Starling v1 GPS Error
-
Thanks for the EKF2 information.
I did end up finding the voxl sdk calibration for the IMU and Gyro. I also was able to use a different wifi dongle that didn't not have a connection issue when I flipped the drone upside down, so I could then do the compass calibration.
Regarding the GPS, what I am seeing if the warnings, but also they are warranted, given that on QGC I can see my phones GPS location and the drone is almost always 15 or more feet away from the correct location. Should I expect at least accuracy within 2-3 feet with the Starling's GPS or no?
And then regarding the position mode failing and forcing the drone to climb, I think when this happens the qvio is at 1% quality, but shouldn't the IMU+GPS+barometer be able to override this? I assume that is sort of what the EKF is aggregating together, but I am not very knowledgeable on this.
Thanks
Sam -
@Sam-Kiley I'm not sure that comparing against a cell phone is valid. Some newer cell phones have dual band (L1 / L5) capability and also may augment with WiFi or other sources. Standard GPS L1 is supposed to give an accuracy within 3m to 5m. But 15 or more feet off from true location seems a bit much. We'll have to do some accuracy testing to see what we get at our test sites.
-
@Sam-Kiley EKF can fuse measurements between GPS and External Vision (EV) (e.g. VIO). However, our parameter sets are separated between indoor parameters (VIO) and outdoor parameters (GPS). But you should be able to enable both at the same time and have EKF choose the best. We just haven't done that yet. With our current parameter sets we expect VIO for indoor use and GPS for outdoor use. Until we can generate and test a unified set of parameters you will have to undertake that exercise yourself if you want to combine GPS and VIO outdoors.
-
Alright, I will look into the parameter sets. Could you point me in the direction of those parameter sets and how you typically swap them on the starling? I think I saw a post about something like that on this forum.
I will also post a screenshot of my GPS error for your reference, as I do think its error is much higher than it should be. 15 ft might have been an understatement.
thanks
-
@Sam-Kiley How many satellites are you getting on average?
-
@Sam-Kiley Our parameter sets are located here: https://gitlab.com/voxl-public/voxl-sdk/utilities/voxl-px4-params/-/tree/master/params/v1.14/EKF2_helpers?ref_type=heads. These can be loaded via QGC.
-
probably around 8 satellites on average. Sometimes going up to 10. I also had difficulty getting any satellites after a reboot a few times. It took multiple minutes to see anything at all.
-
@Sam-Kiley Yeah, that doesn't seem right. In my backyard with some trees around and 2 story buildings I get at least 15, sometimes up to 22. But position does drift around and appears to be up to 15 ft off occasionally but usually within about 6 or 7 ft. But this is testing with VOXL 2 board and M10 GPS near it but not mounted on a drone. So will have to test on an actual Starling to see what I get with that. There is perhaps some interference being generated in it's mounted location on the Starling.
-
@Sam-Kiley Seems pretty clear that the plastic mount holding the GPS unit in place is degrading performance of the unit. We are investigating a new design for the GPS mount to help improve things.
-
Let me know what you figure out. Is your 3D filament something like Onyx, that has some sort of carbon in it?
-
@Eric-Katzfey facing similar issues with our v1 drones. GNSS manufacturer recommends 10cm of distance from ESC/power lines/motors. Is there a location for cad files so we can modify the sensor mount and reprint it.
