?) was indeed impacting your accuracy. I think it make sense that, with corrections taking too long to be received, the error starts compounding and impact accuracy. Thank you for pointing that. I will document it.
Now, about the requirements to use autoplot, you need to be within 80 cm of one point of your path and heading the way you want to go. The paths are bidirectional, but autoplot won’t turn the robot to the right direction.
btw, you can change that 80 cm threshold at Settings → Robot Localization → Path Deviation Threshold.
Thank you very much! Amiga autoplot and track following works, now I just need to fix my path generation script. Again, thank you for the effort and time!
great to here! Thanks for letting us know… enjoy!
