Would this GPS behaviour perhaps be a receiver specific thing? Especially if yours is designed for ag use?
I donât know. Andreas said he noticed the same thing. Which is why heading with dual is kind of immune to the rocking back and forth.
A tiny bit of kalman filtering though made a huge difference.
Are you both using the CMPS14? Would be interesting to replicate your test with the BNO085 to see if it is similar / worse / better. Even different modes to see the effect they have.
or, to keep the library standard, <your IMU name>.begin(0x4A) in your setup code.
So this is using the outback receiver? Iâm pretty sure the F9P does not do a similar kind of kalman filtering on the GPS position, at least not in my limited experience.
1 Like
I would have thought for drone use etc a feature like that would be the last thing you would want! Hence why I wondered if @BrianTee_Admin receiver has application specific features.
It might explain why some of us have had roll related issues that some developers havenât experienced. The MMA was more of an issue to me than my drifting BNO055 most of the time.
Hello,
Iâve eddited my previous autosteer .ino for BNO08x: now you can choose between Rotation Vector and Game Rotation Vector:
Sorry Brian, I donât understand: what is âeastingâ and âroll corrected eastingâ ?
And what are âFixâ and âCorrâ in your graph ?
@Larsvest interesting: I never succed to calibrate magnetometer in the steering box in my tractor cab while doing the 180° turns. Perhaps I should do it one before out of the cab and one after.
A thing I was thinking while making all my test: could it be possible to add to AOG a data logger feature, to logs data usefull for our test, in order to proceed them out of th tractor with a software like Matlab or Scilab, or Excel ? A simple text file with data logged into him, separate with coma.
If we faced to a problem, it could be easier to analyse data out of the tractor cab.
Math
1 Like
How would you know ? Without looking at high speed and seeing the point to point corrections vs roll from something like the CMPS, it is impossible to just see.
Well the CMPS accurately portrays the roll occuring. I set up a 6x6 and drove over it and stopped at the top. measured the angle, then did the same while moving over it. the cmps showed exactly what i drove over.
Northing and Easting. LAtitude and longitude are degree based point represenation but northing and easting are an XY grid representation without the curvature. Think of small grids placed over a basketball. northing is the Y axis and easting is the x axis. Now if i drive exactly straight north, then the easting value represents the roll back and forth on the antenna. If perfectly smooth, the easting would show a straight line. The Green line shows the actual points coming from the antenna. The orange line shows the sin of the roll times the height to show a âcorrectedâ value that eliminates that side to side effect of the bumps and sidehills.
It all happens very quickly and considering only a cm or two of accuracy as a corrected result on an antenna moving 10 to 30 cm back and forth is truly amazing.
https://en.wikipedia.org/wiki/Universal_Transverse_Mercator_coordinate_system
Sorry I wasnât clear. When you speak of the GPS having its own bit of filtering, are you talking about the Outback RTK receiver (thatâs designed for ag of course so filtering wouldnât be surprising) or are you using the F9P?
Last time I fed the F9P straight into my tractorâs commercial system without any roll compensation at all, it did not seem like the F9P was filtering anything. Every little bump on flat ground, even quick one, would cause the track to wiggle. But if you have found the F9P does a bit of filtering out of the quick bumps, then that is great!
To clarify my understanding of the above: Is back and forth a general term for right left or is it actually back and forth?
âŚwell it is until you have accurate roll figures then you canât factor out roll as efficiently because it is tied to an inaccurate (filtered) fix.
I would have never even noticed it if the roll correction with the CMPS wasnât so good. if the bump was âslowerâ then it was the same amplitude, but quicker and you could just see the point to point limitation. Iâm putting up the video right now on youtube.
Without the capability of graphing, it would be impossible to see if there is any filtering.
1 Like
back and forth would be considered local coordinates whereas northing and easting are world coordinates. Terms like back and forth are pretty vague and i should not use them. swaying back and forth would be roll in this case. Rolling left to right is roll as well.
Hello @BrianTee_Admin
Ok thank for your explanation. UTM is the projection system used by AOG ?
So if I understand well, if you drive straight north, on a perfect flat surface, easting value will be a constant, like this:
Now, if you always drive straight north, but not on a flat surface (a surface where your tractor will roll left and right): theoretically, you should still have a constant for easting value, but du to the left/right roll of the antenna, you will have a oscillating value of Easting like this:
And what you does is using the roll angle from the roll sensor to correct the Easting value du to the roll of the antenna ?
Did I understand right ?
But in case of you are not driving straight to the north, Northing and Easting value shall be corrected with roll, right ?
Otherwise, about the use of the BNO08x, I was thinking about the use of Rotation Vector.
Adantage of using Rotation Vector is to have a heading normaly referenced to the north. But is it really useful for AOG autosteering ? Because, there is the following drawback with Rotation Vector:
- Itâs painfull to calibrate the magnetometer, specially in tractor environment
- We could be subject to âjumpâ of heading as I encounter (but I still should try to confirm if itâs caused by a non calibrated magnetometer, background calibration of magnetometer, or just du to the use of magnetometer
On another side we can use Game Rotation Vector (that donât use magnetometer). Advantage is that:
-
We only need to calibrate accelerometer and gyroscope, which is really easier.
-
In this mode, accuracy annonced by Hillcrest Lab is better in this mode:
The only drawback is that: -
Heading is not referenced to north: is it really a drawback ?
-
A drift of 0,5°/min in heading is annonced. As I experencied in this mode, it doesnât seem to have any influence on the guidance, and for @Alan.Webb too.
What do you think ?
@BrianTee_Admin would you do a magnet test, to check if the output that you use from your cmps, is of the rotation vector, or game rotation vector type
Iâve emailed Robot-Electronics UK about it. See what response I get! 
No effect with a magnet. It always starts out in random heading. Currently this isnât an issue as the heading of the imu can be anything and gets corrrected to match the gps heading. You do need to drive a bit initially from cold start.
No mag, less problems.
2 Likes
So the CMPS14 isnât using the magnetometer at all then? So it should be equivalent to the BNO085 directly in one of the modes that also doesnât use it. Or am I getting that wrong?