Driving on a side-hill: problem with the roll (sidehill compensation)


I’m currently seeding sugar beets. My seeder is a 12 rows, 45cm spacing. So ideally, between 2 pass, the space shall be 45cm. But, as AOG is driving with +/-2cm accuracy, between 2 pass, I shall measure 45cm +/-4cm.

I’have seed in a flat field, around 1° or 2° of tilt. If I disable the roll in AOG, everything works perfectly. Between 2 pass, I always measure 45cm +/-4cm.

But if I enable the roll, it doesn’t work at all: between 2 pass, I’m always under 43cm, more often at 40cm and sometime less.
It’s not a autosteer problem: on AOG I’m on the line. But the line “is not correctly positionned” compare to real world.
To illustrated, some pictures, with roll enable:

  • On AOG, I’m on the line:
  • In real world, the tractor is shited from the seed drill plotter (in this case space between this space and the previous was less than 40cm):

If I disable the roll, everything go back to normal work: when I’m on the line in AOG, I’m align with the seed drill ploter. Space between 2 pass is 45cm.

So it’s look’s like when the roll in enable “the line in AOG is not in the correct place in the real world”.

So correct me if I’m wrong: in fact in AOG the line is always at the same place (every 5,4m in my case): it’s the mesured GPS position that is corrected by the roll value.
A schematic of a tractor on a side hill:

  • The tractor real position is point A
  • The GPS position mesured by GPS du to hill is point B

So mesured GPS position shall be corrected from the distance L between point A and B. L is calculated with tilt measurement angle alpha. Then when we know L, we can corrected tractor measured position B to obtain real tractor position A.

So in my case it’s look like the tilt compensation calculation: tilt is overcompensated.

But I have checked 10 times:

  • Antena height correctly configure in AOG
  • IMU zeroed on flat surface
  • All other tractor dimentions
  • All tools parameters also: but as it work without roll, it’s confirm that it’s correctly configured

So for now, roll is disable, the first part of the field I’m working in as a little tilt: 1-2°. So the error du to tilt seems negligible: positionning of the tractor in real world is bette without roll.
But the tilt of the field is going to increase, I things to 5-6°. Not neglitible, so I will have to enable the roll.

So anyone have an idea of what’s going wrong when I enable the roll ? Does I miss something ?

Thank for your help.


You will probably be able to correct me pretty fast, but if line L is the correction amount for roll, then the error will grow with an increase in sidehill. Shouldn’t L be horizontal to sea level so correction is tangent to horizontal ? I noticed the same thing 2 weeks ago when I was strip tilling, and thought I probably have something incorrect in my setup, but all measures well also.


Perhaps L shall be parallel to sea level. Don’t exactly know. Perhaps someone who know exactly how it works can confirm ?
But, yes whether the line is as drawn or parallel to sea level, the distance between “real tractor position” (A) and “Mesured tractor position” (B) will increase with the increase of tilt angle.

So you mean than: when roll is disable, AOG giving a error of 0, in the real world your tractor is correctly positionned compare to your strip till. When roll is enable, AOG giving a error of 0, in the real world your tractor is not correctly positionned compare to your strip till ?


Following it perfectly across slopes will be difficult because the gps create parallels on a horizontal plane but if but projecting this horizontal plane on an inclined plane (the ground) the distance between the lines increases it is necessary to compensate for the roll of the tractor take into account that the fact of being on a slope increases the ground surface and the distance between two parallels and reduces accordingly the projecting width of the tool I could not explain it mathematically but draw on a material that is loose and flat several parallel then create a bump or a sag in the center of it the deformation will be obvious

Yes, on an incline that is measured with true horizontal, the middles would have to be wider than the row spacing, but can you do the math according to the roll ? Because, who cares if the row spacing is correct from the sattelites, we don’t want wide middles with weeds.

No, you are correct. If it is horizontal, then a 5 degree inclination with 120 inch antenna height would be 10 inch error . Yes ?


So I re-test again the roll compensation while seeding in a tilt field.

Conclusion : it’s finally works good. I was always between 2 pass at 45cm +/-4cm with roll enable.

Some measurement I’ve done:

  • To check the roll measure:
    On the length of the field I put in AOG 4 flags, in area with various tilt. During 4 pass, each time I pass near the flag I record the tilt value given by AOG. The results:
    So tilt measurement was ok: in one direction and in the other the values are the same with the sign reversed. it confirm that the IMU was correctly zeroed.

  • Then check the row space between 2 pass on 5 successive passes (3 or 4 measurements randomly on the field lenght)
    Between pass 1 (No roll) and pass 2 (roll), I made 2 measures and obtain 34cm & 37cm → result logical, du to the activation of roll.

Between pass 2 (roll) and pass 3 (roll), I made 4 measures and obtain: 45cm, 50cm, 47cm, 44cm.

Between pass 3 (roll) and pass 4, I made 4 measures and obtain: 45cm, 45cm, 43.5cm, 43,5cm.

Between pass 4 (roll) and pass 5, I made 3 measures and obtain: 42cm, 43cm 47cm.

The field was finished (tilt grows up to 7°), I continue to randomly check the raw space between 2 passes: I always have a measure of 45cm +/-4cm.

So the tilt compensation work perfectly ! Really happy of the reasult of the sugar beet seeding with AOG !